WO2025221685A1 - Bioreactor systems and methods for the use thereof - Google Patents

Abstract

Provided herein are methods for manufacturing and using engineered tissues. The methods and compositions described herein may be useful for biocomputing, regenerative medicine, non-clinical trials, or cellular agriculture.

Description

BIOREACTOR SYSTEMS AND METHODS FOR THE USE THEREOF

CROSS-REFERENCE

[0001] This application claims the benefit of U.S. Provisional Application No. 63/634,033, filed on April 15, 2024, which is incorporated herein by reference in its entirety.

SUMMARY

[0002] In certain aspects, described herein is a bioreactor system comprising: a first module comprising a plurality of bioprocess controls; and a second module comprising a system for electrical stimulation and electrical recording, wherein the first module is isolated from electrical stimulation. In some embodiments, each electrical component is grounded. In some embodiments, the electrical component comprises at least the pump, for the addition and removal of liquids, and pressure control. In some embodiments, the first module is physically isolated from the second module. In some embodiments, the first module comprises a first chamber and the second module comprises a second chamber. In some embodiments, the first module comprises an incubator and the second module comprises a bioreactor placed within the incubator. In some embodiments, the second module comprises a cell culture or a tissue culture plate. In some embodiments, the plurality of bioprocess controls comprises at least one of a pH control, a dissolved oxygen control, a temperature control, a pressure control, and control of gas, liquid, and solid components added and removed from each reactor. In some embodiments, the bioreactor system comprises a system for additive manufacturing, subtractive manufacturing, or a combination thereof. In some embodiments, additive manufacturing comprises bioprinting and injection molding of biomaterials. In some embodiments, the system for additive and subtractive manufacturing comprises a scaffold material and a plurality of cells. In some embodiments, the scaffold material is biocompatible. In some embodiments, the plurality of cells comprise a neural progenitor cell, a stem cell, a primary tissue cell, a differentiated neuron, an astrocyte, an oligodendrocyte, a t-cell, a vascular cell, or a combination thereof. In some embodiments, the second module comprises a system assembly, growth, and control of tissues by additive or subtractive manufacturing. In some embodiments, the second module comprises an electrode array. In some embodiments, the bioreactor system comprises both a three-dimensional electrode or microelectrode array for recording and stimulation of engineered tissues. In some embodiments, the bioreactor system comprises both a surface and three-dimensional electrode or microelectrode array for recording and stimulation of engineered tissues. In some embodiments, each electrode is configured to stimulate and record electronic cellular messaging. In some embodiments, the bioreactor system comprises at least 1 port for flow of liquid into the bioreactor. In some embodiments, the liquid comprises growth factors, nutrients, metabolites, stabilizers, pH indicators and controllers, living and non-living components . In some embodiments, the system for electrical stimulation comprises at least one electrical component comprising at least one of a pump, a microcontroller, or controller, a probe, and a microelectrode array. In some embodiments, the bioreactor system comprises: a module comprising a data processing and control center; a module comprising a biological to digital decoder; a module comprising a neural signal preprocessing block; a module comprising a digital-to-biological encoder; a module comprising AI/ML assisted postprocessing; a module for additional ML training and offloading; a module for local data storage; a module for cloud data storage; a module for CMOS-based biocomputation; or a combination thereof. In some embodiments, described herein is a method of using the bioreactor system described herein, wherein additive manufacturing is used to directly biomanufacture tissues into said reactor. In some embodiments, the bioreactor system allows stimulation and recording of engineered tissue electrophysiochemistry signals through isolation and reduction of background electrical noise. In some embodiments, engineered tissue viability is maintained by way of active perfusion, or passive perfusion in which avascular engineered tissues of diameter greater than 1000 microns, are maintained through the use of porous biomaterials. In some embodiments, engineered tissue viability is maintained by way of active perfusion, or passive perfusion in which avascular engineered tissues of diameter greater than 1000 microns, are maintained through the use of tissue-free spaces, such as avascular channels for media transport, by way of additive or subtractive manufacturing perfusion. In some embodiments, one or more digital modules serve as a digital twin to a plurality of biocomputation parameters, allowing for the biological information to be stored in a non-biological substrate. In some embodiments, the digital twin is used to train novel biological components in a biocomputation context. In some embodiments, model features are directly transferred from one biological module to another without routing information back to through digital components. In some embodiments, the biocomputing system comprises a neuromorphic platform that integrates living biological components with reconfigurable digital hardware. In some embodiments, an automatic conversion layer translates high-level, python-like code to perform neuromorphic tasks that comprise: in-silico functionality; in-vitro functionality; in-vivo functionality; ASIC wavelet transforms; dynamic neural signal ingestion; Al post-processing; biological-to-digital (ADC) conversion; digital-to-biological (DAC) conversion; Al preprocessing; or a combination thereof. In some embodiments, a module stores biological data including spike trains and synaptic weights, and runs on-chip ML models to extract neural data features and perform predictive analysis while reducing latency and energy consumption. In some embodiments, animal or human neural tissues are decellularized, recellularized with new cells, and then used for biocomputation. In some embodiments, a module performs real-time task redistribution in neural tissue computation, wherein workload, health, and cohesion metrics are continuously assessed to dynamically reallocate tasks, optimizing resource utilization while maintaining tissue stability. In some embodiments, neural computation task migration is performed based on spike features, gamma levels, complexity scores, and connectivity measures. In some embodiments, a Task Table is made wherein active sub-tasks are tracked with cluster assignment, priority, progress state, and/or workload metrics. In some embodiments, training and inference can simultaneously occur across silicon and biological substrates. In some embodiments, a module automatically converts between alphanumeric formats and neural signals in one or two directions. In some embodiments, a module trains biological components to execute basic scripts by associating specific input patterns with desired output responses. In some embodiments, digital data is automatically converted into neural spike trains and other signals using encoding schemes compatible with neural processing. In some embodiments, neural spike trains and other neural signals are automatically converted into digital data. In some embodiments, data is sent to biological components such as neural cells or tissue(s) in a compressed format. In some embodiments, data is decompressed in biological components such as neural cells or tissue(s). In some embodiments, a module performs matrix operations. In some embodiments, a module performs matrix operations within biological components such as neural cells or tissue(s). In some embodiments, a biocomputer is used for autonomous biological/silicon drone applications. In some embodiments, the biocomputing system provides initial bias to prevent inadvertent stimulation or sensing or provide gating threshold function. In some embodiments, any electrode can be connected directly to any other to force specific neural paths in thinking. In some embodiments, neurons can control a switch to provide new selfgenerated hybrid thinking routes. In some embodiments, neurons can control functions outside of their specific well, outside a given module, and/or outside the system. In some embodiments, delays in action potentials is leveraged to achieve specific thinking mechanisms - circuit equivalent of delay line. In some embodiments, electronic circuits provide digital logic type functions in a biocomputation context. In some embodiments, a through silicon via is used to guide biological component growth, such as dendrite or neural growth paths. In some embodiments, electrode arrays and biological components such as tissue cultures are alternatively stacked. In some embodiments, at least one through wafer vias (TSV) and electrodes are built in vertical stacks to provide compact connections. In some embodiments, biological sensing is accomplished through developed biological systems for sight, smell, touch sensing, or a combination thereof. In some embodiments, biological components directly connect to digital camera chip. In some embodiments, biological components directly connect to an inertial sensing system(s). Also described herein is or more non-transitory computer-readable media comprising computer-executable instructions that, when executed by at least one processor, cause the at least one processor to perform the method described herein.

[0003] In certain aspects, described herein is a method of manufacturing a three- dimensional tissue, comprising manufacturing the three-dimensional tissue directly into a bioreactor; wherein the tissue is greater than 1000 pm³. In some embodiments, the three- dimensional tissue is integrated with a three-dimensional multi el ectrode array. In some embodiments, the bioreactor comprises a first module comprising bioprocess controls and a second module comprising, the three-dimensional electrode array, wherein the first module is isolated from electrical stimulation. In some embodiments, the three-dimensional tissue comprises a three-dimensional tissue for use in regenerative medicine. In some embodiments, the three-dimensional tissue comprises neural tissue or muscle tissue. In some embodiments, the three-dimensional tissue is for use in a non-clinical trial. In some embodiments, the three- dimensional tissue is for use in cellular agriculture. In some embodiments, the three- dimensional tissue is for use in biocomputing. Also described herein is or more non-transitory computer-readable media comprising computer-executable instructions that, when executed by at least one processor, cause the at least one processor to perform the method described herein.

[0004] In some aspects, described herein is a method of analyzing a tissue comprising a three-dimensional electrode array, comprising: additive and subtractive manufacture of the tissue enmeshed with three-dimensional electrode array in a bioreactor, wherein the first module comprises bioprocess controls and the second module comprises the three- dimensional electrode array, wherein the first module is isolated from electrical stimulation; sending training signals to the tissue in the second module; and receiving signals from the three-dimensional electrode array. In some embodiments, the engineered tissue and the three- dimensional electrode array comprise a three-dimensional biocomputing system. In some embodiments, the method comprises a brain machine interface. In some embodiments, the engineered tissue is greater than 1000 pm³. In some embodiments, manufacturing comprises at least one method of additive manufacturing. In some embodiments, the tissue is vascularized. In some embodiments, the tissue comprises at least one cell type, at least two cell types, or at least three cell types. In some embodiments, at least one cell type comprises a neural progenitor cell, a stem cell, a primary tissue cell, a differentiated neuron, an astrocyte, an oligodendrocyte, a T cell, or a vascular cell. In some embodiments, the three-dimensional microelectrode array is embedded into engineered tissues. In some embodiments, at least one electrode in the three-dimensional electrode array can stimulate and record electronic cellular messaging. In some embodiments, a plurality of bioprocess controls regulate a plurality of bioprocess parameters simultaneously with at least one electrode in the three dimensional electrode array stimulating and recording electronic cellular messaging. In some embodiments, the three-dimensional microelectrode array comprises read and write capabilities. In some embodiments, the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues. In some embodiments, the three-dimensional microelectrode array with read and write capabilities is embedded into engineered tissues supported by engineered vascular networks. In some embodiments, the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues, supported by vascular networks. In some embodiments, the three- dimensional microelectrode array with read and write capabilities is embedded into engineered tissues supported by engineered avascular, active perfusion networks. In some embodiments, the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues, supported by avascular passive perfusion networks through the use of negative space in engineered tissues. In some embodiments, the method comprises manufacturing of the engineered tissue comprising a plurality of cells with the three-dimensional electrode array in a bioreactor system, with a three-dimensional electrode array capable of: sending relevant electrophysiology signals to the tissue; and receiving signals from the engineered tissue cells. In some embodiments, the method comprises receiving at least one signal from a plurality of cells in the engineered tissue, wherein the signal is received by the three-dimensional electrode array. In some embodiments, the method comprises receiving at least one signal from a plurality of cells in the engineered tissue, wherein the signal is received by an external receiver. In some embodiments, the method comprises training of cells within an engineered tissue using microelectrode pulses at physiologically relevant ranges embedded within engineered tissues. In some embodiments, the method comprises training of cells within an engineered tissue using microelectrode pulses at physiologically relevant ranges embedded within the tissues through the use of the three-dimensional arrangement of microelectrode arrays. In some embodiments, the three-dimensional electrode array emits a three- dimensional arrangement of electrical pulses to map networks of engineered tissues. In some embodiments, the method comprises reading data communication between a plurality of bioprocess controllers and a plurality of electric cell firings by multi-electrode arrays in a biomanufactured three-dimensional environment. In some embodiments, the use of training and test phases with three-dimensional multi-electrode arrays and which train and evaluate the quality of a neural network. In some embodiments, the method comprises controlling a bioprocess with a plurality of neural spikes. In some embodiments, the method comprises controlling a bioprocess with a neural spike pattern. In some embodiments, the method comprises storing information via in-silico and in vitro neural networks within a manufactured biological three-dimensional environment exceeding >1000 pm cross-section. In some embodiments, the method comprises writing, reading, and interpreting cell electrical firings with an artificial neural network within a manufactured biological three-dimensional environment exceeding >1000 pm cross-section. In some embodiments, multiple organoids or tissues are used together in biocomputation and information storage. In some embodiments, the pulse train, strength, frequency, pattern, duration, waveform, amplitude, shape, and physical location are used for read, write, and storage of information in a biocomputation environment. In some embodiments, the method comprises correlating gene expression to neuronal activity over time and three-dimensional space to indicate learning patterns and mechanisms. In some embodiments, the method comprises: the use of electrical, chemical, and/or physical stimulation to engineered tissue interior for use in biocomputation; the use of living neurons as perceptrons in an artificial neural network; the use of artificial neurons and biological neurons as enmeshed nodes in a neural network; or a combination thereof. In some embodiments, biological neuron firings from the three-dimensional multi el ectrode embedded in engineered tissue populates a data table which is then interpreted by artificial intelligence. In some embodiments, biological neuron firings from 3D multi el ectrode embedded in engineered tissue is directly interpreted by artificial intelligence in real-time. Also described herein is or more non-transitory computer-readable media comprising computer-executable instructions that, when executed by at least one processor, cause the at least one processor to perform the method described herein.

[0005] In some aspects, described herein is a method for converting annotated neural recordings into biocomputational inputs, the method comprising: capturing an annotated neuroscience dataset during subject cognitive activities using brain recording devices; processing the annotated neuroscience dataset to extract neural data comprise a plurality of neural spike trains; and introducing the plurality of neural spike trains into an engineered tissue via a brain-machine interface. In some embodiments, the annotated neuroscience dataset are derived from in vivo measurements. In some embodiments, the annotated neuroscience dataset comprises emergent data generated via biocomputation. In some embodiments, the method comprises a tokenization module configured to convert the neural data into discrete tokens representing distinct neural activity patterns, raw data, or concepts. In some embodiments, the stimulation inputs applied to the biocomputing tissue comprise a combination of tokenized waveforms and/or spike trains and simple spike signals that encode values such as numbers or letters. In some embodiments, the method comprises processing nodes operable in both a biological tissue layer and a digital computing layer to allow dynamic conversion, transmission, and processing of tokenized neural data. In some embodiments, the annotated neural data is further processed to generate emergent properties that can be used to enhance performance of the biocomputation network. In some embodiments, the method comprises an encoding module that facilitates digital-to-biological conversion by converting biological data such as neural waveform signals into ASCII representations, binary representations, or a combination thereof. In some embodiments, data are treated as multi-dimensional vectors that serve as embeddings for conceptual information. In some embodiments, the biological data is further processed to generate emergent properties that can be used to enhance biocomputational performance. Also described herein is or more non-transitory computer-readable media comprising computer-executable instructions that, when executed by at least one processor, cause the at least one processor to perform the method described herein.

[0006] In some aspects, described herein is a method for encoding and/or dual encoding of neural data, comprising: tokenizing neural data comprising a plurality of neural spike trains derived from annotated neuroscience datasets into discrete symbols; and/or forming a dual stimulation scheme in which both tokenized waveform representations and simple numerical spike signals are delivered to biocomputational tissue. [0007] In some aspects, described herein is a method for processing and manipulating conceptual information, the method comprising: converting neural spike train-derived waveforms into multi-dimensional vectors; tokenizing said vectors to form a plurality of embeddings; and combining the plurality of embeddings to deliver an output.

[0008] In some aspects, described herein is a method for training and inference in a biocomputation network comprising biocomputational tissues, comprising the steps of: preserving the temporal and quantitative attributes of a plurality of original annotated neural recordings during conversion into a plurality of spike trains delivering the plurality of spike trains to the biocomputational tissues; and processing the plurality of spike trains to enable language-related and concept-based tasks in a manner inspired by biological neural networks. [0009] In some aspects, described herein is a method for comprehensive encoding and decoding of neural information, comprising integrating layered conversion techniques that translate biological neural signals into digital representations, forming multi-dimensional embedding vectors that preserve semantic integrity.

INCORPORATION BY REFERENCE

[0010] All publications, patents, and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference.

BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The novel features of the invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the present invention will be obtained by reference to the following detailed description that sets forth illustrative embodiments, in which the principles of the invention are utilized, and the accompanying drawings of which:

[0012] FIGS. 1A-1C shows one example of a tissue bioreactor at different angles.

[0013] FIGS. 2A-2B depicts one example of a tissue bioreactor lid.

[0014] FIGS. 3A-3E depicts additional tissue bioreactor parts and attachments. [0015] FIGS. 4A-4B depicts bioprinting of tissues directly into tissue bioreactor.

[0016] FIG. 5 depicts an additional view of the tissue bioreactor now mounted to the stage and stereotaxic manipulator. [0017] FIG. 6 depicts a two-bioreactor system without electronics.

[0018] FIG. 7 depicts a two-bioreactor system with electronics.

[0019] FIG. 8 depicts a two-bioreactor system flow diagram.

[0020] FIG. 9 depicts microcontrollers used for prototype two-bioreactor system with cables removed to show core components.

[0021] FIG. 10 depicts an alternative view of the tissue bioreactor and multi-electrode arrays.

[0022] FIGS. 11A-11E depict quality by design applied to tissue engineering.

[0023] FIGS. 12A-12D depict brain-like and tendon-like bioprinted tissues.

[0024] FIGS. 13A-13E depicts cellular agriculture bioprinted tissues.

[0025] FIG. 14 depicts an example of biocomputing.

[0026] FIGS. 15A-15D depicts an additional example of biocomputing.

[0027] FIG. 16 depicts data communication between process microcontrollers and multielectrode arrays.

[0028] FIG. 17 depicts an example of signals sent by an artificial neural network.

[0029] FIGS. 18A-18B depicts Intan RHX Software.

[0030] FIG. 19 depicts weighted neural spike score explanation.

[0031] FIGS. 20A-20C depicts results from an initial experiment with weighted neural spikes.

[0032] FIG. 21 depicts an example of a random forest.

[0033] FIG. 22 depicts a biocomputer chipset and flow diagram.

[0034] FIG. 23 depicts a biocomputer block diagram.

[0035] FIG. 24 depicts a conceptual biocomputer module.

[0036] FIG. 25 depicts an integrated circuit MEA interface chip.

[0037] FIG. 26 depicts a local ‘passive’ non-digitized switching concept.

[0038] FIG. 27 depicts a passive switching block.

[0039] FIG. 28 depicts a multiplexing block.

[0040] FIG. 29 depicts axon tunnels with electrodes.

[0041] FIG. 30 depicts micromachined channels.

[0042] FIG. 31 depicts a MEA tissue well concept.

[0043] FIG. 32 depicts a secondary MEA tissue well concept.

[0044] FIG. 33 depicts a rack and stack concept.

[0045] FIG. 34 depicts a front-end circuit.

[0046] FIG. 35 depicts stimulation pulse shape. DETAILED DESCRIPTION

[0047] Described herein are systems and methods for tissue culture, tissue culture media perfusion, and bi-directional electrophysiologic control, using multi-electrode arrays as an interface between a biological and non-biological controller. This system enables growth, maintenance, manipulation, interpretation, manufacture, of large diameter tissues (>1000 pm cross section) for computational, medical, diagnostic, and cellular agriculture uses via three- dimensional assembly and a dedicated bioreactor system.

[0048] Despite the advantages of culturing cells in a three-dimensional environment, relatively few programs focus on culturing cells larger than a spheroid, or single layer of bioprinted tissue. This is in part because tissues are limited by their ability to transport nutrients, metabolites, and gasses critical to their survival through a tissue. For example, neurons are commonly cultured as organoids, also called spheroids or neurospheres. Neurospheres exhibit beneficial three-dimensional structural arrangement of cells from sizes of 100 pm to 500 pm. However, beyond sizes of 500 pm, neurospheres tend to exhibit significant necrosis at the center of the sphere because of limited gas and nutrient exchange. The lack of oxygen and nutrients causes cell death, effectively limiting neural tissue to sizes below 1000 pm.

[0049] Described herein are methods and manufacture of vascularized tissue culture with electrophysiologic circuitry, biological and non-biological control for growth, maintenance, manipulation and interpretation of large diameter tissue cultures. In some instances, described herein is a tissue culture method in which a three-dimensional bioprinted, vascularized tissue, is enmeshed with a three-dimensional multi-electrode array and bioprocess controls are separated into a secondary process bioreactor to minimize background electrical stimulation. Together, this enables long duration tissue culture, with read and write of electrical stimulus throughout a vascularized three-dimensional tissue. This enables a method of tissue culture, assessment of cellular activity, and biocomputation using electrical pulses.

I. BIOREACTOR SYSTEMS

[0050] Disclosed herein is a bioreactor system comprising a first module comprising a plurality of bioprocess controls; and a second module comprising a system for electrical stimulation and electrical recording. The second module may be isolated from electrical interference from the first module.

[0051] One example of a tissue bioreactor system is depicted in FIGS. 1A-1C. FIGS. 1A-1C shows 3 views of the tissue bioreactor. Individual features are highlighted, either as a model in a Computer Assisted Design (CAD) software (FIGS. 1A-1B) or three-dimensional printed in ABS plastic printed in a Bambulab PIS three-dimensional printer, sanded, chemically smoothed and treated with biocompatible sealant where appropriate (FIG. 1C). In some embodiments, ports 101 may allow flow of liquid in and out of the bioreactor. The bioreactor may comprise between 0 to 16 ports, or more than 16 ports. FIGS. 1A-1C depict 8 ports. This iteration of the tissue bioreactor base has 8 total ports, but versions have been made with anywhere from 0 to 16. All ports are printed at sufficient height and dimensions to ensure no interference with the three-dimensional bioprinter.

[0052] In some embodiments, the tissue bioreactor may comprise an “anchor ring” ring 102. The anchor ring may anchor the tissue bioreactor to the tissue bioreactor base/staging area. In some embodiments, the anchor ring has a hole 103 which fits a screw in order to properly secure the bioreactor to the anchor. In some embodiments, the bioreactor comprises a small lip 104 is present at the bottom of the bioreactor, this lip is necessary to fit securely into the bioprinter with no slippage during printing. In some embodiments, a channel in the top of the bioreactor fits an O-ring 105 . When the bioreactor lid is pressed against this Ciring, it creates a water-tight seal. In some embodiments, the interior of the bioreactor 106 fits large tissue cultures. In some embodiments, the bioreactor comprises holes 107 for screws to secure the lid to the bioreactor and base and stage are present. In some embodiments, the bioreactor interior has attachment ports 108 for tubing, which is necessary for certain tissue vascularization techniques. In some embodiments, silicon tubing 109 is attached to each of the ports. In some embodiments, Luer lock connectors 1110 are attached to the bioreactor tubing. When combined with a closed lid, having closed tubes allows for the bioreactor to be completely sealed for transport. This allows the tissue bioreactor to move from the bioprinter, to the laminar flow hood, to the process bioreactor staging area with appropriate aseptic technique.

[0053] In some embodiments, the bottom of the bioreactor has holes 112 which fit securing nuts, allowing for a secure connection of the lid to the bioreactor. In some embodiments, some tissue bioreactor inlets and outlets have 3 way valves 111 . These may be used to connect temperature/pressure probes, and additional media lines or syringes. The ability to connect syringes is particularly helpful when conducting tissue washes and crosslinking in the bioprinter.

[0054] In some embodiments, the bioreactor system comprises a tissue bioreactor lid. An example of a tissue bioreactor lid is depicted in FIGS. 2A-2B. In some embodiments, small lid plates 201 are used to connect the three-dimensional multi-electrode array and the surface multi-electrode array to the bioreactor aseptically. The gap in image A shows the hole where a multi -electrode array interfaces with its respective lid plate. An O-ring in the middle of this hole serves as a sealing gasket, which ensures a watertight connection. In some embodiments, the multi -el ectrode array blocking plate has a ring 202 where a gasket sits to make a watertight seal to the reactor lid. In some embodiments, holes 203 in the lid allow for the reactor lid and bioreactor to be screwed securely together. In some embodiments, a small hose barb and port 204 exists on the lid on which to secure a pressure/temperature probe. In some embodiments, holes 205 which interface with the multi-electrode array blocking plate allow for the multi -el ectrode arrays to pass into the interior of the bioreactor while maintaining an aseptic seal. In some embodiments, a larger hole 206 exists in the bioreactor lid, in which a Plexiglass plate can be firmly and permanently secured. In some embodiments, two hose barbs and ports 207 may connect to tubing for gas inlet and outlet, enabling DO, pH, and pressure control when appropriate.

[0055] Additional embodiments of a bioreactor are depicted in FIGS. 3A-3E. In some embodiments, the bioreactor system comprises portions such as depicted in FIG. 3A to secure the tissue bioreactor stage to the stereotaxic frame, and tubing holder on the left side. In some embodiments, the tissue bioreactor stage such as depicted in FIG. 3B has a spot to securely fit the tissue bioreactor. Screw holes allow the tissue bioreactor to be secured directly to the stage. In some embodiments, the bioreactor system comprises perfusion aids such as those depicted in FIG. 3C which can be printed directly into the tissue bioreactor alongside tissue using a thermoplastic printhead to ensure media perfusion directly into tissue.

[0056] In some embodiments, of the tissue bioreactor as depicted in FIG. 3D, only one hose/barb/port enters the tissue culture chamber and other ports run directly through the bottom of the bioreactor to facilitate perfusion of vascularized tissue. In some embodiments, the bioreactor system comprises injection molds such as depicted in FIG. 3D for biomaterials with ports to perfuse crosslinking reagent, and plates which facilitate removal of tissue from the injection molds.

[0057] One example of a bioreactor mounted to a stag is depicted in FIG. 5. In some embodiments, one of three temperature/pressure sensors (LPS33HW Water Resistant Pressure Sensor - Stemma QT Adafruit) 501. The tissue bioreactor may have a temperature pressure probe on 1 inlet tube, 1 outlet tube, and the tissue bioreactor lid. In some embodiments, a three-dimensional printed ‘anchor’ 502 connects the tissue bioreactor, tissue bioreactor stage, and provides support for tubing on the tissue bioreactor. In some embodiments, screws 503 are used to secure the tissue bioreactor to mounting anchor. In some embodiments, tubing 504 is secured to the tissue bioreactor lid. In some embodiments, O-rings 505 are used to maintain an aseptic connection at multi -el ectrode array ports. In some embodiments, the stereotaxic manipulator 506 allows for precise insertion of the multielectrode arrays into the tissue. In some embodiments, an acrylic sheet 506 provides the tissue bioreactor with a closed viewing port. In some embodiments, the three-dimensional printed tissue bioreactor stage 508 securely holds the tissue bioreactor. In some embodiments, the stereotaxic manipulator is on an antistatic pad 509 to prevent electrical charges from travelling through the table to the tissue bioreactor.

[0058] FIG. 6 depicts an example of a bioreactor system comprising two modules. In some embodiments, many of the bioprocess control electronics (probes, pumps, controllers) are controlled through a series of Arduino microcontroller devices 601. These may be housed in individual cases so that they may be easily swapped for other microcontrollers. In some embodiments, series of pumps 602 to move liquid through the system. In some embodiments, syringes 603 attached to tissue bioreactor inlet for washes independent of the pumps. In some embodiments, the bioreactor system comprises a tissue bioreactor 604 on its stage. In some embodiments, the system comprises a temperature controller 605. The temperature controller may comprise a magnetic stirrer connected to a Digital Loggers Enclosed High-Power Power loT Relay. In some embodiments, an electronic microscope 606 hung from a retort stand can view through the viewing window without touching the tissue bioreactor and causing electrical interference. In some embodiments, a plurality of bottles 607 are connected to the bioreactors to add or remove media components including without limitations, nutrients, metabolites, collect samples, and adjust pH. In some embodiments, the stimulation and recording controller 608 can both stimulate and record electrical pulse within the tissue bioreactor at physiologically relevant ranges across a plurality of channels using small, affordable hardware and free, open-source software. In some embodiments, the controller can both stimulate and record electrical pulse across 128 channels. In some embodiments, air is moved in and out of the system to provide oxygen and pH control. The air controller 609 may be connected to a Digital Loggers Enclosed High-Power Power loT Relay. In some embodiments, the process bioreactor 610 in which temperature, pH, dissolved oxygen, and other process parameters are controlled without interfering with the tissue reactor multielectrode arrays.

[0059] In some embodiments, the two-module biosystem reactor comprises the system in FIG. 7. FIG. 7 comprises FIG. 6 with the addition of cables to the Intan Technologies stimulation and recording controller, microcontroller connections, multi-electrode array headstages, and various grounding hubs. In some embodiments, the components of the bioreactor system are arranged as depicted in FIG. 8. Media flow is generally in the direction shown but can be reversed. For instance, fluid may flow from media bottles 801 through a first pump 802 into the tissue bioreactor 803. The media bottles 801 may be kept in a fridge 804. Fluid may flow out of the tissue bioreactor 803 into a sample container 805. Fluid may flow out of the tissue bioreactor 803 through a second pump 806 to a stirred tank bioreactor 807. Fluid may flow out of the stirred tank bioreactor 807 through a third pump 808 into the custom tissue reactor 803. Fluid may flow from a media bottle 809 into the stirred tank bioreactor 807. Fluid may flow out of the stirred tank bioreactor 807 through a fourth pump 810 into a waste unit 811. Fluid for adjusting pH, such as a base, may flow from a pH regulator 812 through a fifth pump 813 into the stirred tank bioreactor 807. A pump controller 816 may communicate with the first pump 802, the second pump 806, the third pump 808, the fourth pump 810, and the fifth pump 813. The pump controller 816 may communicate with the computer 817. The computer 817 may communicate with the tissue culture controllers 818. The tissue culture controllers 818 may communicate with the tissue culture probes and control devices 819. The computer 817 may communicate with the bioprocess controllers 820. The bioprocess controllers 820 may communicate with the bioprocess probes and control devices 821. All components are modular, allowing for relative ease to switch in and out components location or function.

[0060] The bioreactor system described herein may comprise additional components. In some instances, the bioreactor comprises at least 1 port for flow of liquid into the bioreactor. In some embodiments, the liquid comprises growth factors, nutrients, metabolites, stabilizers, pH indicators and controllers, living and non-living components. In some embodiments, the bioreactor system comprises a perfusion aid. In some embodiments, the bioreactor system comprises least one syringe. In some embodiments, the system for electrical stimulation comprises at least one electrical component comprising at least one of a pump, a microcontroller, or controller, a probe, and a microelectrode array.

[0061] In some embodiments, a central data repository is used to link process control in an Arduino environment and stimulation/recording in a python environment with or without TensorFlow and Juypter notebook integration for machine based neural network integration. An example is depicted in FIG. 16. The integrated computer system 1601 may comprise electrophysiology-python environment 1602, a data repository 1603, and process record and control 1604. The electrophysiology-python environment may comprise a Jupyter notebook integration. The electrophysiology -python environment may comprise tensorflow integration. The process record and control may be an Arduino environment. Data may flow between the electrophysiology-python environment 1602 and the data repository 1603. Data may flow between the process record and control 1604 and the data repository 1603. Data may flow between the process record and control 1604 and the Arduino based microcontrollers 1605. Data may flow between the Arduino based microcontrollers 1605 and the process probes 1606. Data may flow between the Arduino based microcontrollers 1605 and the process controllers 1607. Data may flow between the electrophysiology-python environment 1602 and the MEA record/stimulation device 1608. Data may flow between the MEA record/stimulation device 1608 and the multi el ectrode array 1609. The multi el ectrode array 1609, the process controllers 1607, the process probes 1606 or a combination thereof may interact directly with the reactors 1610.

A. Isolation from Electrical Interference

[0062] The bioreactors system described herein comprises at least two modules, wherein the second module is isolated from electrical interference from the first module. In some embodiments, the second module is not exposed to at least 70%, at least 80%, at least 90%, or 100% of the electrical signature generated by the first module.

[0063] The first module may comprise bioprocess controls that emit electrical interference. In some embodiments, the bioprocess controls comprises at least one electrical component comprising at least one of a pump, a microcontroller or controller , and a probe. Isolating the second module from electronic interference from the first module may comprise grounding at least one electrical component. In some embodiments, each electrical component is grounded. The electrical components may include one or pump, for the addition and removal of liquids, and pressure control.

[0064] Isolating the second module from the first module may comprise placing the second module on an anti-static mat. Isolating the second module from the first module may comprise physically isolating the first module from the second module. In some embodiments, the first module comprises a first chamber and the second module comprises a second chamber. In some embodiments, the first module comprises an incubator and the second module comprises a bioreactor placed within the incubator. The second module may comprise a cell culture or a tissue culture plate. In some embodiments, the first module is not physically isolated from the second module. B. First module

[0065] In the system describe herein, the first module comprises a plurality of bioprocess controls. The plurality of bioprocess controls comprises at least one of a pH control, a dissolved oxygen control, a temperature control, a pressure control, and control of gas, liquid, and solid components added and removed from each reactor, and combinations thereof. The plurality of bioprocess controls may comprise at least a pH control, a dissolved oxygen control, a temperature control and a pressure control.

[0066] An example of systems for bioprocess controls is depicted in FIG. 9. This allows for custom control of the bioprocess as discussed in later figures. In some embodiments, the bioreactor system comprises an Arduino Uno microcontroller used to control a Digital Loggers Enclosed High-Power Power loT Relay 901. This relay can turn power on and off for various heating elements, the process bioreactor magnetic stirrer, and air flow via air pumps. In some embodiments, the bioreactor system comprises an Arduino Uno microcontroller with an Atlas Scientific i2 InterLink shield and a black temperature control module 902. In some embodiments, the bioreactor system comprises an Atlas scientific red pH module used to record pH 903. In some embodiments, the bioreactor system comprises an Atlas scientific yellow Dissolved Oxygen module 904 used to record Dissolved Oxygen. In some embodiments, the bioreactor system comprises a plurality of Arduino Uno microcontrollers with an attached CNC shield 905, used to control stepper motors and associated KPHM-100 pump heads. Each red or purple stepper motor driver may be used to control a single pump head. The nine stepper motor drivers shown across 3 microcontrollers can control 9 motors. Only 6 pumps are used in the displayed setup. In some embodiments, the bioreactor system comprises an Arduino UNO R4 WIFI enabled microcontroller board with an integrated Qwiic connector 906. The Qwiic connector may be used to interface with 3 of the pressure/temperature sensors (LPS33HW Water Resistant Pressure Sensor - Stemma QT Adafruit).

C. Second module

[0067] The system described herein, in some embodiments, comprises a system for manufacturing in the second module. The manufacturing may comprise additive manufacturing, subtractive manufacturing, or a combination thereof. In some embodiments, the second module comprises a system for assembly, growth, and control of tissues by additive or subtractive manufacturing. [0068] For example, one method of biomanufacturing is depicted in FIGS. 4A-4B. In some embodiments, a mix of cells and scaffolding material are printed directly into the tissue bioreactor using a Cellink BioX Bioprinter. In some embodiments, the Cellink bioprinter allows 3 different cell/ scaffolding mixtures to be printed within an aseptic laminar flow environment. In some embodiments, the external ports on the bioreactor within the bioprinter allow for aseptic addition of cell washes, crosslinking reagents, growth media and other liquids during the printing process. In some embodiments, two bioink/cell mixtures are interfaced using multiple bioprinter printheads. In some embodiments, the lighter of the two inks contains cells, and the darker of the two services as a scaffold which is washed out and removed. In some embodiments, the cavity is washed with vascular cells, which seed a vascular network and long-term tissue viability.

[0069] In some embodiments, the second module comprises a system for additive manufacturing. In some embodiments, additive manufacturing comprises bioprinting and injection molding of biomaterials. In some embodiments, the system for additive and subtractive manufacturing comprises a scaffold material and a plurality of cells. In some embodiments, the scaffold material is biocompatible.

[0070] Biomanufacturing processes may comprise use of a plurality of cells. In some embodiments, the plurality of cells comprises a neural progenitor cell, a stem cell, a primary tissue cell, a differentiated neuron, an astrocyte, an oligodendrocyte, a t-cell, a vascular cell, or a combination thereof. In some embodiments, the plurality of cells comprises a plurality of neural progenitor cells. In some embodiments, the plurality of cells comprises a plurality of stem cells. In some embodiments, the plurality of cells comprises a plurality of primary tissue cells. In some embodiments, the plurality of cells comprises a plurality of differentiated neurons. In some embodiments, the plurality of cells comprises a plurality of astrocytes. In some embodiments, the plurality of cells comprises a plurality of oligodendrocytes. In some embodiments, the plurality of cells comprises a plurality of t-cells. In some embodiments, the plurality of cells comprises a plurality of vascular cells. Vascular cells include, without limitations, endothelial cells, angioblasts, and smooth muscle-like cells.

[0071] Cell culture media generally include essential nutrients and, optionally, additional elements such as growth factors, salts, minerals, vitamins, etc., that may be selected according to the cell type(s) being cultured. Particular ingredients may be selected to enhance cell growth, differentiation, secretion of specific proteins, etc. In some embodiments, growth media include Dulbecco's Modified Eagle Medium, low glucose (DMEM), with 110 mg/L pyruvate and glutamine, supplemented with 10-20% fetal bovine serum (FBS) or calf serum and 100 U/ml penicillin, 0.1 mg/ml streptomycin are appropriate as are various other standard media well known to those in the art. In some embodiments, cells are cultured under sterile conditions in an atmosphere of 3-15% CO2, or about 5% CO2, at a temperature at or near the body temperature of the animal of origin of the cell.

[0072] The cells can also be cultured with cellular differentiation agents to induce differentiation of the cell along the desired line. For instance, cells can be cultured with growth factors, cytokines, etc. The term “growth factor” as used herein refers to a protein, a polypeptide, or a complex of polypeptides, including cytokines, that are produced by a cell and which can affect itself and/or a variety of other neighboring or distant cells. Typically growth factors affect the growth and/or differentiation of specific types of cells, either developmentally or in response to a multitude of physiological or environmental stimuli. Some, but not all, growth factors are hormones. Growth factor include, without limitations, insulin, insulin-like growth factor (IGF), nerve growth factor (NGF), vascular endothelial growth factor (VEGF), keratinocyte growth factor (KGF), fibroblast growth factors (FGFs), including basic FGF (bFGF), platelet-derived growth factors (PDGFs), including PDGF-AA and PDGF-AB, hepatocyte growth factor (HGF), transforming growth factor alpha (TGF-a), transforming growth factor beta (TGF-0), including TGF01 and TGF03, epidermal growth factor (EGF), granulocyte-macrophage colony-stimulating factor (GM-CSF), granulocyte colony-stimulating factor (G-CSF), interleukin-6 (IL-6), IL-8, and the like.

1. Electrode arrays

[0073] In some instances, the second module comprises an electrode array. The electrode array may be arranged as a three-dimensional electrode array or as a surface electrode array. In some embodiments, the electrode array is arranged in a grid. In some embodiments, the electrode array comprises both a surface and three-dimensional electrode or microelectrode array for recording and stimulation of engineered tissues.

[0074] One example of an electrode array is depicted in FIG. 10. In some embodiments, the Headstage 1001 for the surface multi -electrode array connects through a plate in the tissue bioreactor. In some embodiments, a custom surface multi -electrode array 1002 records and stimulates tissue through a grid of multi-electrodes. In some embodiments, cables 1003 connect the multi-electrode array headstage to the stimulation/recording module. In some embodiments, the multi -el ectrode array headstage 1004 connects the cable to a custom made Atlas scientific three-dimensional multi-electrode arrays. In some embodiments, the system comprises a plurality of grounding hubs. In some embodiments, one of the grounding hubs 1005 grounds various devices in order to minimize background electrical noise within and around the tissue bioreactor. In some embodiments, the system comprises a three- dimensional multi-electrode array 1006 raised from tissue using stereotaxic manipulator for visualization. The multi -el ectrode array may comprise sharp tipped shafts containing no less than 6 multi-electrodes each. The shafts may be arranged in a 2 x 4 arrangement, allowing for a true three-dimensional multi -el ectrode environment within printed, vascularized, process controlled tissue.

[0075] In some embodiments, the electrode array comprises at least 4 electrodes. In some embodiments, the electrode array comprises at least 1, 2, 3, 4, 5, 6, 7, 8,9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900 or more electrodes, including increments therein.

[0076] The electrode array may be integrated with the biomanufactured tissue. In some embodiments, the electrode array is configured for recording and stimulation of engineered tissues. In some embodiments, each electrode is configured to stimulate and record electronic cellular messaging. In some embodiments, each electrode is configured to stimulate at least one cell in the biomanufactured tissue. In some embodiments, data may be recorded using Complementary metal-oxide-semiconductor (CMOS) based probes such as the neuropixel probe. In some embodiments, CMOS are advanced neural recording devices that comprise dense arrays of electrodes on a single shank, capable of simultaneously recording from hundreds of neurons across multiple brain regions. In some instances, CMOS are manufactured with a metal-oxide-semiconductor field-effect transistor (MOSFET) fabrication process that uses complementary and symmetrical pairs of p-type and n-type MOSFETs for logic functions.

[0077] Many of the biocomputation modalities discussed are presented with a multielectrode array context. However, the biocomputation methods described are not necessarily reliant on multi -el ectrode arrays to function. In some embodiments, other tools could be used to accomplish similar results in the described biocomputation framework. Some alternative embodiments are present herein.

[0078] In some embodiments, Genetically Encoded Voltage Indicators (GEVIs) are used. GEVIs are proteins engineered to sense membrane potential changes within neurons and emit fluorescence in response. In some embodiments, by introducing genes encoding these indicators into specific neurons, the systems described herein may be used to optically monitor electrical activity with high temporal precision. In some embodiments, the use of GEVIs enables non-invasive, cell-type-specific recordings, facilitating detailed studies of neural circuits and their computational properties.

[0079] In some embodiments, optogenetic methods are combined with fluorescent indicators to enable system modeling. In some embodiments, optogenetics involves using light-sensitive proteins to control neuronal activity. When combined with fluorescent indicators, this approach may allow simultaneous manipulation and recording of neural activity. For example, In some instances, activating neurons with light while recording their responses through fluorescence provides insights into neural processing and can be utilized to modulate biocomputational pathways dynamically.

[0080] In some embodiments, optetrodes are leveraged. In some configurations, these devices combine optical fibers with electrodes, enabling simultaneous optical stimulation and electrical recording of neurons. In some embodiments, this integration allows precise modulation and monitoring of neural circuits, enhancing the biocomputer's capability to interact with neural tissues.

[0081] In some embodiments, the system integrates ultrasound methods to enhance neural control. Ultrasound-based stimulation and monitoring may involve non-contact ultrasound pulses that modulate neural activity as an alternative to electrical stimulation. In some embodiments, these modalities work within a combined feedback loop, complementing electrical and chemical signals for nuanced control. In some instances, the integration of ultrasound methods into neuromorphic systems introduces a new level of precision by targeting specific neural populations beyond what is achievable with standard electrical -only systems. This is particularly significant as electrical -only stimulation may have limitations in deep structures.

[0082] In some embodiments, the combination of electrical, optical, and ultrasound stimulation facilitates deeper and more selective neural control

[0083] In some embodiments, recording is performed using calcium imaging with Genetically Encoded Calcium Indicators (GECIs). GECIs are proteins that fluoresce upon binding to calcium ions, which are indicative of neuronal activity. In some embodiments, by expressing GECIs in neurons, our can visualize and record activity patterns across populations of neurons. In some instances, this technique is useful for monitoring network dynamics and understanding how neural ensembles contribute to computation.

[0084] Functional Magnetic Resonance Imaging (fMRI) is used to measure brain activity by detecting changes in blood flow, reflecting neural activity. In some biocomputation embodiments, fMRI can be adapted to assess large-scale neural activity patterns and their correlation with computational tasks, providing a non-invasive method to study the functional architecture of neural networks.

[0085] Each of these methods offers distinct advantages for recording neural information, and their integration into biocomputation systems can enhance our understanding and utilization of neural processing for computational purposes.

D. Additional Modules

[0086] The bioreactor systems disclosed herein may comprise many additional, distinct units, blocks, and/or modules that can be used individually or in conjunction with each other to achieve specific tasks. These blocks may represent distinct tasks or operations performed by the system. They may exist in various formats including on the same board, or on separate boards or multiple locations. For instance, the bioreactor systems may comprise a module comprising a biological to digital decoder; a module comprising a neural signal preprocessing block; a module comprising a digital-to-biological encoder; a module comprising AI/ML assisted post-processing; a module for additional ML training and offloading; a module for local data storage; a module for cloud data storage; a module for CMOS-based biocomputation; or a combination thereof.

2. Data aggregation

[0087] In some embodiments, the system comprises a module comprising a global system for data aggregation, and control commands. In some instances, this module is referred to as a global FPGA controller. In some embodiments, this system is comprised of subcomponents that may include but are not limited to components in charge of signal collection, data aggregation, real-time workload scoring, a load manager, health monitor, task and table scheduler, a checkpoint, transfer blocks, adaptive stimulation controller(s), and a microfluidic control interface.

[0088] In some embodiments, the global FPGA is responsible for aggregating data from all clusters. This includes various types of signal features, for example: Wavelet Transforms: Capture transient, time-varying events in neural signals; Fourier Transforms: Break down the neural signals into their constituent frequency components; Lyapunov Exponents: Indicate how sensitive neural activity is to small changes, helping to assess chaotic behavior, which could be indicative of neural tissue distress; Sample Entropy: Measures the predictability or randomness in the neural data; and Local Field Potentials (LFPs) and Gamma Waves: Provide a broader picture of neural group activity and can indicate overall tissue health or stress.

[0089] In some embodiments, the module is configured to aggregate data. In some embodiment, the global FPGA collects these metrics in real time, combining them to form a comprehensive view of the state of the neural tissue across the chip. This data is then used to decide if any clusters are overloaded, underperforming, or showing signs of stress.

[0090] In some embodiments, the module is a dedicated DSP module. In some embodiments, there are dedicated Digital Signal Processing (DSP) modules with the FPGA that process the incoming neural data. These modules may handle tasks such as Real-Time Workload Scoring. For Real-time workload scoring the FPGA computes a workload score for each cluster by combining the normalized outputs of these analyses using weighted summation. This workload score may help determine whether a cluster needs stimulation adjustments or a shift to additional clusters to maintain proper tissue compute functions.

[0091] In some embodiments, the load manager continuously receives “Load Score” data from each neural cluster and compares these values to predefined thresholds, such as 0.8 for overload. It may coordinate with the Health Monitor to assess whether a cluster is stable or exhibiting chaotic signals, ensuring efficient task distribution and preventing computational bottlenecks.

[0092] In some embodiments, the health monitor collects “Health Score” metrics, including chaotic spikes and gamma anomalies, from each neural cluster. When a cluster approaches unhealthy conditions or exhibits irregular activity, the Health Monitor may alert the Load Manager, allowing proactive load balancing and system stabilization.

[0093] In some embodiments, the task table & scheduler maintains a record of active tasks, such as partial matrix multiplications, along with their priority and progress. It may determine which neural cluster should receive new or migrated tasks, optimizing workload distribution. For fast lookups and real-time scheduling, the task table may be stored in on- chip Block RAM, ensuring minimal latency.

[0094] In some embodiments, the checkpoint & transfer block manages the checkpointing of partial computation states, such as intermediate matrix multiplication results, from an overloaded cluster’s local memory. It may utilize DMA engines to efficiently transfer these states via the AXI bus for on-chip movement or LVDS/SERDES links for off- chip or cross-module communication, ensuring seamless computational handoff between clusters. [0095] In some embodiments, the adaptive stimulation controller interfaces with the neural clusters’ stimulation lines to dynamically adjust pulse frequencies. Based on task priority and cluster stress signals, it may increase or decrease stimulation to optimize computational efficiency and maintain neural stability.

[0096] In some embodiments, the microfluidic control interface communicates with microcontrollers responsible for nutrient and chemical delivery. It may issue real-time commands to regulate flow rates, adjusting to the cluster’s updated load and health status to maintain optimal environmental conditions for sustained neural activity.

3. Biological to digital decoders

[0097] In some embodiments, the bioreactor system further comprises a biological-to- digital decoder that performs digital conversion of neural signals. In some instances, this unit converts analog neural signals into digital data using high-resolution ADCs. In some instances, digital conversion of neural signals is the first step in processing, forming the basis for subsequent analysis and computational task allocation. Specifically, the biological-to- digital decoder may convert the neural tissue activity into a structured digital format that can be processed by downstream FGPA based or ASIC based computation via high speed LVDS/SERDES communication links. This may allow real-time analysis of ongoing stimulation or computational workloads. The biological to digital decoder may transform raw neural signals into structured digital data. The biological to digital decoder may allow the real-time monitoring of neural tissue compute performance. The decoding of the neural tissue’s electrical activity may create feedback loops where FGPA- based signal processing can modify tissue stimulation dynamically. The conversion of the neural tissue’s electrical activity into a digital format may permit the detection of the tissue’s stress states , potential overactivity or chaotic activity for load balancing and self-regulation via high speed LVDS/SERDES communication with the global FGPA controller.

4. Pre-processing modules

[0098] In some embodiments, the bioreactor system further comprises a module comprising a pre-processing block. The pre-preprocessing block may process raw digital neural signals; perform transformations (applying adaptive filtering, Fourier/wavelet transforms, and extracting speech features); send data to relevant blocks, such as the Global FPGA Controller, to storage, or, directly back to tissue, via LVDS/SERDES and high-speed ADC/I/O connections; or a combination thereof. This module may also be referred to as FPGA-Based Pre-processing.

5. Digital to Biological encoders

[0099] In some embodiments, the bioreactor system further comprises a module comprising a digital-to-biological encoder. In some embodiments, this encoder converts digital commands into analog stimulation pulses delivered to neural tissue using dedicated analog signal lines. In some embodiments, the digital-to-biological encoder applies preprocessing steps such as filtering, normalization and gain adjustment to assure that signals match neural tissue thresholds. In some embodiments, digital signal data is converted into biologically relevant stimulation pulses. In some embodiments, the core component of the digital-to-biological decoder is a high speed digital-to-analog converter(DAC) array that translates the digital signal into precise electrical pulses. In some embodiments, each DAC channel corresponds to an electrode group in the multi electrode array (MEA) that interfaces with the neural tissue.

6. AI/ML-assisted post-processing modules

[0100] In some embodiments, the bioreactor system further comprises a module comprising an AI/ML-assisted post-processing module. In some embodiments, this module processes data with error correction and optional Al routines. In some instances, the AI/ML- assisted post-processing unit and data handling mechanism further refine signal interpretation and output management. In some embodiments, this unit ingests filtered signals from the FPGA neural signal processing modules and applies machine learning sub-modules to detect and correct anomalies, ensuring a consistent data structure. In some embodiments, the processed data is then delivered to network attached storage such as ASIC, SSD, or cloud storage for further analysis. In some embodiments, implementation approaches can leverage FPGA, ASIC, or a hybrid solution, balancing reprogrammability and power efficiency based on task requirements. In some embodiments, the integration of ALassisted post-processing with biological signals bridges the gap between digital error correction and neural tissue computation, ensuring the accurate and timely processing of neural signals for downstream applications.

[0101] In some embodiments, the Digital to biological Encoder converts digital signals into pre. 7. ASIC Modules for additional ML Training & Offloading

[0102] In some embodiments, the bioreactor system further comprises a module comprising an ASIC Module for Additional ML Training & Offloading. In some embodiments, this block offloads stable DSP tasks from the FPGA, stores spike trains and synaptic weights, and runs an on-chip ML model to extract features and make predictions. In some embodiments, the ASIC module offers a power-efficient, low-cost solution for executing stable DSP functions and on-chip ML tasks. In some instances, the module allows rapid classification or prediction based on neural data while reducing latency and energy consumption.

8. Network attached storage

[0103] In some embodiments, the bioreactor system further comprises a module comprising network attached storage. In some embodiments, the module comprises a top layer connectivity module that connects externally via Optical/TCP-IP links. In some embodiments, it serves as the primary data output interface. In some embodiments, the module comprises network attached storage that uses high-speed optical or serial links (with TLS/SSL encryption) to connect other modules to itself, external networks, and data centers.

9. Remote storage

[0104] In some embodiments, the bioreactor system further comprises a module comprising remote storage. In some embodiments, the module is configured to use highspeed optical or serial links (with TLS/SSL encryption) to connect other modules to itself, network attached storage, external networks, and data centers.

E. Use of system

[0105] In some embodiments, additive manufacturing is used to directly biomanufacture tissues into said reactor. In some embodiments, the bioreactor system allows stimulation and recording of engineered tissue electrophysiochemistry signals through isolation and reduction of background electrical noise. In some embodiments, engineered tissue viability is maintained by way of active perfusion, or passive perfusion in which avascular engineered tissues of diameter greater than 1000 microns, are maintained through the use of porous biomaterials. In some embodiments, engineered tissue viability is maintained by way of active perfusion, or passive perfusion in which avascular engineered tissues of diameter greater than 1000 microns, are maintained through the use of tissue-free spaces, such as avascular channels for media transport, by way of additive or subtractive manufacturing perfusion.

[0106] The bioreactor systems described within may be used to generate tissue for use in a variety of functions, including without limitations, artificial neural networks, regenerative medicine, non-clinical trials, or cellular agriculture. In some embodiments, the three- dimensional tissue comprises a three-dimensional tissue for use in regenerative medicine. In some embodiments, the three-dimensional tissue comprises neural tissue. In some embodiments, three-dimensional tissue comprises muscle tissue. In some embodiments, the three-dimensional tissue is for use in a non-clinical trial. In some embodiments, the three- dimensional tissue is for use in cellular agriculture. In some embodiments, the three- dimensional tissue is for use in biocomputing.

1. Artificial Neutral Networks

[0107] In some instances, the bioreactor used within can create an artificial neutral network. Artificial neural networks aim to simulate biological neural networks. Similar communication and processing modalities are therefore self-evident. Methods of electrode communication include pulse train strength, frequency, pattern, duration, waveform amplitude, shape and electrode location. Neural network outputs are best interpreted through multifactorial artificial neural nets. When attempting to use cell activity as an analytical method, pulses can be visualized and scored with a proprietary weighted pulse score. Where the weighted pulse score takes all aforementioned pulse quality metrics and compares them to the desired output. One example of an artificial neural network is depicted in FIG. 17.

2. Regenerative Medicine

[0108] Regenerative medicine is a field focused on the development and application of new treatments to restore function to tissues lost to aging, disease, damage or defects. Physically and functionally complete tissues (intact tissues) capable of long-term survival are useful regenerative medicine. However, commercially available methods and equipment are incapable of supporting most types of intact tissues at a therapeutic scale, limiting their development. Potential regenerative medicine targets include, without limitations, disability or injury of tendons or the brain, including traumatic brain injury and glioblastomas.

[0109] Various brain conditions are treated through the removal of damaged brain tissue. Brains with resected tissue may recover faster or more completely if cells can replace the removed tissues. However, implanted cells can suffer from washout and low survivability when implanted in resection cavities. Intact tissues may prove more capable of providing sustained therapeutic benefit.

[0110] Tendon surgeries such as ACL reconstruction rely on translocation of muscle tissue from a patient’s own native muscle which weakens support structures, resulting in a cascade of degeneration and subsequent injury. By instead growing the tissue necessary for ACL surgery as an external allogeneic product it reduces the burden of weakened muscles.

3. Non-clinical Trials

[0111] Regenerative medicine development relies on non-clinical trials such as animal trials to test products prior to testing in human patients. These trials can be expensive, are concerning to animal ethicists, and do not always relevant to clinical pathology. Intact tissues may provide additional means and controls for serve non-clinical trials. For example, with brain tissue replacement therapy, a section of engineered brain tissue could be removed and a secondary tissue placed in the resection cavity to test the ability of the two tissues to integrate, with no animal trials.

4. Cellular Agriculture

[0112] The animal agriculture sector is the single largest anthropogenic user of land, contributing to soil degradation, dwindling water supplies, and air pollution. Further, overfishing erodes not only the over-fished species but also cascade across the food web, leading to loss of other important marine life.

[0113] Cellular agriculture comprises using lab grown cells as an alternative protein source. By making intact muscle cellular tissue food products with the same appearance and form as animal agriculture products at a lower cost, it may be possible to ensure widespread adoption of cell cultured meat. Many cell and plant based alternative meat companies have limited capability to produce intact muscle products. Those that do lack the ability to produce tissues thicker than a few millimeters because of oxygen and media delivery constraints.

II. METHODS OF USE

A. Methods of manufacturing a three-dimensional tissue

[0114] The bioreactor systems described herein can be used in a method of manufacturing a three-dimensional tissue. The tissue may be greater than 1000 pm³.

[0115] In some embodiments, the three-dimensional tissue is integrated with a three- dimensional multi el ectrode array as described herein. In some embodiments, the bioreactor comprises a first module comprising bioprocess controls and a second module comprising. the three-dimensional electrode array, wherein the second module is isolated from electrical interference from the first module. In some embodiments, the manufacturing occurs in the second module.

[0116] In some embodiments, manufacturing comprises at least one method of additive manufacturing. Additive biomanufacturing is the process of a joining biomaterials to make objects from three-dimensional model data, usually layer upon layer. To create tissues living cells are mixed with a carrier hydrogel, sometimes referred to as a bioink, and placed in precise, biomemetic, three-dimensional arrangements. The semi-liquid hydrogel-cell mixtures are cross-linked via chemical or physical processes to increase robustness of their three- dimensional arrangement.

[0117] Bioprinting is a common form of additive tissue biomanufacturing. In bioprinting, multiple printheads each containing a unique mixture of cells and hydrogels place each bioink layer by layer until a complete tissue is formed.

[0118] A second type of additive manufacturing is injection molding. Injection molding is a manufacturing process for producing parts by injecting liquids into a mold and subsequently solidifying the liquid so that it keeps the shape of the mold. Injection molding is a highly scalable alternative to three-dimensional bioprinting for the creation of tissue.

[0119] In some embodiments, manufacturing comprises at least one method of subtractive manufacturing. In some embodiments, subtractive manufacturing comprises the removal of a structure, the removal of a porogen or the removal of a bioink. In some embodiments, manufacturing comprises at least one method of additive manufacturing and one method of subtractive manufacturing.

[0120] An injection mold may be used to develop a three-dimensional engineered tissue. In some embodiments, the injection mold is used for biomaterials with ports to perfuse crosslinking reagent, subtractive manufacturing media, cells, bioprocess media, and other living and non-living components. In some embodiments, the injection mold comprises at least one plate that facilitates the removal of tissue from the injection molds.

[0121] Cells embedded within an engineered tissue require constant maintenance of critical cellular process components such as oxygen, nutrients, metabolites, and removal of byproducts. This maintenance requirement limits many cultured tissues to sizes supported by passive exchange and avascular cellular transport of critical components. For example, cellular aggregates, commonly referred to as organoids, or cell spheres, cannot generally exceed 1000 pm in diameter, before the centers begin to necrose due to poor maintenance of critical components. One method to maintain appropriate levels of critical cellular process components within tissues is the vascularization of tissues. Where vascularization is defined as the process of growing blood vessels into a tissue to improve oxygen and nutrient supply. One method of tissue vascularization comprises bioprinting. In some embodiments, the tissue is vascularized. In some embodiments, the tissue comprises at least one cell type. In some embodiments, the tissue comprises at least two cell types. In some embodiments, the tissue comprises at least three cell types as described herein.

[0122] In some embodiments, the plurality of cells comprises a neural progenitor cell, a stem cell, a primary tissue cell, a differentiated neuron, an astrocyte, an oligodendrocyte, a t- cell, a vascular cell, or a combination thereof. In some embodiments, the plurality of cells comprises a plurality of neural progenitor cells. In some embodiments, the plurality of cells comprises a plurality of stem cells. In some embodiments, the plurality of cells comprises a plurality of primary tissue cells. In some embodiments, the plurality of cells comprises a plurality of differentiated neurons. In some embodiments, the plurality of cells comprises a plurality of astrocytes. In some embodiments, the plurality of cells comprises a plurality of oligodendrocytes. In some embodiments, the plurality of cells comprises a plurality of t-cells. In some embodiments, the plurality of cells comprises a plurality of vascular cells. Vascular cells include, without limitations, endothelial cells, angioblasts, and smooth muscle-like cells. [0123] Vasculature is the blood vessels or arrangement of blood vessels in an organ. To enable the growth of tissues larger than that allowed by passive diffusion and avascular mechanisms, avascular tissues can be vascularized to enable the transport of oxygen and nutrients to cells. In-vivo neovasculature creation is referred to as either vasculogenesis, in which new blood vessels are formed from angioblasts (endothelial precursor cells) or angiogenesis, in which endothelial cells in existing blood vessels move and grow to allow new capillaries to form. In-vitro vascularization can be accomplished through the creation of hollow chambers seeded with successive washes of vascular cells (endothelial cells, angioblast, smooth muscle, and/or other associated cells). Together, this creates a tissue with an empty tube coated with the appropriate cell types. Over time, the cells become established and can spread into tissues allowing for vascularized delivery of critical components throughout the tissue. As methods described herein refer to a combination of cell types, perfusion techniques, and vessel lumens investigated using multi-factorial analysis the various methods collectively are referred to as vascularization.

[0124] Manufacturing may comprise adding a primary vascularized tissue directly into a cavity or plurality of cavities within a secondary tissue to combine the primary vascularized tissue and the secondary tissue. A matrix or a gel layer may be used to combine the primary vascularized tissue and the secondary tissue. Alternatively, the primary vascularized tissue and the secondary tissue are combined without a matrix or a gel layer.

[0125] Manufacturing may comprise adding avascular tissues directly into a cavity or plurality of cavities within a secondary tissue to combine the primary avascularized tissue and the secondary tissue. A matrix or a gel layer may be used to combine the primary avascularized tissue and the secondary tissue. Alternatively, the primary avascularized tissue and the secondary tissue are combined without a matrix or a gel layer. The method may additionally comprise scanning a plurality of resection cavities, and that biomanufacturing bespoke tissues in the shape of the scanned plurality of resection cavities.

[0126] Manufacturing may comprise an approach by which scaffolds are decellularized and/or seeding with cells. This method may comprise a process by which native tissues such as neural tissues from animal or human sources are stripped of their original cellular components while preserving the extracellular matrix structure. In some manufacturing methods the scaffold is seeded with relevant cells. In some embodiments, this method enables the creation of highly biomimetic, functionally mature neural tissues with native-like connectivity and synaptic organization.

B. Biocomputing

[0127] Biocomputers use biologically derived materials to perform computational functions. Both read and write capabilities are necessary to interface traditional silicon computation with cellular computation. In recent years, there has been a wave of Brain Machine Interfaces as well as use of in-vitro culture of neuronal cells in basic computational tasks. These in-vitro cultured cell brain machine interfaces may be conducted using surface level stimulation and recording of neural spikes using multi-electrode arrays. In these multielectrode arrays, each electrode can stimulate and record electric cellular messaging at physiologically relevant ranges.

[0128] Brain function relies on its three-dimensional environment. Brain tissues are highly heterogeneous and comprise many different cell types, including neurons, astrocytes, and oligodendrocytes. Each of these cells along with other characterized and uncharacterized cells are vital for proper brain function. For this reason it is likely necessary to combine many cell types to create an intact biomimetic brain tissue. To appropriately interface with viable three-dimensional tissues long term a bioreactor system as described herein can be used to can support general tissue growth, vascularized tissue, while separating electrical bioprocess stimulus from controlled electrical stimulus of the tissue. [0129] In some embodiments, the biocomputing system described herein is referred to alternatively as a biocomputer, or a HBANPU system. The Hybrid Bio- Al Neural Processing Unit (HBANPU) system described herein comprises a neuromorphic platform that integrates living neural tissue with reconfigurable digital hardware. In some embodiments, the Hybrid Bio Operating System (HBOS) merges Verilog-like hardware descriptions with Python-like high-level scripting through our Bio-Neural Programming Language (BNPL). This unified approach may permit dynamic FPGA reconfiguration in real time and allows stable modules to be exported as ASIC -ready designs. The system may continuously monitor neural activity, including advanced signal processing features, and adapts via a closed-loop feedback mechanism.

[0130] By unifying biological computation with flexible digital hardware, HBOS and BNPL may enable rapid prototyping, adaptive control, and efficient deployment of neuromorphic applications. This integration may allow for innovative applications in medical diagnostics, assistive communication, and Al-driven neuromorphic computing, establishing both neural tissue health and system scalability. In some embodiments, the biocomputing system described herein abstracts low-level hardware coding so that software engineers can develop high-level code in Python, while HBOS automatically translates it into Verilog for FPGA and ASIC programming.

[0131] Also described herein is a method of analyzing a tissue comprising a three- dimensional electrode array. In some embodiments, the method comprises additive and subtractive manufacture of the tissue enmeshed with three-dimensional electrode array in a bioreactor as described herein; sending training signals to the tissue in the second module; and receiving signals from the three-dimensional electrode array. In some embodiments, the engineered tissue and the three-dimensional electrode array comprise a three-dimensional biocomputing system. In some embodiments, the engineered tissue is greater than 1000 pm³. [0132] In certain aspects, the manufactured tissue and the embedded array may be for use in a method of biocomputing. In some embodiments, the three-dimensional microelectrode array is embedded into engineered tissues. In some embodiments, at least one electrode in the three-dimensional electrode array can stimulate and record electronic cellular messaging. In some embodiments, a plurality of bioprocess controls regulate a plurality of bioprocess parameters simultaneously with at least one electrode in the three dimensional electrode array stimulating and recording electronic cellular messaging.

[0133] In some embodiments, the three-dimensional microelectrode array comprises read and write capabilities. In some embodiments, the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues. In some embodiments, the three-dimensional microelectrode array with read and write capabilities is embedded into engineered tissues supported by engineered vascular networks. In some embodiments, the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues, supported by vascular networks. In some embodiments, the three- dimensional microelectrode array with read and write capabilities is embedded into engineered tissues supported by engineered avascular, active perfusion networks. In some embodiments, the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues, supported by avascular passive perfusion networks through the use of negative space in engineered tissues. [0134] In some embodiments, the method comprises manufacturing of the engineered tissue comprising a plurality of cells with the three-dimensional electrode array in a bioreactor system, with a three-dimensional electrode array capable of: sending relevant electrophysiology signals to the tissue; and receiving signals from the engineered tissue cells. In some embodiments, the methods further comprise receiving at least one signal from a plurality of cells in the engineered tissue. In some embodiments, the signal is received by the three-dimensional electrode array. In some embodiments, the signal is received by an external receiver.

[0135] The method may additionally comprise training of cells within an engineered tissue using microelectrode pulses at physiologically relevant ranges embedded within engineered tissues. In some embodiments, training of cells within an engineered tissue using microelectrode pulses at physiologically relevant ranges embedded within the tissues through the use of the three-dimensional arrangement of microelectrode arrays. In some embodiments, the three-dimensional electrode array emits a three-dimensional arrangement of electrical pulses to map networks of engineered tissues. In some embodiments, the method comprises reading data communication between a plurality of bioprocess controllers and a plurality of electric cell firings by multi-electrode arrays in a biomanufactured three- dimensional environment.

[0136] In some embodiments, the use of training and test phases with three-dimensional multi-electrode arrays and which train and evaluate the quality of the neural network. The method may further comprise controlling a bioprocess with a plurality of neural spikes. In some embodiments, the method comprises controlling a bioprocess with a neural spike pattern. [0137] In some embodiments, the method comprises storing information via in-silico and in vitro neural networks within a manufactured biological three-dimensional environment exceeding >1000 pm cross-section. In some embodiments, the method comprises writing, reading, and interpreting cell electrical firings with an artificial neural network within a manufactured biological three-dimensional environment exceeding >1000 pm cross-section. In some embodiments, multiple organoids or tissues are used together in biocomputation and information storage. In some embodiments, cells or tissues smaller than >1000 pm are utilized.

[0138] The pulse train, strength, frequency, pattern, duration, waveform, amplitude, shape, and physical location may be used for read, write, and storage of information in a biocomputation environment. Gene expression may be correlated to neuronal activity over time and three-dimensional space to indicate learning patterns and mechanisms.

[0139] In some cases, the devices or systems described herein may implement computing systems. For example, these computing systems may be implemented to serve as controllers for the devices or systems disclosed herein (e.g., controlling nutrient flow, controlling pH levels, controlling neuromodulators, controlling temperature, etc.). The computer systems disclosed herein may implement one or more non-transitory computer-readable media comprising computer-executable instructions that, when executed by at least one processor, cause the at least one processor to perform the methods described herein.

[0140] In some cases, the systems, the methods, the computer-readable media, and the techniques disclosed herein include one or more non-transitory computer readable storage media encoded with a program including instructions executable by the operating system of an optionally networked computing device. In further cases, a computer readable storage medium is a tangible component of a computing device. In still further cases, a computer readable storage medium is optionally removable from a computing device. In some cases, a computer readable storage medium includes, by way of non-limiting examples, CD-ROMs, DVDs, flash memory devices, solid state memory, magnetic disk drives, magnetic tape drives, optical disk drives, distributed computing systems including cloud computing systems and services, and the like. In some cases, the program and instructions are permanently, substantially permanently, semi -permanently, or non-transitorily encoded on the media.

[0141] In some cases, the systems, the methods, the computer-readable media, and the techniques disclosed herein include at least one computer program, or use of the same. A computer program includes a sequence of instructions, executable by one or more processor(s) of the computing device’s CPU, written to perform a specified task. Computer readable instructions may be implemented as program modules, such as functions, objects, Application Programming Interfaces (APIs), computing data structures, and the like, that perform particular tasks or implement particular abstract data types. A computer program may be written in various versions of various languages.

[0142] The functionality of the computer readable instructions may be combined or distributed in various ways across various environments. In some cases, a computer program comprises one sequence of instructions. In some cases, a computer program comprises a plurality of sequences of instructions. In some cases, a computer program is provided from one location. In some cases, a computer program is provided from a plurality of locations. In some cases, a computer program includes one or more software modules. In some cases, a computer program includes, in part or in whole, one or more web applications, one or more mobile applications, one or more standalone applications, one or more web browser plug-ins, extensions, add-ins, or add-ons, or combinations thereof.

[0143] In some cases, the systems, the methods, the computer-readable media, and the techniques disclosed herein include software, server, or database modules, or use of the same. Software modules may be created by techniques using machines, software, and languages. The software modules disclosed herein are implemented in a multitude of ways. In some cases, a software module comprises a file, a section of code, a programming object, a programming structure, a distributed computing resource, a cloud computing resource, or combinations thereof. In some cases, a software module comprises a plurality of files, a plurality of sections of code, a plurality of programming objects, a plurality of programming structures, a plurality of distributed computing resources, a plurality of cloud computing resources, or combinations thereof. In some cases, the one or more software modules comprise, by way of non-limiting examples, a web application, a mobile application, a standalone application, and a distributed or cloud computing application. In some cases, software modules are in one computer program or application. In some cases, software modules are in more than one computer program or application. In some cases, software modules are hosted on one machine. In some cases, software modules are hosted on more than one machine. In some cases, software modules are hosted on a distributed computing platform such as a cloud computing platform. In some cases, software modules are hosted on one or more machines in one location. In some cases, software modules are hosted on one or more machines in more than one location.

[0144] In some cases, the systems, the methods, the computer-readable media, and the techniques disclosed herein include one or more databases, or use of the same. In some cases, various databases may be suitable for storage and retrieval of one or more of (i) wearable data, (ii) responses to health queries, (iii) geographic data, etc., one or more of which may be historical, present, or future data or information. In some cases, suitable databases include, by way of non-limiting examples, relational databases, non-relational databases, object oriented databases, object databases, entity -relationship model databases, associative databases, XML databases, document oriented databases, and graph databases. Further non-limiting examples include SQL, PostgreSQL, MySQL, Oracle, DB2, Sybase, and MongoDB. In some cases, a database is Internet-based. In further cases, a database is web-based. In still further cases, a database is cloud computing-based. In a particular case, a database is a distributed database. In other cases, a database is based on one or more local computer storage devices.

1. Hybrid Bio Operating System (HBOS) & Bio-Neural Programming Language (BNPL)

[0145] Described herein are methods of using a custom operating system and programming language to accomplish computational tasks. In some embodiments, these systems are referred to as a Hybrid Bio Operating System (HBOS), which merges Verilog- like hardware descriptions with Python-like high-level scripting through a custom Bio-Neural Programming Language (BNPL).

[0146] In some embodiments, the HBOS manages neural tasks, schedules spike-based computations, and securely coordinates communication between neural tissue and FGPA/ASIC devices. In some embodiments, the HBOS automates the conversion of high- level code into hardware configurations, permitting rapid reconfiguration and secure operation without developers having to deal with low-level details. In some embodiments, the HBOS architecture includes an interpreter core, kernel services, security, and logging.

[0147] In some embodiments, an interpreter core reads high-level BNPL or Python-like scripts and translates them into either FPGA configurations (partial bitstreams) or ASIC- ready Verilog modules. The interpreter may read the script, identifying “pure Python-like” commands (e.g., loops, if-statements) and “Verilog-like” hardware blocks. For FPGA usage, the interpreter may compile these blocks into partial bitstreams. For ASIC generation, the interpreter may output Verilog modules that can be synthesized using standard tools.

[0148] In some embodiments, inputs and outputs at the Python-like and Verilog-like level include the automatic generation and interpretation of neural spike waveforms, patterns, shapes, frequencies, and other metrics. In some embodiments, these neural spike features are automatically tested, updated, and adjusted as necessary using self-improvement AI/ML practices. In some embodiments, computational tasks, such as AI/ML tasks or interpreting basic scripts are fed to the system through this Python-like, or Verilog-like interface.

[0149] In some embodiments, the system is compatible with standard AI/ML frameworks (PyTorch, TensorFlow, JAX, MXNet, ONNX Runtime, Keras, and Hugging Face Transformers), data science and computational platforms (Jupy ter Notebooks, Google Colab, Apache Zeppelin, Databricks, and RStudio), scientific computing and numerical libraries (NumPy, SciPy, MATLAB, Julia, Wolfram Mathematica, GNU Octave, and SymPy), big data and distributed computing frameworks (Apache Spark, Dask, Ray, TensorFlow Serving, and Kafka), embedded and edge computing platforms (NVIDIA Jetson, Arduino, Raspberry Pi, ESP32, and TinyML), high-performance computing (HPC) and GPU computing environments (CUD A, OpenCL, MPI, and SLURM), cloud computing and virtualization technologies (AWS SageMaker, Google Cloud Al, Azure Machine Learning, Docker, and Kubernetes), and general-purpose programming platforms and integrated development environments (VS Code, PyCharm, Eclipse, JetBrains Intelli J, Xcode, Visual Studio, and Emacs/Vim).

[0150] In some embodiments, kernel services manage memory, scheduling, and real-time event handling (e.g., neural stress signals). It may also log all critical events, enabling developers to audit or debug system behavior.

[0151] In some embodiments, security and logging guarantees that any hardware reconfiguration commands are authenticated. In some embodiments, the system uses blockchain or tamper-proof logs for critical reconfiguration events.

[0152] In some embodiments, the BNPL abstracts complex DSP and hardware control tasks into simple, human-readable commands that blend Python-like logic with embedded Verilog-like blocks. In some instances, this allows software developers to specify algorithmic behavior in a familiar Python-like syntax or Verilog like syntax depending on their expertise. [0153] In some embodiments, the software performs automatic code generation. In some instances, software engineers can write BNPL scripts in a familiar Python-like style. In some instances, HBOS’s interpreter performs a detailed conversion, generating complete Verilog code with proper ports and timing constraints. For example, in some instances, programmers can use a convert bnpl to verilog function to simulate how HBOS parses a high-level BNPL script written in Python-like syntax. In some instances, this system detects module definitions and function calls (like ADAPTIVE FILTER) and produces a Verilog module template. In some instances, the conversion process generates complete, synthesizable Verilog code with proper port definitions and timing constraints. [0154] In some embodiments, automatic conversion allows software engineers to develop neuromorphic applications using familiar Python-like code. It may allow software engineers to focus on high-level algorithm design while HBOS handles the low-level implementation required for programming FPGAs and ASICs.

[0155] In some instances, hardware oriented BNPL exposes the low-level hardware logic (Verilog-like e.g., always comb blocks) so that hardware engineers can directly control and optimize the module behavior. In some embodiments, netlist generation, place-and-route, timing analysis, and power optimization are used to optimize the module behavior. These may help transform a high-level hardware description (like Verilog) into a final, manufacturable design (ASIC) or a configurable bitstream (for an FPGA).

[0156] In some embodiments, a netlist is used. A netlist may comprise a list or description of all the components (like logic gates, flip-flops, or memory blocks) in a digital circuit and how they are connected to each other (the “wires” or signals). In some instances, Verilog is used to create a netlist, allowing the user and/or system to create a precise blueprint of what the final circuit is at the logic level. In some instances, Electronic Design Automation tools then use this netlist to generate the actual physical layout of the chip or FPGA configuration.

[0157] In some embodiments, place-and-route (P&R) is used. In some instances, P&R is the process of deciding where each component from the netlist should physically go on the chip (the “place” step) and how the interconnecting wires should be drawn between them (the “route” step). A netlist may comprise hundreds, or even millions of small components. In some instances, there components need to be placed in a way that fits the overall shape (the chip) and connect them with “paths” (wires) that minimize delay and power consumption. In some instances, Electronic Design Automation tools automatically figure out how to pack everything together efficiently while meeting performance requirements.

[0158] In some embodiments, timing analysis is used. In some instances, timing analysis checks whether signals travel through the circuit fast enough to meet the required clock speeds or operational deadlines. In some instances, timing analysis looks at the time it takes for signals to move from one register (or gate) to another and affirms that the design won’t malfunction at the target clock frequency, f the circuit doesn’t meet timing requirements, the chip may produce incorrect results or fail to run at the intended speed. In some instances, timing analysis ensures the design is reliable and operates correctly at the specified clock rate. [0159] In some embodiments, a power optimization approach is employed. In some instances, this process involves adjusting the design, at both the logical and physical levels, to reduce the amount of electricity the chip consumes. In some instances, this might include turning off unused sections, lowering voltages where possible, or minimizing unnecessary signal toggling. In some instances, chips with lower power consumption generate less heat, require smaller (or no) cooling solutions, and can extend battery life in portable devices. For the bio-hybrid systems described herein, efficient power usage is also important for preventing damage to living neural tissue and maintaining a stable operating environment.

2. Advanced signal processing

[0160] Also described herein are methods for advanced signal processing. In some embodiments, the biocomputing system described herein fuses living neural cells/tissue with digital processing to deliver real-time, energy-efficient, and adaptive computation. In some instances, a core component of the biocomputing system's function is signal processing. In some embodiments, signal processing includes cleaning, formatting, and monitoring neural signals, alongside other innovative bio-digital interactions. In some embodiments, by combining living neural tissue with digital electronics, the biocomputing system can create a computing system that adapts more easily, uses less energy, and manages many tasks at once. [0161] In some embodiments, developers can use the described systems to write high- level, human-readable scripts that can define signal processing pipelines. For instance, in some embodiments, scripts may specify which wavelet transforms or adaptive filters to use; set thresholds and weights; determine how much each metric (e.g., gamma power, Lyapunov exponent) influences the workload score; trigger feedback loops; orchestrate additional clusters; trigger a microfluidic response; generate new spike trains, patterns, shapes, codes, or machine learning tasks.

[0162] In some embodiments, the disclosed system employs conditioning and formatting of signals. In some instances, stimulating and recording devices such as MEAs are utilized to record both spike events and low-frequency local field potentials (LFPs). In some instances, high-resolution Analog-to-Digital Converters (ADCs) digitize these analog signals, producing a continuous stream of digital data. In some instances, to refine the captured signals, an initial digital filtering stage is applied, often incorporating a notch filter at 50 or 60 Hz. In some instances, notch filters are designed to eliminate a narrow band of frequencies, such as the 50/60 Hz hum from power lines, ensuring that the neural signals remain free of electrical interference. [0163] In some embodiments, to dynamically adapt to changing noise conditions, the system employs FPGA-based modules that measure noise and signal variance in real time. In some embodiments, these modules update filter coefficients within microseconds to milliseconds to compensate for new noise sources or shifts in neural signal patterns. In some embodiments, a neural feedback loop is integrated into the system, wherein the filtering parameters are refined based on detected stress indicators in the neural tissue, such as chaotic patterns or amplitude drops. In some instances, this ensures that signals are neither overfiltered nor under-filtered, preserving their integrity for analysis.

[0164] In some instances, once filtered, the signals are standardized into uniform data structures to maintain consistent sample rates, labeling, and formatting. In some instances, the uniform data structures include time-stamped frames, fixed-length buffers, structured packets, and multidimensional arrays. These formats enable efficient downstream processing, ensuring compatibility across hardware and software components. This approach allows for bidirectional communication between the neural tissue and the digital filtering system, providing dynamic optimization that surpasses static filtering techniques.

[0165] Regardless of the exact format, one goal is to make sure every piece of hardware or software expects the same data layout, so the neural signals can be processed, stored, or analyzed without confusion. In some instances, the two-way communication between the tissue and the digital filtering system assures that the neural signals are not only clean but also dynamically optimized in real time, a significant improvement over static filtering techniques.

[0166] In some embodiments, the bioreactor system uses Local Field Potential (LFP) monitoring. In some instances, LFPs represent aggregated electrical signals from groups of neurons, capturing low-frequency components of neural activity (1-200 Hz). In some instances, these signals offer insights into the broader neural dynamics within a given tissue region. LFP isolation is achieved through low-pass filtering, which removes high-frequency spikes and noise, leaving only the slow oscillations characteristic of LFPs. In some instances, advanced signal processing methods such as wavelet and Fourier transforms are then applied to break down the LFP signal into distinct frequency bands (e.g., theta, alpha, gamma), allowing for precise pattern detection.

[0167] In some embodiments, a low-pass filtering method is utilized that lets lower- frequency signals (like slow neural waves) pass through while blocking higher frequency noise. In some instances, this step helps focus on important slow waves (LFPs) without the clutter of high frequency noise. [0168] In some embodiments, a wavelet transform takes a signal and breaks it down into tiny wave “packets,” revealing what frequencies are present and when they occur. In some instances, changes in neural signals are analyzed over time.

[0169] In some embodiments, a Fourier Transform splits a signal into its fundamental frequencies. In some instances, the Fourier Transform helps researchers and systems understand if a signal has strong components in, for example, the gamma range (30-100 Hz). These Fourier Transforms matter because gamma band neural oscillations typically range from 30-100 Hz (up to 150 Hz) and are linked to cognitive functions such as attention and memory. Elevated gamma can indicate focused attention. However, excess gamma may signal pathological states such as epileptiform activity, migraines, neurological stress, or psychiatric conditions.

[0170] In some instances, by monitoring the data pipeline excessive gamma power can be flagged as a potential indicator of tissue stress, triggering interventions (e.g., reduced stimulation, switching computation to underutilized neural tissue clusters, or microfluidic adjustments). In some instances, early detection of abnormal gamma activity allows the system to intervene before irreversible tissue damage occurs. Thus, in some instances, monitoring gamma activity permits sustained performance and extending the lifespan of the neural tissue.

[0171] Complexity Metrics such as Recurrence Quantification Analysis (RQA), Hurst exponent, sample entropy, and connectivity measures may be used to assess tissue stability and computational capacity. Shifts in these metrics may indicate changes in neural health or workload distribution.

[0172] In some embodiments, the system enables RQA analysis for assessing the dynamics of neural signals by measuring the recurrence of signal patterns over time. In some instances, by analyzing recurrences within a phase-space representation of neural activity, RQA can distinguish between structured neural communication and random noise. In some instances, RQA measures, such as determinism, laminarity, and entropy, help quantify the stability and complexity of neural activity. In some instances, a decline in recurrence metrics may signal neural degradation, while an increase in determinism may indicate an emerging structured computational state. In some instances, the adaptability of RQA makes it particularly useful for monitoring neural plasticity and early-stage dysfunctions in neural computation.

[0173] In some embodiments, the system enables Hurst Exponent analysis, by which a statistical analysis evaluates the long-term memory and persistence of neural signals. In some instances, values greater than 0.5 suggest that past signal patterns influence future activity, indicative of stable computational processes and efficient memory encoding. Conversely, in some instances, values below 0.5 imply anti -persistent behavior, where past trends are likely to reverse, potentially signaling a loss of structured processing. In some embodiments, by tracking fluctuations in the Hurst exponent over time, the system can infer the evolving computational strategies of the neural tissue and detect anomalies that may indicate cognitive fatigue or inefficiency in information retention.

[0174] In some embodiments, the system enables sample entropy quantification and evaluation. In some instances, sample entropy is used to quantify the predictability of a neural signal by evaluating the likelihood that similar patterns will repeat at varying scales. In some instances, low entropy suggests excessive regularity, which may indicate rigid or overly synchronous activity, whereas high entropy points to chaotic and disorganized signaling. In some instances, an optimal range of sample entropy is indicative of a balanced neural state, supporting adaptive computation. In some instances, this metric is particularly useful for assessing the resilience of neural tissue to external perturbations, as sudden increases or decreases in entropy may reflect shifts in computational efficiency or impending instability. [0175] In some embodiments, the system enables quantification and analysis of connectivity metrics. In some instances, connectivity metrics provide a measure of how well different neural regions interact, reflecting the functional architecture of the tissue. In some instances, techniques such as coherence analysis, phase-locking value (PLV), and Granger causality are employed to map inter-regional communication pathways. In some instances, strong, stable connectivity between regions suggests efficient information transfer and distributed computation, whereas declining connectivity may indicate emerging dysfunction. In some instances, these measures allow for real-time assessment of tissue integrity, aiding in task allocation and optimizing neural workload distribution.

[0176] In some embodiments, the system enables quantification and analysis of unified health-load-synergy metrics. In some instances, a unified health-load-synergy metric employs a composite scoring system, integrating RQA, Hurst exponent, sample entropy, and connectivity metrics into a single score used by the global FPGA for real-time scheduling. In some instances, predictive analytics leverage neural networks or advanced algorithms to forecast cluster stress based on historical patterns, enabling proactive load redistribution. 3. Communication and Dynamic Neural Tissue Workload Management

[0177] Described herein are some embodiments of communication protocols for on-board and between board communication and some embodiments of their uses including their use as it relates to dynamic neural tissue workload management.

[0001] In some embodiments, each cell/tissue processing cluster is a modular unit of living neural tissue that processes information locally within a culture well. In some embodiments, these clusters are connected via on-chip communication buses. In some embodiments, specialized high-speed data links (such as LVDS and SERDES) allow each cluster and/or cluster segment to send its processed signal, like wavelet outputs, Fourier transforms, and complexity metrics, to a global FGPA controller. In some instances, this communication network makes sure that data from every cluster is available for overall system analysis. Herein, reference to a bus may encompass one or more digital signal lines serving a common function, where appropriate. Bus may be any of several types of bus structures including, but not limited to, a memory bus, a memory controller, a peripheral bus, a local bus, and any combinations thereof, using any of a variety of bus architectures. As an example and not by way of limitation, such architectures include an Industry Standard Architecture (ISA) bus, an Enhanced ISA (EISA) bus, a Micro Channel Architecture (MCA) bus, a Video Electronics Standards Association local bus (VLB), a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCLX) bus, an Accelerated Graphics Port (AGP) bus, HyperTransport (HTX) bus, serial advanced technology attachment (SATA) bus, and any combinations thereof.

[0178] In some embodiments, the system uses Low-Voltage Differential Signaling (LVDS). LVDS is a protocol that transmits data by comparing the voltage difference between two wires. It minimizes noise and supports high-speed, reliable communication between onboard components. In some instances, if clusters or modules are placed on different parts of a circuit board or on separate chips, use LVDS lines to communicate. These lines are designed to move large amounts of data reliably and at high speeds.

[0179] In some embodiments, the system uses Serializer/Deserializer (SERDES) Links. SERDES is a technology that converts parallel data into a high-speed serial stream and then back into parallel data. It reduces the number of physical connections and preserves signal integrity, ensuring efficient data transfer. In some instances, where even higher data rates or longer distance communication is needed, SERDES links are used. In some instances, the SERDES links convert parallel data into a high-speed serial stream and then back again, often achieving gigabit speeds.

[0180] In some embodiments, the system uses Advanced extensible Interface (AXI) Bus. AXI is a widely used on-chip communication protocol that is part of the ARM Advanced Microcontroller Bus Architecture (AMB A) specification. AXI is defined by ARM Holdings. It provides a standardized way for master components (such as processors, Direct Memory Access engines, or custom Field Programmable Gate Array logic) to read and write data to slave components (such as memory controllers or peripheral registers). FPGAs because it supports high throughput, low latency, and independent channels for reading and writing. In some embodiments, when clusters are located on the same physical FPGA chip, they can use an on-chip bus to transfer data very quickly and with very little delay. Examples of the utility of AXI in this context are described herein.

[0181] In some embodiments, AXI comprises distinct channels for addresses and data, allowing simultaneous read and write operations.

[0182] In some embodiments, AXI supports burst transfers of data, meaning a master can read or write multiple words in one address transaction. In this context a “master” is any hardware component (such as a processor, direct memory access engine, or custom FPGA logic) that initiates transactions on the bus. This means the master sends out address requests and commands (read or write) to “slave” components, which respond by providing or accepting data. In some instances, the master is the driver of the communication, telling the bus where to read or write and how much data to transfer.

[0183] In some instances, AXI uses valid and ready signals such as handshaking on each channel to coordinate data transfers without requiring strict timing constraints.

[0184] In some instances, AXi provides scalability, as it can be configured with different data widths (for example, 32-bit, 64-bit, 128-bit, etc.) to meet varying performance needs.

[0185] In some instances, AXI allows the system to communicate using common intellectual property cores. Many modern FPGA vendors provide AXI-based intellectual property blocks (such as Direct Memory Access controllers and memory interfaces) to simplify system integration.

[0186] In some embodiments, within the HBANPU, AXI is used to move partial computation states (such as matrix data) between neural clusters or between clusters and memory. In some embodiments, for local clusters on the same FPGA chip, AXI-based interconnects allow fast data transfers. In some embodiments, with remote modules, the Field Programmable Gate Array may bridge AXI signals over Low Voltage Differential Signaling (LVDS) or Serializer/Deserializer (SERDES) links.

[0187] In some embodiments, other protocols are used such as an Optical/TCP-IP Links employed at the top layer for secure external connectivity. In some embodiments, analog signal lines are used by the digital-to-biological encoder to deliver precise stimulation pulses to neural tissue.

[0188] In some embodiments, to choose the right communication protocol, the FPGA refers to an internal location mapping of each cluster. In some embodiments, if both the source and destination clusters are located on the same chip, the FPGA uses an on-chip interconnect such as the AXI bus for low-latency, high-bandwidth transfers. In some embodiments, if the clusters are on different chips or modules, the FPGA selects LVDS or serializer/deserializer SERDES channels, which are better suited for long-distance, high- throughput communication. In some embodiments, the system uses these communication channels to perform adjustments in stimulation and nutrient delivery to help maintain efficient handling of new tasks while preserving tissue health.

[0189] In some embodiments, the FPGA can be reprogrammed in real time based on instructions from HBOS. In some embodiments, if neural tissue shows signs of stress, HBOS can send new parameters or even update the DSP routines on the fly, so that the system remains responsive and balanced. In some instances, if the analysis (like an increase in excessive gamma power or a high Lyapunov exponent) indicates potential tissue stress, the global FPGA can communicate with a microcontroller that adjusts microfluidic pump rates. In some embodiments, this closed-loop feedback mechanism helps maintain optimal nutrient and oxygen delivery.

[0190] In some embodiments, the FPGA-based system continuously redistributes computational tasks across neural clusters, optimizing workload distribution and prolonging tissue lifespan. In some instances, by monitoring synaptic plasticity markers and firing pattern signatures, the system assesses computational state and dynamically reallocates tasks accordingly. In some embodiments, reinforcement protocols, including controlled stimulation pulses and microfluidic adjustments, further optimize computational efficiency based on task priority and tissue condition. This kind of dynamic balancing does not appear in conventional systems.

[0191] In some embodiments, this real-time task redistribution mechanism allows the system to adapt dynamically, distinguishing it from conventional static load-balancing methods. In some embodiments, by integrating workload, health, and cohesion metrics, the system maximizes resource utilization while ensuring tissue stability.

[0192] In some embodiments, the global FPGA monitors load and health metrics across neural clusters, including spike features, gamma levels, complexity scores, and connectivity measures. In some instances, if a cluster becomes overloaded, partial computations are checkpointed and transferred to underutilized clusters via high-speed links. In some embodiments, the FPGA selects an appropriate communication protocol based on cluster locations.

[0193] In some embodiments, finite state machines (FSMs) are a way of designing digital logic so that the system is always in one of a limited number of states, and it transitions from one state to another based on inputs or conditions. In some embodiments, in FPGA(s), FSMs are used to control the step-by-step sequences of operations (like checking load scores, issuing checkpoint commands, or deciding to migrate tasks). In some embodiments, each state in the FSM represents a distinct “phase” of the control logic, and the transitions define when the system moves to the next phase. This structured approach helps guarantee that complex operations happen in the correct order without conflicts.

[0194] In some embodiments, the global FPGA maintains a “Task Table,” listing each active sub-task (e.g., partial matrix multiplication for an LLM, partial protein conformation states).

[0195] In some embodiments, each entry includes: Current Cluster Assignment, Task Priority (e.g., real-time query vs. background compute), Progress/State (how far along the cluster is in solving the sub-task). In some embodiments, if the global FPGA sees that some clusters are idle or lightly loaded (under threshold) while others are nearing overload, it prepares to reassign sub-tasks. In some embodiments, other metrics include Load Score (indicates how busy a cluster is with current tasks), Health Score (reflects tissue well-being by measuring stable vs. chaotic signals)), Utilization Threshold, (quantifies if a cluster’s load score exceeds a preset limit).

[0196] In some embodiments, the system uses an identified overloaded cluster metric. In some embodiments, the system calculates and utilizes this metric by continuously monitoring the computational load and health metrics of each cluster using the global FPGA. In some instances, when a cluster exhibits a high load score — indicative of excessive computational demand — or a low health score due to irregular signal patterns such as chaotic neuronal activity or saturating spike rates, the FPGA identifies it as overloaded. In some embodiments, the system then initiates a redistribution process to optimize workload balance across available resources.

[0197] In some embodiments, upon identifying an overloaded cluster, the FPGA issues a request for the cluster to “checkpoint” its partial computational state before offloading the task. This checkpointing process may involve storing relevant intermediate data, such as partially computed matrices in large language model (LLM) operations or folding state information in protein modeling, into a dedicated digital memory unit. In some embodiments, the memory unit may include an attached solid-state drive (SSD) or high-speed randomaccess memory (RAM) to ensure rapid data retrieval and transfer.

[0198] In some embodiments, the FPGA references the Task Table, a dynamically updated record of cluster workloads and health statuses, to locate an underutilized cluster. The target cluster is selected based on its low computational load score and a stable health score, ensuring that it has sufficient processing capacity to take on the redistributed task. This process optimizes the utilization of available computing resources while maintaining the stability and efficiency of the system.

[0199] In some embodiments, to facilitate the movement of computational state data between clusters, the FPGA employs a specialized communication architecture that integrates high-speed LVDS and SERDES links alongside an on-chip AXI bus to transfer a partial state. For intra-chip transfers between clusters residing on the same physical substrate, an AXI interconnect or multi-lane bus is used to maintain low-latency data movement. In some embodiments, when transferring data between clusters located on separate chips or modules, the system utilizes LVDS/SERDES channels to achieve reliable, high-throughput communication. In some embodiments, the partial state transfer process is managed by the FPGA’s Load Manager or a dedicated “Checkpoint & Transfer” hardware block, which initiates a DMA transaction to migrate the data to the target cluster. In some embodiments, the receiving cluster then updates its local digital signal processing pipeline or memory with the transferred state before resuming computation.

[0200] In some embodiments, the system may automatically adjust stimulation & nutrients in the dynamic neural tissue workload management context. In some scenarios where the newly assigned sub-task is computationally intensive or time-sensitive, the FPGA dynamically adjusts the physiological conditions of the receiving cluster. If a cluster requires increased neural activity, the FPGA incrementally modulates the amplitude or frequency of electrical stimulation pulses to optimize processing speed. Additionally, in some embodiments, where biomarker signals indicate rising cellular stress or metabolic strain, the system issues microfluidic control commands to enhance the delivery of essential nutrients or oxygen to the affected cluster. In some instances, these real-time adjustments ensure computational stability and maintain neural viability under varying workloads.

[0201] In some embodiments, once the underutilized cluster receives the transferred partial state, it resumes the computation from the last checkpointed position, effectively continuing the process without loss of prior work. In some instances, simultaneously, the previously overloaded cluster experiences a reduced workload, allowing it to stabilize and recover from computational stress. In some instances, this dynamic workload management enhances overall system resilience and efficiency.

[0202] In some embodiments, upon successful completion of a computational task, the responsible cluster generates a completion signal, which is transmitted back to the FPGA. In some embodiments, the FPGA captures this signal and updates the Task Table and Load Manager to reflect the completion status. In some embodiments, this signaling mechanism ensures that the system maintains real-time awareness of task progress and resource availability.

[0203] In some embodiments, the computed output — such as a fully processed matrix from an LLM inference or a resolved protein folding structure — is encapsulated into a standardized data structure containing essential metadata. In some embodiments, this metadata may include time stamps, cluster identifiers, processing channel labels, and additional annotations for downstream applications. In some embodiments, the structured output is then relayed to an AI-Assisted Post-Processing Unit for further refinement or analysis, or alternatively, it may be stored in an on-chip memory unit or SSD for retrieval by higher-level computing processes.

[0204] In some embodiments, following task completion, the FPGA reassesses the computational load and health metrics across all clusters to determine the optimal redistribution of future tasks. If the completed task was part of a larger computational pipeline, the FPGA may allocate the next sub-task to the newly available cluster or reassign additional workloads from the Task Table. In some embodiments, the system dynamically recalibrates electrical stimulation and microfluidic support in response to the updated cluster status, ensuring sustained operational efficiency and longevity of the neural computational network.

[0205] In some embodiments, the successful completion of a task and any associated changes in cluster performance are reported to high-level software layers, such as the BNPL scripts operating on the Hybrid Bio Operating System (HBOS). In some embodiments, this feedback loop enables adaptive scheduling, allowing the system to make real-time modifications to task sequencing, workload distribution, and neural stimulation protocols. In some embodiments, the interaction between hardware-level execution and software-driven task management ensures an integrated and responsive computational framework.

[0206] In some embodiments, this framework matters because reallocation of tasks minimizes downtime. In some instances, this prevents any single cluster from being overburdened, permitting continuous high-performance operation. This example demonstrates how the software-level BNPL instructions can map onto hardware-level signals and finite state machines in an FPGA environment, achieving dynamic load shifting in real time.

4. Closed-loop microfluidic feedback and advanced chemical delivery:

[0207] In some embodiments, the integration of multi-chemical delivery with closed-loop feedback may provide control over the neural tissue environment, setting the stage for robust, adaptive computation that responds to both digital and biochemical signals. In some embodiments, by maintaining an optimal biochemical environment, the system supports longterm tissue viability and improved computational accuracy. In some embodiments, the neural signal processing pipeline identifies anomalies such as excessive gamma or chaotic signals. In response, commands may be sent to microcontrollers to adjust nutrient and oxygen flow, reverting to normal once anomalies subside. The system may also expand on microfluidic techniques by enabling multi -chemi cal delivery, allowing precise control over neurotransmitters such as glutamate and GABA, as well as neuromodulators like dopamine. In some embodiments, tissue remodeling agents, including growth factors and scaffolding molecules, facilitate regeneration and neural connectivity modifications. In some embodiments, precision pH and chemical balancing ensure real-time monitoring and adjustment of pH levels to maintain optimal neural firing conditions.

5. Automated tissue calibration & training:

[0208] In some embodiments, automated tissue calibration and training involve selfcalibrating protocols that gradually introduce tasks to the tissue while monitoring complexity metrics to establish an optimal operating zone. In some embodiments, adaptive stimulation ramp-up incrementally increases stimulation intensity while adjusting microfluidic flow to train the tissue effectively. In some embodiments, plasticity control is achieved through specific protocols such as targeted electrical stimulation, lower frequency pulses, Spike- Timing-Dependent Plasticity (STDP), chemical or pharmacological modulation, optogenetic methods, and reinforcement feedback, which encourage synaptic plasticity in targeted regions while stabilizing others.

6. Self-healing and automatic reassignment

[0209] In some embodiments, self-healing and automatic reassignment mechanisms detect damage through persistent abnormal signals, prompting the system to isolate both data and media flow to and from irreversibly damaged clusters and permanently reassign their tasks to healthy clusters. In some embodiments, the tasks from an irreversibly damaged cluster would be assigned to a different portion of the tissue. In some embodiments, regenerative interventions are employed by delivering protective chemicals and growth factors to stimulate tissue regeneration. In some embodiments, regenerative interventions actively maintain and regenerate living tissue through adaptive control mechanisms. In some instances, the self-healing architecture of the Hybrid Bio-AI Neural Processing Unit (HBANPU) extends tissue lifespan and ensures computational stability, addressing a fundamental limitation of traditional neuromorphic systems that lack self-repair capabilities. These protocols represent a major advancement in neuromorphic computing by not only utilizing living tissue for computation but also actively maintaining and regenerating it through adaptive control mechanisms.

7. Multi-modal integration

[0210] The biocomputing system described herein has numerous applications in robotics, AR/VR, and medical diagnostics. In some embodiments, integrated sensory inputs allow realtime video, audio, and other sensory signals to feed into neural clusters, expanding the range of applications. In some embodiments, neural tissue could be specialized to process these types of inputs, improving system performance.

[0211] In some embodiments, secure and encrypted bio-compute sessions are enabled by a trusted execution environment, utilizing specialized FPGA/ASIC routines to encrypt data exchanged between the digital controller and neural tissue. In some embodiments, biometric authentication leverages unique neural fingerprints, derived from LFP patterns, to authenticate the health and identity of tissue clusters before processing sensitive tasks.

[0212] In some embodiments, an electrical cutoff is triggered if chaotic signals exceed thresholds, chemical injection of protective agents is administered, and software-based hibernation modes are activated for stressed clusters. In some embodiments, digital signatures are applied to BNPL scripts to prevent unauthorized reconfiguration. In some instances, this involves preparing a script, generating a cryptographic hash, signing it with a private key, and verifying it through FPGA or microcontroller decryption. In some instances, by ensuring that only authorized developers can sign scripts, the system prevents tampering, such as unauthorized modifications that could compromise stimulation protocols.

[0213] In some embodiments, digital signatures allow the HB ANPU to verify that BNPL scripts come from a trusted source and have not been altered, thereby protecting against unauthorized changes or malicious commands.

8. Thermal management

[0214] In some embodiments, thermal management is achieved through thermo- responsive microfluidics, which dynamically deliver cooler fluids to hotspots, and smart heat sinks with integrated temperature sensors that adjust cooling paths in real time. In some instances, the combination of secure computation, multi-layer safety protocols, and active thermal management creates a resilient system capable of operating reliably in demanding environments. Many conventional neuromorphic Al systems lack an equivalent level of cybersecurity, making HBANPU’s secure and authenticated bio-compute operations a critical advancement.

9. Direct communication between neural clusters

[0215] In some embodiments, direct communication between neural clusters is facilitated through a peer-to-peer protocol that allows adjacent clusters to exchange partial states directly. Additionally, in some embodiments, microfluidic signaling enables adaptive routing of microfluidic channels, allowing nearby clusters to share resources as needed. These advanced scheduling and communication methods provide neuromorphic systems with an unprecedented level of flexibility and robustness, ensuring smooth task distribution and preventing failures in any single area. In some embodiments, optimized task distribution and direct inter-cluster communication maximize efficiency and reliability, which is crucial for applications requiring continuous high-performance computation.

10. Bi-directional weight updating

[0216] In some embodiments, bi-directional weight updating is employed. In some instances, this concept describes a system by which algorithms adjust synaptic weights based on discrepancies between expected and actual neural outputs, permitting in-tissue “backpropagation” style corrections. 11. Adaptive stimulus generation:

[0217] In some embodiments, adaptive stimulation generation ensures the system dynamically adjusts microfluidic and stimulation parameters to optimize processing for specific tasks such as speech recognition, visual processing, or other systems. In some instances, this utilizes living neural tissue's ability to integrate multiple input types simultaneously, an improvement over conventional digital computers that operate strictly on binary data.

12. Learning

[0218] In some embodiments, the biocomputation platform learning can involve training neural cells/tissues to produce desired responses through repeated stimulation patterns. In some embodiments, correct behaviors are reinforced using additional neural stimuli or by adjusting the biochemical environment via mechanisms such as microfluidic delivery systems. In some instances, this method allows the neural tissues to adapt and optimize their responses over time, enhancing computational efficiency. In some embodiments, nonreinforced learning techniques are performed. Non-reinforced learning techniques may include unsupervised learning, where neural tissues identify patterns without explicit training signals. Non-reinforced learning techniques may include supervised learning, where tissues are trained using labeled datasets.

[0219] In some embodiments, living neurons can be trained using specialized techniques rooted in computational neuroscience. a. Spike timing dependent plasticity (STDP)

[0220] In some embodiments, spike timing dependent plasticity adjusts the strength of real or simulated synaptic connections based on the precise timing of neuronal spikes. In some embodiments, if a presynaptic neuron's spike precedes a postsynaptic neuron's spike within a narrow time window, synaptic strengthening occurs; if the order is reversed, weakening ensues. In some embodiments, implementing STDP in our biocomputer may enable neurons to adaptively refine their responses based on temporal patterns of activity. b. Hebbian learning

[0221] In some embodiments, Hebbian learning is used to train neural tissues. Hebbiann learning may be summarized as "cells that fire together, wire together." This principle posits that simultaneous activation of neurons leads to increased synaptic strength between them. In some embodiments, of the system, neurons can naturally develop associations and recognize patterns without explicit programming through Hebbian learning. c. Homeostatic plasticity

[0222] In some embodiments, homeostatic plasticity maintains overall neural activity within optimal ranges by adjusting synaptic strengths up or down in response to prolonged changes in network activity. In some instances, incorporating homeostatic plasticity ensures stable operation of the biocomputer, preventing runaway excitation or depression that could impair functionality. d. Training vs inference

[0223] In some embodiments, training involves adjusting the neural tissue's synaptic weights through repeated stimuli to achieve desired responses, while inference pertains to the tissue's application of learned patterns to new inputs. In our biocomputer, training and inference can occur in neural cells/tissues, and/or silicon processors allowing each system to adapt and learn from stimuli either together or independently. In some instances, this enables efficient computation by leveraging the strengths of both biological and silicon components. e. Functions

[0224] In some embodiments, training neural tissues to execute basic scripts, such as print "Hello, World!", involves associating specific input patterns with desired output responses. In some embodiments, through leveraging methods such as reinforcement learning, neural tissues can learn to recognize input patterns corresponding to simple commands and produce appropriate outputs. In some instances, this allows for coding the tissue using computer languages such as python or Verilog.

[0225] In some embodiments, implementing multi-layered functions involves sending complex queries to neural tissues and processing their responses. In some embodiments, this process, akin to hierarchical or modular programming, allows the biocomputer to handle intricate tasks by decomposing them into simpler sub-tasks managed by different neural tissue modules. In some instance, the biocomputing system programs living tissues to execute multi-layered functions, enabling dynamic adaptability and parallel processing capabilities inherent to biological systems. 13. Use of biocomputing

[0226] In some embodiments, the biocomputing system described herein can automatically convert data between numerical formats and neural spike trains or waveforms, facilitating seamless communication between silicon processors and neural tissues. This bidirectional conversion enables efficient data encoding and decoding, enhancing the system's versatility in handling various data types.

[0227] In some embodiments, inputs are converted into neural spike trains and other relevant signals using encoding schemes compatible with neural processing. This process may involve translating digital data into temporal patterns, shapes, frequencies, etc. of electrical stimuli that neural tissues can interpret. In some embodiments, utilizing neuromorphic chips can facilitate this conversion by emulating neural architectures, providing a bridge between silicon-based systems and biological tissues.

[0228] In some embodiments, outputs from neural tissues are decoded by analyzing signals such as neural spike trains and waveforms. In some embodiments, techniques such as neural decoding algorithms, or machine learning models can interpret these patterns, translating them back into digital data. In some instances, this approach allows for real-time monitoring and analysis of neural tissue responses, facilitating effective communication between biological and silicon components.

[0229] In some embodiments, automatic conversion of alpha-numeric values into neuron readable formats involves encoding characters into neural process features. In some embodiments, each character can be represented by a unique feature such as a neural spike, or pattern of neural spikes, shape, location, frequency, of neural spike, etc. enabling neural tissues to process textual/mathematical data. In some embodiments, this method allows the biocomputer to handle language-based information, expanding its application scope.

[0230] Beyond alpha-numeric encoding, in some embodiments, the system can directly be programmed and process data representations, such as Binary, Hexidecimal, or Unicode. In some embodiments, by efficiently encoding packaged data, including compressed files, into neural-compatible formats, the biocomputer can manage diverse data types.

[0231] Data compression reduces data size by removing redundancy or using encoding techniques. Decompression reverses this process to restore the original data. In some embodiments, compressed data packages are sent into the neural cells/tissues. In some embodiments the neural cells/tissues are trained to decompress files. In some embodiments, data packages are compressed in formats like ZIP, RAR, or custom, compressed, neuromorphic data packages in order to minimize data sent through the brain-machine interface, while maximizing the data within the neural tissues. In some embodiments the tissue can compress files and return compressed files back through the brain machine interface.

[0232] In some embodiments, users can define specific encoding schemes for data input into the biocomputer, tailoring the system's processing capabilities to particular applications. [0233] In some embodiments, the biocomputer can autonomously refine its data encoding and processing algorithms based on performance feedback. In some embodiments, by analyzing the efficacy of different encoding schemes, the ability to extract features, and model performance, the system can optimize its operations, signal shapes and types, leading to improved computational efficiency over time. In some embodiments, this self-improving capability leverages the adaptive nature of neural tissues, offering a dynamic approach to computation. This is reminiscent of neural architecture search algorithms in silicon based ANN. In some embodiments, self-improvement in the biocomputing system expands beyond the scope of neural architecture search, in that all components of the biocomputing system are available to be optimized across neural tissues and hardware systems.

[0234] In some embodiments, implementing hierarchical processing layers allows the biocomputer to manage information at varying levels of abstraction, optimizing resource allocation and enhancing processing efficiency.

[0235] In some embodiments, matrix mathematics can form the foundation of many computational models, including neural networks. In some embodiments, matrix math enables efficient representation and manipulation of data structures, essential for operations like transformations, optimizations, and modeling in various applications. In some embodiments, our biocomputer can perform matrix operations within neural tissues and/or non-neural systems, leveraging their inherent parallel processing capabilities. In some embodiments, this approach enables efficient handling of complex mathematical computations, essential for advanced data processing tasks.

[0236] In some embodiments, either through direct training of the tissues to respond to stimuli or by utilizing matrix math capabilities, our biocomputer can support the development of large language models. In some embodiments, these models, implemented within neural tissues, can process and generate human-like text, offering applications in natural language processing and understanding. In some embodiments, the integration of LLMs within a biocomputational framework represents allows for language modeling. [0237] In some embodiments, we are building AI/ML models that can exist across silicon and biological hardware. In some embodiments, reasoning models are models in which certain aspects of the model are given extra time for data processing/thought. In some embodiments, by distributing computational loads between silicon processors and neural tissues, the biocomputer can efficiently manage intricate problem-solving activities. For example, in some embodiments, the most computationally expensive reasoning portions of a model can be offloaded onto the neural tissues, the neural tissue can “perform computations” and information can then be sent back to other modules to inform the larger model. In some embodiments, this hybrid approach leverages the strengths of all biocomputer components, enhancing the system's overall reasoning capabilities.

[0238] In some embodiments, biocomputation models are not limited to reasoning models, and reasoning models are not limited to LLM or other speech applications. Some other useful applications of the biocomputing systems described herein includes the following:

[0239] In some embodiments, a mathematical and theoretical biocomputation system enables automated theorem proving by starting with a small set of axioms or conjectures, leading to deep computational exploration and producing a concise yet valuable proof. In some embodiments, it facilitates prime factorization and cryptanalysis by factoring large numbers, which is critical for cryptographic applications, or by solving complex mathematical problems such as the Riemann Hypothesis, where the proof itself serves as the key output.

[0240] In some embodiments, capabilities include neural architecture search, where the system autonomously evolves optimal Al architectures. In some embodiments, the primary output comprises of trained model weights or optimized hyperparameters, eliminating the need for external design intervention. In some embodiments, the system supports symbolic regression and algorithm discovery, generating compact mathematical formulas or algorithms that describe and predict complex systems efficiently.

[0241] In some embodiments, biocomputer functionality allows for fundamental physics simulations, including quantum gravity explorations, dark matter interactions, and exotic material property analyses. In some embodiments, the output of such simulations may be a singular, groundbreaking discovery, such as a new fundamental equation governing physical phenomena. In some embodiment in the realm of protein folding and drug discovery, the system may simulate molecular interactions to produce a ranked list of promising drug candidates, streamlining the identification of potential therapeutics. [0242] In some embodiments, the biocomputer supports game theory, economic modeling, air, sea, or car routing by running large-scale economic simulations to derive a refined, optimal policy recommendation.

[0243] In some modalities capabilities encompass data compression research, where the system identifies new encoding schemes, with the primary output being an optimized compression algorithm. In some modalities, the system performs scientific knowledge synthesis, condensing vast amounts of simulated or inferred data into concise, human- readable insights, allowing for efficient knowledge extraction and application across various scientific disciplines.

[0244] In some modalities by simulating and/or interfacing with biological sensory systems, the biocomputer can interpret complex sensory inputs, such as visual or auditory data.

[0245] In some modalities, by interacting with and/or modeling motor neuron circuits, the system can contribute to advancements in prosthetics and robotics, enabling more natural and adaptive movements.

[0246] In some modalities simulating and/or interfacing with aspects of human cognition, such as memory and decision-making, can lead to improved human-computer interaction and personalized Al systems.

[0247] In some modalities, low powered drones can be made in which neural cells/tissues complete remote, dangerous, or otherwise useful tasks in autonomous or semi -autonomous manners.

14. Use and/or Generation of Annotated Neuroscience Datasets

[0248] In one embodiment, the biocomputation system described herein integrates annotated neuroscience datasets that are obtained when a subject thinks, speaks, or reads with a brain recording device. In such systems, annotated neural recordings are processed to extract neural data. This data can be converted into relevant patterns such as neural spike trains and then introduced into engineered tissues through brain-machine interfaces including but not limited to MEAs. In some embodiments, similar datasets may be emergent properties derived from biocomputation.

[0249] In some iterations, this annotated neural data is further tokenized into discrete symbols representing distinct neural activity patterns or concepts. For example, a neural spike train may either represent the letters “hello,” the neural patterns that represent “hello” derived from an annotated neuroscience dataset, or the concept of greeting someone. [0250] In other embodiments, the stimulation inputs to the biocomputing tissue consist of a combination of tokenized waveforms and simple spike signals that encode data such as numerical values. This dual encoding approach allows the system to represent both complex waveform structures and quantitative information concurrently. By correlating the annotated conceptual data with specific patterns of neural spikes, the system is capable of learning and processing language-related tasks in a biologically inspired manner.

[0251] Within this framework, neural data representations function as vectors that can be tokenized and subsequently combined using vector math. For example, vector addition may be applied where the embedding vector for "king" adjusted by an element representing "woman" results in a vector corresponding to "queen." In another example the embedded vector for the word “hello” may result in the vector for “world.”

[0252] The system described herein thereby provides an adaptable and scalable method for interfacing high-fidelity neuroscience data with biocomputational devices. It facilitates efficient training and inference in biocomputation AI/ML tissue networks, while preserving the temporal and quantitative aspects of the original neural information. An additional layer of conversion processes the neural data by translating waveforms into ASCII or binary representations, further bridging the gap between biological signals and digital encoding. [0253] This capability allows the system to manipulate, combine or generate concepts dynamically, with individual nodes capable of operating across both the biological tissue and the digital computing layers. The integration of these layered conversion techniques ensures a comprehensive encoding and decoding process, preserving the semantic integrity of the original neural information.

C. Machine learning and artificial intelligence

[0254] Also described herein are method comprising using electrical, chemical, and/or physical stimulation to engineered tissue interior in biocomputation. The methods may comprise using living neurons as perceptrons in an artificial neural network. In some embodiments, the method comprises the use of artificial neurons and biological neurons as enmeshed nodes in a neural network. In some embodiments, the method comprises the integration of artificial neural network and biological neural network with forward data flow. In some embodiments, the method comprises the integration of artificial neural network and biological neural network with a combination of forward data flow and backpropagation. In some embodiments, the method comprises the use of biological neuron firings from three- dimensional multi el ectrode embedded in engineered tissue to populate a data table which is then interpreted by artificial intelligence. In some embodiments, the method comprises the use of biological neuron firings from three-dimensional multi el ectrode embedded in engineered tissue directly interpreted by artificial intelligence in real-time.

[0255] As used in this specification and the appended claims, the terms “artificial intelligence,” “artificial intelligence techniques,” “artificial intelligence operation,” and “artificial intelligence algorithm” generally refer to any system or computational procedure that may take one or more actions that simulate human intelligence processes for enhancing or maximizing a chance of achieving a goal. The term “artificial intelligence” may include “generative modeling,” “machine learning” (ML), or “reinforcement learning” (RL) or “biocomputation.”

[0256] As used in this specification and the appended claims, the terms “machine learning,” “machine learning techniques,” “machine learning operation,” and “machine learning model” generally refer to any system or analytical or statistical procedure that may progressively improve computer performance of a task. In some cases, ML may generally involve identifying and recognizing patterns in existing data in order to facilitate making predictions for subsequent data. ML may include a ML model (which may include, for example, a ML algorithm). Machine learning, whether analytical or statistical in nature, may provide deductive or abductive inference based on real or simulated data. The ML model may be a trained model. ML techniques may comprise one or more supervised, semi-supervised, self-supervised, or unsupervised ML techniques. For example, an ML model may be a trained model that is trained through supervised learning (e.g., various parameters are determined as weights or scaling factors). ML may comprise one or more of regression analysis, regularization, classification, dimensionality reduction, ensemble learning, meta learning, association rule learning, cluster analysis, anomaly detection, deep learning, or ultra-deep learning. ML may comprise, but is not limited to: k-means, k-means clustering, k-nearest neighbors, learning vector quantization, linear regression, non-linear regression, least squares regression, partial least squares regression, logistic regression, stepwise regression, multivariate adaptive regression splines, ridge regression, principal component regression, least absolute shrinkage and selection operation (LASSO), least angle regression, canonical correlation analysis, factor analysis, independent component analysis, linear discriminant analysis, multidimensional scaling, non-negative matrix factorization, principal components analysis, principal coordinates analysis, projection pursuit, Sammon mapping, t-distributed stochastic neighbor embedding, AdaBoosting, boosting, gradient boosting, bootstrap aggregation, ensemble averaging, decision trees, conditional decision trees, boosted decision trees, gradient boosted decision trees, random forests, stacked generalization, Bayesian networks, Bayesian belief networks, naive Bayes, Gaussian naive Bayes, multinomial naive Bayes, hidden Markov models, hierarchical hidden Markov models, support vector machines, encoders, decoders, auto-encoders, stacked auto-encoders, perceptrons, multi-layer perceptrons, artificial neural networks, feedforward neural networks, convolutional neural networks, recurrent neural networks, long short-term memory, deep belief networks, deep Boltzmann machines, deep convolutional neural networks, deep recurrent neural networks, or generative adversarial networks.

[0257] Training the ML model may include, in some cases, selecting one or more untrained data models to train using a training data set. The selected untrained data models may include any type of untrained ML models for supervised, semi-supervised, selfsupervised, or unsupervised machine learning. The selected untrained data models may be specified based upon input (e.g., user input) specifying relevant parameters to use as predicted variables or other variables to use as potential explanatory variables. For example, the selected untrained data models may be specified to generate an output (e.g., a prediction) based upon the input. Conditions for training the ML model from the selected untrained data models may likewise be selected, such as limits on the ML model complexity or limits on the ML model refinement past a certain point. The ML model may be trained (e.g., via a computer system such as a server) using the training data set. In some cases, a first subset of the training data set may be selected to train the ML model. The selected untrained data models may then be trained on the first subset of training data set using appropriate ML techniques, based upon the type of ML model selected and any conditions specified for training the ML model. In some cases, due to the processing power requirements of training the ML model, the selected untrained data models may be trained using additional computing resources (e.g., cloud computing resources). Such training may continue, in some cases, until at least one aspect of the ML model is validated and meets selection criteria to be used as a predictive model.

[0258] In some cases, one or more aspects of the ML model may be validated using a second subset of the training data set (e.g., distinct from the first subset of the training data set) to determine accuracy and robustness of the ML model. Such validation may include applying the ML model to the second subset of the training data set to make predictions derived from the second subset of the training data. The ML model may then be evaluated to determine whether performance is sufficient based upon the derived predictions. The sufficiency criteria applied to the ML model may vary depending upon the size of the training data set available for training, the performance of previous iterations of trained models, or user-specified performance requirements. If the ML model does not achieve sufficient performance, additional training may be performed. Additional training may include refinement of the ML model or retraining on a different first subset of the training dataset, after which the new ML model may again be validated and assessed. When the ML model has achieved sufficient performance, in some cases, the ML may be stored for present or future use. The ML model may be stored as sets of parameter values or weights for analysis of further input (e.g., further relevant parameters to use as further predicted variables, further explanatory variables, further user interaction data, etc.), which may also include analysis logic or indications of model validity in some instances. In some cases, a plurality of ML models may be stored for generating predictions under different sets of input data conditions. In some embodiments, the ML model may be stored in a database (e.g., associated with a server).

[0259] Examples of machine learning include, without limitations, random walk and biased random walk, decision tree and random forest, computer vision, support vector machine, LSTM, vision transformer, and masked autoencoder.

1. Decision Tree and Random Forest

[0260] As described above, the machine learning model may implement a decision tree. A decision tree may be a supervised ML algorithm that can be applied to both regression and classification problems. Decision trees may mimic the decision-making process of a human brain. For example, a decision tree may grow from a root (base condition), and when it meets a condition (internal node/feature), it may split into multiple branches. The end of the branch that does not split anymore may be an outcome (leaf). A decision tree can be generated using a training data set according to the following operations: (1) Starting from a root node (the entire dataset), the algorithm may split the dataset in two branches using a decision rule or branching criterion; (2) each of these two branches may generate a new child node; (3) for each new child node, the branching process may be repeated until the dataset cannot be split any further; (4) each branching criterion may be chosen to maximize information gain (e.g., a quantification of how much a branching criterion reduces a quantification of how mixed the labels are in the children nodes). The labels may be the data or the classification that is predicted by the decision tree.

[0261] A random forest regression is an extension of the decision tree model that tends to yield more robust predictions by stretching the use of the training data partition. Whereas a decision tree may make a single pass through the data, a random forest regression may bootstrap 50% of the data (e.g., with replacement) and build many trees. Rather than using all explanatory variables as candidates for splitting, a random subset of candidate variables may be used for splitting, which may enable trees that have completely different data and different variables (hence the term random). The predictions from the trees, collectively referred to as the “forest,” may be then averaged together to produce the final prediction. Many trees (e.g., one hundred trees) may be included in a random forest model, with a number (e.g., 3, 6, 10, etc.) of terms sampled per split, a minimum of number (e.g., 1, 2, 4, 10, etc.) of splits per tree, and a minimum split size (e.g., 16, 32, 64, 128, 256, etc.). Random forests may be trained in a similar way as decision trees. Specifically, training a random forest may include the following operations: (1) select randomly k features from the total number of features; (2) create a decision tree from these k features using the same operations as for generating a decision tree; and (3) repeat the previous two operations until a target number of trees is created.

[0262] FIG. 21 illustrates a random forest 800. The random forest 800 (which may also be referred to as random forest model) is an ensemble of decision trees 805, 810, and 815 with randomly selected features in each of the decision trees 805, 810, and 815 so that it can provide more stable and accurate outcomes. Outcomes may be determined by majority voting in the case of a classification problem. In the example of FIG. 8, the random forest 800, which has been trained previously by a training method, is used to decide between classifications A, B and C. For example, the random forest 800, with only the three decision trees shown in FIG. 8, would return the classification A by majority voting.

2. Computer Vision

[0263] The systems, the methods, the biocomputation systems, and the techniques disclosed herein may implement one or more computer vision techniques. Computer vision is a field of artificial intelligence that uses computers to interpret and understand the visual world at least in part by processing one or more digital images from cameras and videos. In some instances, computer vision may use deep learning models (e.g., convolutional neural networks). Bounding boxes may be used in object detection techniques within computer vision. Bounding boxes may be annotation markers drawn around objects in an image. Bounding boxes, are often, although not always, may be rectangularly shaped. Bounding boxes may be applied by humans to training data sets. However, bounding boxes may also be applied to images by a trained machine learning that is trained to detect one or more different objects (e.g., humans, hands, faces, cars, etc.). In addition or in alternative to bounding boxes detection and tracking techniques may use any object detection annotation techniques, such as semantic segmentation, instance segmentation, polygon annotation, non-polygon annotation, landmarking, 3D cuboids, etc.

3. Support Vector Machine

[0264] As also described above, the machine learning model may implement support vector machine learning techniques. In machine learning, support vector machines (SVMs) may be supervised learning models with associated learning algorithms that analyze data for classification and regression analysis. SVMs may be a robust prediction method, being based on statistical learning. SVMs may be well-suited for domains characterized by the existence of large amounts of data, noisy patterns, or the absence of general theories.

[0265] In general terms, SVMs may map input vectors into high dimensional feature space through non-linear mapping function, chosen a priori. In this high dimensional feature space, an optimal separating hyperplane may be constructed. The optimal hyperplane may then be used to determine things such as class separations, regression fit, or accuracy in density estimation. More formally, a SVM constructs a hyperplane or set of hyperplanes in a high or infinite-dimensional space, which can be used for classification, regression, or other tasks like outlier detection.

[0266] Support vectors may be defined as the data points that lie closest to the decision surface (or hyperplane). Support vectors may therefore be the data points that are most difficult to classify and may have direct bearing on the optimum location of the decision surface. Given a set of training examples, each marked as belonging to one of two categories, an SVM training algorithm may build a model that assigns new examples to one category or the other, making it a non-probabilistic binary linear classifier (although methods such as Platt scaling exist to use SVM in a probabilistic classification setting). SVM may map training examples to points in space so as to maximize the width of the gap between the two categories. New examples may then be mapped into that same space and predicted to belong to a category based on which side of the gap they fall. In addition to performing linear classification, SVMs can efficiently perform a non-linear classification using what is called the kernel trick, implicitly mapping their inputs into high-dimensional feature spaces.

[0267] Within a support vector machine, the dimensionally of the feature space may be large. For example, a fourth-degree polynomial mapping function may cause a 200- dimensional input space to be mapped into a 1.6 billionth dimensional feature space. The kernel trick and the Vapnik-Chervonenkis dimension may allow the SVM to thwart the “curse of dimensionality” limiting other methods and effectively derive generalizable answers from this very high dimensional feature space. Accordingly, SVMs may assist in discovering knowledge from vast amounts of input data.

[0268] Patent applications directed to support vector machines include, U.S. patent application Ser. Nos. 09/303,386; 09/303,387; 09/303,389; 09/305,345; all filed May 1, 1999; and U.S. patent application Ser. No. 09/568,301, filed May 9, 2000; and U.S. patent application Ser. No. 09/578,011, filed May 24, 2000 and also claims the benefit of U.S. Provisional Patent Application No. 60/161,806, filed Oct. 27, 1999; of U.S. Provisional Patent Application No. 60/168,703, filed Dec. 2, 1999; of U.S. Provisional Patent Application No. 60/184,596, filed Feb. 24, 2000; and of U.S. Provisional Patent Application Ser. No. 60/191,219, filed Mar. 22, 2000; all of which are herein incorporated in their entireties.

4. LSTM

[0269] Long short-term memory (LSTM) may be an artificial neural network used in the fields of artificial intelligence and deep learning. Unlike standard feedforward neural networks, LSTM may use feedback connections. The LSTM architecture may provide a short-term memory for a recurrent neural network (RNN). Such RNN can process not only single data points (such as images), but also entire sequences of data (such as speech or video). This characteristic may mean that LSTM networks are well-suited for processing and predicting data. The name of LSTM may refer to the analogy that a standard RNN has both “long-term memory” and “short-term memory.” The connection weights and biases in the RNN may change once per episode of training, analogous to how physiological changes in synaptic strengths store long-term memories; the activation patterns in the network may change once per time-step, analogous to how the moment-to-moment change in electric firing patterns in the brain store short-term memories. The LSTM architecture may provide a shortterm memory for an RNN that can last many (e.g., thousands) timesteps.

[0270] In some cases, a LSTM unit may comprise a cell, an input gate, an output gate, and a forget gate. The cell may remember values over arbitrary time intervals and the input gate, the output gate, and the forget gate may regulate the flow of information into and out of the cell. Forget gates may be used to decide what information to discard from a previous state by assigning a previous state, compared to a current input, a value between 0 and 1 (e.g., a (rounded) value of 1 may mean to keep the information, and a value of 0 means to discard it). The input gate may decide which pieces of new information to store in the current state, using the same system as the forget gates. The output gate may control which pieces of information in the current state to output (e.g., by assigning a value from 0 to 1 to the information, considering the previous and current states). Selectively outputting relevant information from the current state may allow the LSTM network to maintain useful, long-term dependencies to make predictions, both in current and future time-steps. LSTM networks may be well-suited to classifying, processing and making predictions based on time series data, since there can be lags of unknown duration between important events in a time series. LSTMs may resolve the vanishing gradient problem that can be encountered when training traditional RNNs. Relative insensitivity to gap length may be an advantage of LSTM over RNNs, hidden Markov models and other sequence learning methods in numerous applications.

[0271] In some cases, LSTMs may be used with one or more various types of neural networks (e.g., convolutional neural networks (CNNs), deep neural network (DNNs), recurrent neural networks (RNNs), etc.). In some cases, CNNs, LSTM, and DNNs are complementary in their modeling capabilities and may be combined a unified architecture. For example, in such unified architecture, CNNs may be well-suited at reducing frequency variations, LSTMs may be well-suited at temporal modeling, and DNNs may be well-suited for mapping features to a more separable space. For example, input features to a ML model using LSTM techniques in the a unified architecture may include segment features for each of a plurality of segments. To process the input features for each of the plurality of segments, the segment features for the segment may be processed using one or more CNN layers to generate first features for the segment; the first features may be processed using one or more LSTM layers to generate second features for the segment; and the second features may be processed using one or more fully connected neural network layers to generate third features for the segments, where the third features may be used for classification operations. In some examples, to process the first features using the one or more LSTM layers to generate the second features, the first features may be processed using a linear layer to generate reduced features having a reduced dimension from a dimension of the first features; and the reduced features may be processed using the one or more LSTM layers to generate the second features. Short-term features having a first number of contextual frames may be generated based on the input features, where features generated using the one or more CNN layers may include long-term features having a second number of contextual frames that are more than the first number of contextual frames of the short-term features. In some cases, the one or more CNN layers, the one or more LSTM layers, and the one or more fully connected neural network layers may have been jointly trained to determine trained values of parameters of the one or more CNN layers, the one or more LSTM layers, and the one or more fully connected neural network layers. In some cases, the input features may include log-mel features having multiple dimensions. The input features may include one or more contextual frames indicating a temporal context of a signal (e.g., input data). Advantageously, implementations for such unified architecture may leverage complementary advantages associated with each of a CNN, LSTM, and DNN. For example, convolutional layers may reduce spectral variation in input, which may help the modeling of LSTM layers. Having DNN layers after LSTM layers may help reduce variation in the hidden states of the LSTM layers. Training the unified architecture jointly may provide a better overall performance. Training in the unified architecture may also remove the need to have separate CNN, LSTM and DNN architectures, which may be expensive (e.g., in computational resource, in network traffic, in financial resources, in energy consumption, etc.). By adding multi-scale information into the unified architecture, information may be captured at different time scales.

5. Vision Transformer

[0272] A vision transformer (ViT) is a transformer-like model that handles vision processing tasks. While CNNs use convolution, a “local” operation bounded to a small neighborhood of an image, ViTs use self-attention, a “global” operation, since the ViT draws information from the whole image. This allows the ViT to capture distant semantic relevances in an image effectively. Advantageously, ViTs may be well-suited catching longterm dependencies. In some cases, ViTs may be a competitive alternative to convolutional neural networks as ViTs may outperform the current state-of-the-art CNNs by almost four times in terms of computational efficiency and accuracy. ViTs may be well-suited to object detection, image segmentation, image classification, and action recognition. Moreover, ViTs may be applied in generative modeling and multi-model tasks, including visual grounding, visual-question answering, and visual reasoning. In some cases, ViTs may represent images as sequences, and class labels for the image are predicted, which allows models to learn image structure independently. Input images may be treated as a sequence of patches where every patch is flattened into a single vector by concatenating the channels of all pixels in a patch and then linearly projecting it to the desired input dimension. For example, a ViT architecture may include the following operations: (A) split an image into patches; (B) flatten the patches; (C) generate lower-dimensional linear embeddings from the flattened patches;

(D) add positional embeddings; (E) provide the sequence as an input to a standard transformer encoder; (F) pretrain a model with image labels (e.g., fully supervised on a huge dataset); and (G) finetune on the downstream dataset for image classification. In some cases, there may be multiple blocks in a ViT encoder, with each block comprising three major processing elements: (1) Layer Norm; (2) Multi-head Attention Network; and (3) MultiLayer Perceptrons. The Layer Norm may keep the training process on track and enable the model adapt to the variations among the training images. The Multi-head Attention Network may be a network responsible for generating attention maps from the given embedded visual tokens. These attention maps may help the network focus on the most critical regions in the image, such as object(s). The Multi-Layer Perceptrons may be a two-layer classification network with a Gaussian Error Linear Unit at the end. The final Multi-Layer Perceptrons block may be used as an output of the transformer. An application of SoftMax on this output can provide classification labels (e.g., if the application is image classification).

6. Masked Autoencoder

[0273] Masked autoencoders (MAE) are scalable self-supervised learners for computer vision. The MAE leverages the success of autoencoders for various imaging and natural language processing tasks. Some computer vision models may be trained using supervised learning, such as using humans to look at images and created labels for the images, so that the model could learn the patterns of those labels (e.g., a human annotator would assign a class label to an image or draw bounding boxes around objects in the image). In contrast, selfsupervised learning may not use any human-created labels. One technique for self-supervised image processing training using an MAE is for before an image is input into an encoder transformer, a certain set of masks are applied to the image. Due to the masks, pixels are removed from the image and therefore the model is provided an incomplete image. At a high level, the model’s task is to now learn what the full, original image looked like before the mask was applied.

[0274] In other words, MAE may include masking random patches of an input image and reconstructing the missing pixels. The MAE may be based on two core designs. First, an asymmetric encoder-decoder architecture, with an encoder that operates on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, masking a high proportion of the input image, e.g., 75%, may yield a nontrivial and meaningful self- supervisory task. Coupling these two core designs enables training large models efficiently and effectively, thereby accelerating training (e.g., by 3* or more) and improving accuracy. MAE techniques may be scalable, enabling learning of high-capacity models that generalize well, e.g., a vanilla ViT-Huge model. As mentioned, the MAE may be effective in pretraining ViTs for natural image analysis. In some cases, the MAE uses the characteristic of redundancy of image information to observe partial images to reconstruct original images as a proxy task, and the encoder of the MAE may have the capability of deducing the content of the masked image area by aggregating context information. This contextual aggregation capability may be important in the field of image processing and analysis.

III. DEFINITIONS

[0275] Unless defined otherwise, all terms of art, notations and other technical and scientific terms or terminology used herein are intended to have the same meaning as is commonly understood by one of ordinary skill in the art to which the claimed subject matter pertains. In some cases, terms with commonly understood meanings are defined herein for clarity and/or for ready reference, and the inclusion of such definitions herein should not necessarily be construed to represent a substantial difference over what is generally understood in the art.

[0276] Throughout this application, various embodiments may be presented in a range format. It should be understood that the description in range format is merely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the disclosure. Accordingly, the description of a range should be considered to have specifically disclosed all the possible subranges as well as individual numerical values within that range. For example, description of a range such as from 1 to 6 should be considered to have specifically disclosed subranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6 etc., as well as individual numbers within that range, for example, 1, 2, 3, 4, 5, and 6. This applies regardless of the breadth of the range.

[0277] As used in the specification and claims, the singular forms “a”, “an” and “the” include plural references unless the context clearly dictates otherwise. For example, the term “a sample” includes a plurality of samples, including mixtures thereof.

[0278] The terms “determining,” “measuring,” “evaluating,” “assessing,” “assaying,” and “analyzing” are often used interchangeably herein to refer to forms of measurement. The terms include determining if an element is present or not (for example, detection). These terms can include quantitative, qualitative or quantitative and qualitative determinations. Assessing can be relative or absolute. “Detecting the presence of’ can include determining the amount of something present in addition to determining whether it is present or absent depending on the context. [0279] The terms “subject,” “individual,” or “patient” are often used interchangeably herein. A “subject” can be a biological entity containing expressed genetic materials. The biological entity can be a plant, animal, or microorganism, including, for example, bacteria, viruses, fungi, and protozoa. The subject can be tissues, cells and their progeny of a biological entity obtained in vivo or cultured in vitro. The subject can be a mammal. The mammal can be a human. The subject may be diagnosed or suspected of being at high risk for a disease. In some cases, the subject is not necessarily diagnosed or suspected of being at high risk for the disease.

[0280] The term “zzz vivo" is used to describe an event that takes place in a subject’s body.

[0281] The term “ex vivo" is used to describe an event that takes place outside of a subject’s body. An ex vivo assay is not performed on a subject. Rather, it is performed upon a sample separate from a subject. An example of an ex vivo assay performed on a sample is an “zzz vitro" assay.

[0282] The term “zzz vitro" is used to describe an event that takes places contained in a container for holding laboratory reagent such that it is separated from the biological source from which the material is obtained. In vitro assays can encompass cell-based assays in which living or dead cells are employed. In vitro assays can also encompass a cell-free assay in which no intact cells are employed.

[0283] As used herein, the term “about” a number refers to that number plus or minus 10% of that number. The term “about” a range refers to that range minus 10% of its lowest value and plus 10% of its greatest value.

[0284] As used herein, the terms “treatment” or “treating” are used in reference to a pharmaceutical or other intervention regimen for obtaining beneficial or desired results in the recipient. Beneficial or desired results include but are not limited to a therapeutic benefit and/or a prophylactic benefit. A therapeutic benefit may refer to eradication or amelioration of symptoms or of an underlying disorder being treated. Also, a therapeutic benefit can be achieved with the eradication or amelioration of one or more of the physiological symptoms associated with the underlying disorder such that an improvement is observed in the subject, notwithstanding that the subject may still be afflicted with the underlying disorder. A prophylactic effect includes delaying, preventing, or eliminating the appearance of a disease or condition, delaying or eliminating the onset of symptoms of a disease or condition, slowing, halting, or reversing the progression of a disease or condition, or any combination thereof. For prophylactic benefit, a subject at risk of developing a particular disease, or to a subject reporting one or more of the physiological symptoms of a disease may undergo treatment, even though a diagnosis of this disease may not have been made.

[0285] The term “production bioreactor” or “bioreactor,” as used herein, generally refers to a bioreactor device suitable for scaling production of cells and/or products produced by cells. A production bioreactor may include one or more channels or other openings for inputting cells, for providing liquid media, gas composition and other cell environment factors and one or more channels for harvesting cells and/or products produced by cells.

[0286] As used herein, the term “bioreactor computing system,” “biocomputer” or “biocomputing system” may refer to a system that uses biologically derived materials to perform computational functions. In some instances, the biocomputing system described herein comprises a bioreactors described herein. In some instances, the biocomputer comprises in vitro cultured cell brain machine interfaces that may be conducted using surface level stimulation and recording of neural spikes using multi-electrode arrays. In these multielectrode arrays, each electrode can stimulate and record electric cellular messaging at physiologically relevant ranges. A bioreactor computing system may also refer to a system inspired by a biologically inspired system without active biological components.

[0287] The term “field programmable gate array” (FPGA) is used to describe a type of “integrated circuit” (IC) that can be reconfigured after it's been manufactured. Components of the biocomputing system described herein may comprise FGPA’s, because FGPA’s represent a reconfigurable chip that is useful during development. However, in each case, the use of FGPA is not meant as an exclusionary term. In certain configurations, each chip could be excluded, or replaced with an “Application Specific Integrated Circuit” (ASIC), other chips, or biological components. Specific architectures are also not exclusionary. Von Neumann and neuromorphic architectures are each considered in isolation and conjunction for our various biocomputer modules.

[0288] "Spike train" or "neural spike train" is representative of data transmitted to cells or tissues. However, the term may be used to indicate that data is captured or transmitted, and is not intended to limit concepts to the transmission of electrical spikes into or out of tissues. In some embodiments, spike trains may consist of a single spike of electricity. In other embodiments, a spike train may consist of multiple spikes of electricity. In some embodiments, spike trains consist of various waveforms. In some embodiments, spike trains may exist as data within a silicon environment, or consist of other biological or physical data. For example, spike trains may consist of spikes of media provided to the reactor or biological signaling patterns transmitted through means other than electricity such as protein interactions.

[0289] The section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described.

IV. EMBODIMENTS

[0290] Also described herein are the following embodiments:

1. A bioreactor system comprising:

(a) a first module comprising a plurality of bioprocess controls; and

(b) a second module comprising a system for electrical stimulation and electrical recording, wherein the first module is isolated from electrical stimulation.

2. The bioreactor system of embodiment 1, wherein each electrical component is grounded.

3. The bioreactor system of embodiment 1 or 2, wherein the electrical component comprises at least the pump, for the addition and removal of liquids, and pressure control

4. The bioreactor system of any one of embodiments 1-3, wherein the second module is on an anti-static mat.

5. The bioreactor system of any one of embodiments 1-4, wherein the first module is physically isolated from the second module.

6. The bioreactor system of embodiment 1-4, wherein the first module comprises a first chamber and the second module comprises a second chamber.

7. The bioreactor system of embodiment 1-5, wherein the first module comprises an incubator and the second module comprises a bioreactor placed within the incubator.

8. The bioreactor system of embodiment 1-6, wherein the second module comprises a cell culture or a tissue culture plate.

9. The bioreactor system of any one of embodiments 1-7, wherein the plurality of bioprocess controls comprises at least one of a pH control, a dissolved oxygen control, a temperature control, a pressure control, and control of gas, liquid, and solid components added and removed from each reactor.

10. The bioreactor system of embodiment 1-8, wherein the plurality of bioprocess controls comprises at least a pH control, a dissolved oxygen control, a temperature control and a pressure control. 11. The bioreactor system of any one of embodiments 1-10, further comprising a system for additive manufacturing, subtractive manufacturing, or a combination thereof.

12. The method of embodiment 9, wherein additive manufacturing comprises bioprinting and injection molding of biomaterials.

13. The bioreactor system of any one of embodiments 9 or 10, wherein the system for additive and subtractive manufacturing comprises a scaffold material and a plurality of cells.

14. The bioreactor system of embodiment 11, wherein the scaffold material is biocompatible.

15. The bioreactor system of embodiment 11 or 12, wherein the plurality of cells comprise a neural progenitor cell, a stem cell, a primary tissue cell, a differentiated neuron, an astrocyte, an oligodendrocyte, a T-cell, a vascular cell, or a combination thereof.

16. The bioreactor system of embodiment 13, wherein the vascular cell comprises at least one of an endothelial cell, an angioblast, and a smooth muscle-like cell.

17. The bioreactor system of any one of embodiments 1-16, wherein the second module comprises a system assembly, growth, and control of tissues by additive or subtractive manufacturing.

18. The bioreactor system of any one of embodiments 1-14, wherein the second module comprises an electrode array.

19. The bioreactor system of embodiment 15, wherein the electrode array is a three- dimensional electrode array.

20. The bioreactor system of any one of embodiments 18 or 19, wherein the electrode array comprises at least 4 electrodes.

21. The bioreactor system of any one of embodiments 18-20, wherein the electrode array is arranged in a grid.

22. The bioreactor system of any one of embodiments 18-21, wherein the electrode array is a surface electrode array.

23. The bioreactor system of any one of embodiments 18-22, comprising both a three- dimensional electrode or microelectrode array for recording and stimulation of engineered tissues.

24. The bioreactor system of any one of embodiments 18-23, comprising both a surface and three-dimensional electrode or microelectrode array for recording and stimulation of engineered tissues.

25. The bioreactor system of any one of embodiments 18-24, wherein each electrode is configured to stimulate and record electronic cellular messaging. 26. The bioreactor system of any one of embodiments 1-25, further comprising at least 1 port for flow of liquid into the bioreactor.

27. The bioreactor system of embodiment 26, wherein the liquid comprises growth factors, nutrients, metabolites, stabilizers, pH indicators and controllers, living and non-living components .

28. The bioreactor system of any one of embodiments 1-27, further comprising a perfusion aid.

29. The bioreactor system of any one of embodiments 1-28, further comprising at least one syringe.

30. The bioreactor system of any one of embodiments 1-29, wherein the system for electrical stimulation comprises at least one electrical component comprising at least one of a pump, a microcontroller, or controller, a probe, and a microelectrode array.

31. A method of using the bioreactor system of any one of embodiments 1-30, wherein additive manufacturing is used to directly biomanufacture tissues into said reactor.

32. The method of embodiment 1-31, wherein the bioreactor system allows stimulation and recording of engineered tissue electrophysiochemistry signals through isolation and reduction of background electrical noise.

33. The method of embodiment 1-30, wherein engineered tissue viability is maintained by way of active perfusion, or passive perfusion in which avascular engineered tissues of diameter greater than 1000 microns, are maintained through the use of porous biomaterials.

34. The method of embodiment 33, wherein engineered tissue viability is maintained by way of active perfusion, or passive perfusion in which avascular engineered tissues of diameter greater than 1000 microns, are maintained through the use of tissue-free spaces, such as avascular channels for media transport, by way of additive or subtractive manufacturing perfusion.

35. A method of manufacturing a three-dimensional tissue, comprising manufacturing the three-dimensional tissue directly into a bioreactor; wherein the tissue is greater than 1000 pm³.

36. The method of embodiment 35, wherein the three-dimensional tissue is integrated with a three-dimensional multi el ectrode array.

37. The method of embodiment 36, wherein the bioreactor comprises a first module comprising bioprocess controls and a second module comprising the three-dimensional electrode array, wherein the first module is isolated from electrical stimulation.

38. The method of embodiment 37, wherein the bioprinting occurs in the second module. 39. The method of any one of embodiments 31-38, wherein the bioprinting comprises a system comprising a carrier fluid and a plurality of cells.

40. The method of any one of embodiments 31-39, wherein the three-dimensional tissue comprises a three-dimensional tissue for use in regenerative medicine.

41. The method of embodiment 40, wherein the three-dimensional tissue comprises neural tissue.

42. The method of embodiment 40, wherein the three-dimensional tissue comprises muscle tissue.

43. The method of embodiment 40, wherein the three-dimensional tissue is for use in a non-clinical trial.

44. The method of any one of embodiments 31-43, wherein the three-dimensional tissue is for use in cellular agriculture.

45. The method of any one of embodiments 31-44, wherein the three-dimensional tissue is for use in biocomputing.

46. A method of analyzing a tissue comprising a three-dimensional electrode array, comprising

(a) additive and subtractive manufacture of the tissue enmeshed with three-dimensional electrode array in a bioreactor, wherein the first module comprises bioprocess controls and the second module comprises the three-dimensional electrode array, wherein the first module is isolated from electrical stimulation;

(b) sending training signals to the tissue in the second module; and

(c) receiving signals from the three-dimensional electrode array.

47. The method of embodiment 46, wherein the engineered tissue and the three- dimensional electrode array comprise a three-dimensional biocomputing system.

48. The method of any one of embodiments 46 or 47, wherein the engineered tissue is greater than 1000 pm³.

49. The method of embodiment 46, wherein each electrical component of the first module of the bioreactor system is grounded.

50. The method of embodiment 49, wherein the electrical component comprises at least the pump, for the addition and removal of liquids, and pressure control.

51. The method of any one of embodiments 31-50, wherein the second module is on an anti-static mat or embedded in a faraday cage.

52. The method of any one of embodiments 31-51, wherein the first module is physically isolated from the second module. 53. The method of embodiment 31-52, wherein the first module comprises a first chamber and the second module comprises a second chamber.

54. The method of embodiment 31-53, wherein the first module comprises an incubator and the second module comprises a bioreactor placed within the incubator.

55. The method of embodiment 31-54, wherein the second module comprises a cell culture or a tissue culture plate.

56. The method of any one of embodiments 31-55, wherein the plurality of bioprocess controls comprises at least one of a pH control, a dissolved oxygen control, a temperature control, a pressure control, and control of gas, liquid, and solid components added and removed from each reactor.

57. The method of embodiment 31-56, wherein the plurality of bioprocess controls comprises at least a pH control, a dissolved oxygen control, a temperature control and a pressure control.

58. The method of any one of embodiments 31-57, further comprising at least 1 port for flow of liquid into the bioreactor.

59. The method of embodiment 58, wherein the liquid comprises growth factors, nutrients, metabolites, stabilizers, pH indicators and controllers, living and non-living components .

60. The method of any one of embodiments 31-59, further comprising a perfusion aid.

61. The method of any one of embodiments 31-60, further comprising at least one syringe.

62. The method of any one of embodiments 31-61, wherein the system for electrical stimulation comprises at least one electrical component comprising at least one of a pump, a microcontroller, or controller, a probe, and a microelectrode array.

63. The method of any one of embodiments 31-62, wherein manufacturing comprises at least one method of additive manufacturing.

64. The method of embodiment 63, wherein the method of additive manufacturing comprises bioprinting or injection molding.

65. The method of embodiment 31-64, wherein manufacturing comprises at least one method of subtractive manufacturing.

66. The method of embodiment 65, wherein subtractive manufacturing comprises the removal of a structure, the removal of a porogen or the removal of a bioink.

67. The method of embodiment 31-66, wherein manufacturing comprises at least one method of additive manufacturing and one method of subtractive manufacturing. 68. The method of embodiment 31-67, comprising the use of an injection mold to develop a three-dimensional engineered tissue.

69. The method of embodiment 68, wherein the injection mold is used for biomaterials with ports to perfuse crosslinking reagent, subtractive manufacturing media, cells, bioprocess media, and other living and non-living components.

70. The method of embodiment 68, wherein the injection mold comprises at least one plate that facilitates the removal of tissue from the injection molds.

71. The method of any one of embodiments 31-70, wherein the tissue is vascularized.

72. The method of any one of embodiments 31-71, wherein the tissue comprises at least one cell type.

73. The method of any one of embodiments 31-71, wherein the tissue comprises at least two cell types.

74. The method of any one of embodiments 31-71, wherein the tissue comprises at least three cell types.

75. The method of any one of embodiments 72-74, wherein at least one cell type comprises a neural progenitor cell, a stem cell, a primary tissue cell, a differentiated neuron, an astrocyte, an oligodendrocyte, a T cell, or a vascular cell.

76. The method of embodiment 77, wherein the vascular cell comprises at least one of an endothelial cell, an angioblast, and a smooth muscle-like cell.

77. The method of any one of embodiments 36-76, wherein the three-dimensional microelectrode array is embedded into engineered tissues.

78. The method of any one of embodiments 36-77, wherein at least one electrode in the three-dimensional electrode array can stimulate and record electronic cellular messaging.

79. The method of any one of embodiments 36-78, wherein a plurality of bioprocess controls regulate a plurality of bioprocess parameters simultaneously with at least one electrode in the three dimensional electrode array stimulating and recording electronic cellular messaging.

80. The method of any one of embodiments 36-79, wherein the three-dimensional microelectrode array comprises read and write capabilities.

81. The method of any one of embodiments 36-80, wherein the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues. 82. The method of any one of embodiments 36-81, wherein the three-dimensional microelectrode array with read and write capabilities is embedded into engineered tissues supported by engineered vascular networks.

83. The method of any one of embodiments 36-82, wherein the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues, supported by vascular networks.

84. The method of any one of embodiments 36-83, wherein the three-dimensional microelectrode array with read and write capabilities is embedded into engineered tissues supported by engineered avascular, active perfusion networks.

85. The method of any one of embodiments 36-84, wherein the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues, supported by avascular passive perfusion networks through the use of negative space in engineered tissues.

86. The method of any one of embodiments 36-85, comprising manufacturing of the engineered tissue comprising a plurality of cells with the three-dimensional electrode array in a bioreactor system, with a three-dimensional electrode array capable of: a) sending relevant electrophysiology signals to the tissue; and b) receiving signals from the engineered tissue cells.

87. The method of embodiment 86, further comprising receiving at least one signal from a plurality of cells in the engineered tissue.

88. The method of embodiment 87, wherein the signal is received by the three- dimensional electrode array.

89. The method of embodiment 87, wherein the signal is received by an external receiver.

90. The method of any one of embodiments 35-89, wherein manufacturing comprises adding a primary vascularized tissue directly into a cavity or plurality of cavities within a secondary tissue to combine the primary vascularized tissue and the secondary tissue.

91. The method of embodiment 90, wherein a matrix or a gel layer is used to combine the primary vascularized tissue and the secondary tissue.

92. The method of embodiment 90, wherein the primary vascularized tissue and the secondary tissue are combined without a matrix or a gel layer.

93. The method of any one of embodiments 35-89, wherein manufacturing comprises adding avascular tissues directly into a cavity or plurality of cavities within a secondary tissue to combine the primary avascularized tissue and the secondary tissue. 94. The method of embodiment 93, wherein a matrix or a gel layer is used to combine the primary avascularized tissue and the secondary tissue.

95. The method of embodiment 93, wherein the primary avascularized tissue and the secondary tissue are combined without a matrix or a gel layer.

96. The method of any one of embodiments 35-95, further comprising scanning a plurality of resection cavities, and that biomanufacturing bespoke tissues in the shape of the scanned plurality of resection cavities.

97. The method of any one of embodiments 36-96, further comprising training of cells within an engineered tissue using microelectrode pulses at physiologically relevant ranges embedded within engineered tissues.

98. The method of any one of embodiments 36-97, further comprising training of cells within an engineered tissue using microelectrode pulses at physiologically relevant ranges embedded within the tissues through the use of the three-dimensional arrangement of microelectrode arrays.

99. The method of any one of embodiments 36-98, wherein the three-dimensional electrode array emits a three-dimensional arrangement of electrical pulses to map networks of engineered tissues.

100. The method of any one of embodiments 36-99, comprising reading data communication between a plurality of bioprocess controllers and a plurality of electric cell firings by multi -el ectrode arrays in a biomanufactured three-dimensional environment.

101. The method of any one of embodiments 36-100, wherein the use of training and test phases with three-dimensional multi -el ectrode arrays and which train and evaluate the quality of a neural network.

102. The method of any one of embodiments 36-101, further comprising controlling a bioprocess with a plurality of neural spikes.

103. The method of any one of embodiments 36-101, further comprising controlling a bioprocess with a neural spike pattern.

104. The method of any one of embodiments 36-103, further comprising storing information via in-silico and in vitro neural networks within a manufactured biological three- dimensional environment exceeding >1000 pm cross-section.

105. The method of any one of embodiments 36-104, further comprising writing, reading, and interpreting cell electrical firings with an artificial neural network within a manufactured biological three-dimensional environment exceeding >1000 pm cross-section.

-n - 106. The method of embodiment 36-105, wherein multiple organoids or tissues are used together in biocomputation and information storage.

107. The method of any one of embodiments 36-106, wherein the pulse train, strength, frequency, pattern, duration, waveform, amplitude, shape, and physical location are used for read, write, and storage of information in a biocomputation environment.

108. The method of any one of embodiments 36-107, comprising correlating gene expression to neuronal activity over time and three-dimensional space to indicate learning patterns and mechanisms.

109. The method of any one of embodiments 36-108, further comprising the use of electrical, chemical, and/or physical stimulation to engineered tissue interior for use in biocomputation.

110. The method of any one of embodiments 36-109, further comprising the use of living neurons as perceptrons in an artificial neural network.

111. The methods of any one of embodiments 36-110, further comprising the use of artificial neurons and biological neurons as enmeshed nodes in a neural network.

112. The method of any one of embodiments 36-111, wherein an artificial neural network and a biological neural network is integrated with forward data flow.

113. The method of any one of embodiments 36-111, wherein an artificial neural network and a biological neural network is integrated with a combination of forward data flow and backpropagation .

114. The method of any one of embodiments 36-113, wherein biological neuron firings from the three-dimensional multi el ectrode embedded in engineered tissue populates a data table which is then interpreted by artificial intelligence.

115. The method of any one of embodiments 36-113, wherein biological neuron firings from 3D multi el ectrode embedded in engineered tissue is directly interpreted by artificial intelligence in real-time.

116. The method of any one of embodiments 59-115, wherein an operating system merges python-like scripting with verilog-like scripting to enable the same level of functionality at two layers of abstraction.

117. The method of any one of embodiments 59-116, wherein an operating system merges Verilog-like and Python-like scripting to dynamically program FPGAs or generate ASIC snippets. 118. The method of any one of embodiments 59-117, wherein the biocomputing system continuously monitors neural activity, including advanced signal processing features, and adapts via a closed-loop feedback mechanism.

119. The method of any one of embodiments 59-118, wherein the biocomputing system comprises a neuromorphic platform that integrates living neural tissue with reconfigurable digital hardware.

120. The method of any one of embodiments 59-119, wherein a unified approach permits dynamic FPGA reconfiguration in real time and allows stable modules to be exported as ASIC -ready design.

121. The method of any one of embodiments 59-120, wherein a software-oriented (Python- Like) code focuses on and enables high-level logic programming (e.g., wavelet transforms, routing decisions) without requiring hardware expertise.

122. The method of any one of embodiments 59-121, wherein hardware-oriented (Verilog- Like) code embeds hardware constructs (like modules, wires, parameters) for direct FPGA or ASIC design.

123. The method of any one of embodiments 59-122, wherein an automatic conversion layer translates high-level, python-like code to perform neuromorphic tasks that comprise: a) in-silico functionality; b) in-vitro functionality; c) in-vivo functionality; d) ASIC wavelet transforms; e) dynamic neural signal ingestion; f) Al post-processing; g) biological-to-digital (ADC) conversion; h) digital-to-biological (DAC) conversion; i) Al preprocessing; or j) a combination thereof.

124. The method of any one of embodiments 59-123, wherein a plurality of CMOS-based probes are used in lieu of or in conjunction with systems to acquire or enhance real-time neural signal acquisition, or is used as the basis for biocomputation.

125. The method of any one of embodiments 59-124, wherein a plurality of CMOS-based probes function as both neural signal acquisition interfaces and in situ neuromorphic computing nodes, enabling real-time hybrid biological-digital processing for adaptive computation and neuromodulation. 126. The method of any one of embodiments 59-125, wherein CMOS-based probes function as in situ computing nodes, enhancing real-time neural signal interpretation and adaptive workload distribution.

127. The method of any one of embodiments 59-126, comprising biological neural tissue, an FPGA-based digital controller, and CMOS-based bioelectronic circuits, wherein the CMOS-based components autonomously execute localized neuromorphic computations to augment biological signal processing and dynamically balance workload across tissue clusters.

128. The method of any one of embodiments 59-127, that leverages CMOS-based neural interfaces for long-term tracking of synaptic plasticity, enabling adaptive neuromodulation and real-time recalibration of biological and digital processing layers.

129. The method of any one of embodiments 59-128, wherein CMOS-based neural probes dynamically adjust stimulation waveforms in response to real-time neural activity, using closed-loop FPGA-driven feedback to optimize bioelectrical modulation in living neural tissue.

130. The method of any one of embodiments 59-129, wherein a plurality of probes, a plurality of neural cells, a plurality of neural tissues, or a plurality of systems are dynamically reconfigured to target specific neural populations based on ongoing computational demands.

131. The method of any one of embodiments 59-130, wherein Genetically Encoded Voltage Indicators are used in lieu of or in conjunction with other signal acquisition systems to acquire or enhance real-time neural signal acquisition, or is used as the basis for biocomputation.

132. The method of any one of embodiments 59-131, wherein optogenetic methods are combined with fluorescent indicators in lieu of or in conjunction with other signal acquisition systems to acquire or enhance real-time neural signal acquisition, or is used as the basis for biocomputation.

133. The method of any one of embodiments 59-132, wherein Genetically Encoded Calcium Indicators are used in lieu of or in conjunction with other signal acquisition systems to acquire or enhance real-time neural signal acquisition, or is used as the basis for biocomputation.

134. The method of any one of embodiments 59-133, wherein optetrodes are used in lieu of or in conjunction with other signal acquisition systems to acquire or enhance real-time neural signal acquisition, or is used as the basis for biocomputation. 135. The method of any one of embodiments 59-134, wherein Functional Magnetic Resonance Imaging are used in lieu of or in conjunction with other signal acquisition systems to acquire or enhance real-time neural signal acquisition, or is used as the basis for biocomputation.

136. The method of any one of embodiments 59-135, wherein ultrasound methods are used in lieu of or in conjunction with other signal acquisition systems to acquire or enhance realtime neural signal acquisition, or is used as the basis for biocomputation.

137. The method of any one of embodiments 59-136, wherein a combination of electrical, optical, GECI’s, and/or ultrasound, is used to acquire or enhance real-time neural signal acquisition, or is used as the basis for biocomputation.

138. The method of any one of embodiments 59-137, comprising a module comprising a global controller configured to aggregate neural data and execute control commands, comprising subcomponents for signal collection, data aggregation, workload scoring, load management, health monitoring, task scheduling, checkpointing, adaptive stimulation, and/or microfluidic control.

139. The method of any one of embodiments 59-138, wherein comprising a module responsible for aggregating data from all clusters.

140. The method of any one of embodiments 59-139, wherein comprising a module responsible for extracting neural signal features using wavelet transforms, Fourier transforms, Lyapunov exponents, Hurst exponent, sample entropy, local field potentials, and gamma wave analysis to assess neural activity and tissue health.

141. The method of any one of embodiments 59-140, comprising a module continuously collects and aggregates neural data in real time to evaluate cluster performance, detect overload conditions, and identify signs of computational inefficiency or neural stress.

142. The method of any one of embodiments 59-141, wherein a dedicated digital signal processing modules within an FGPA or similar devices, perform real-time workload scoring by computing weighted summations of normalized neural signal features to determine computational load distribution.

143. The method of any one of embodiments 59-142, wherein a load manager receives workload scores from neural clusters, compares them against predefined thresholds, and coordinates with the health monitor to ensure computational load balancing.

144. The method of any one of embodiments 59-143, wherein a health monitor collects metrics such as chaotic spikes and gamma anomalies, generating health scores. 145. The method of any one of embodiments 59-144, wherein a task table and scheduler track active computational tasks, prioritize execution, and dynamically allocate tasks to neural clusters using on-chip block RAM.

146. The method of any one of embodiments 59-145, wherein the biocomputing system utilizes CMOS-based high density neural interfaces for real time signal acquisition, spike sorting, and neuromorphic digital signal processing (DSP), wherein the CMOS interfaces directly with FGPA - based processing units to extract computationally relevant neural features.

147. The method of any one of embodiments 59-146, wherein a checkpoint and transfer block manages intermediate computation states, transferring workload data between neural clusters.

148. The method of any one of embodiments 59-147, wherein an adaptive stimulation controller dynamically adjusts neural cluster pulse frequencies based on workload and stress signals.

149. The method of any one of embodiments 59-148, wherein a microfluidic control interface issues real-time commands to microcontrollers, adjusting chemical and nutrient flow rates based on neural cluster load and health status.

150. The method of any one of embodiments 59-149, wherein a biological-to-digital decoder converts analog neural signals into digital data using high-resolution ADCs, enabling further computational processing and analysis.

151. The method of any one of embodiments 59-150, wherein a pre-processing block processes raw digital neural signals by applying adaptive filtering, Fourier/wavelet transforms, and speech feature extraction.

152. The method of any one of embodiments 59-151, wherein a digital-to-biological encoder converts digital commands into analog stimulation pulses, delivering controlled electrical signals to neural tissue via dedicated analog signal lines.

153. The method of any one of embodiments 59-152, wherein an AI/ML-assisted postprocessing module applies machine learning techniques to filter signals, detect and correct anomalies.

154. The method of any one of embodiments 59-153, wherein an ASIC module stores spike trains and synaptic weights, and runs on-chip ML models to extract neural data features and perform predictive analysis while reducing latency and energy consumption. 155. The method of any one of embodiments 59-154, wherein a module stores spike trains and synaptic weights, and runs on-chip ML models to extract neural data features and perform predictive analysis while reducing latency and energy consumption.

156. The method of any one of embodiments 59-155, wherein a module runs on-chip ML models to extract neural data features and perform predictive analysis.

157. The method of any one of embodiments 59-156, wherein a top layer connectivity module facilitates external communication via Optical/TCP-IP links.

158. The method of any one of embodiments 59-157, wherein network-attached storage is utilized to save biocomputation information.

159. The method of any one of embodiments 59-158, wherein information can be pulled off at any layer or module in the system.

160. The method of any one of embodiments 59-159, wherein remote storage module connects to network-attached storage, external networks, and data centers via high-speed optical or serial links with TLS/SSL encryption.

161. The method of any one of embodiments 59-160, wherein animal or human neural tissues are decellularized, recellularized with new cells, and then used for biocomputation.

162. The method of any one of embodiments 59-161, wherein animal or human neural tissues are decellularized, recellularized with new cells, and then used for engraftment studies.

163. The method of any one of embodiments 59-162, wherein engineered cells or tissues with a cross-section larger than 1000 pm are utilized.

164. The method of any one of embodiments 59-163, wherein engineered cells or tissues smaller than 1000 pm are utilized.

165. The method of any one of embodiments 59-164, wherein no living cells are utilized.

166. The method of any one of embodiments 59-165, comprising a module to manage neural tasks, schedules spike-based computations, and facilitates secure communication between neural tissue and FPGA/ASIC devices.

167. The method of any one of embodiments 59-166, wherein an interpreter core translates BNPL scripts into FPGA bitstreams or ASIC Verilog modules.

168. The method of any one of embodiments 59-167, wherein a module automatically generates and analyzes neural spike waveforms, frequencies, and patterns.

169. The method of any one of embodiments 59-168, wherein a module adjusts computational models using AI/ML-based self-improvement techniques. 170. The method of any one of embodiments 59-169, wherein the bioreactor system is compatible with standard AI/ML frameworks (PyTorch, TensorFlow, JAX, MXNet, ONNX Runtime, Keras, and Hugging Face Transformers), data science and computational platforms (Jupyter Notebooks, Google Colab, Apache Zeppelin, Databricks, and RStudio), scientific computing and numerical libraries (NumPy, SciPy, MATLAB, Julia, Wolfram Mathematica, GNU Octave, and SymPy), big data and distributed computing frameworks (Apache Spark, Dask, Ray, TensorFlow Serving, and Kafka), embedded and edge computing platforms (NVIDIA Jetson, Arduino, Raspberry Pi, ESP32, and TinyML), high-performance computing (HPC) and GPU computing environments (CUD A, OpenCL, MPI, and SLURM), cloud computing and virtualization technologies (AWS SageMaker, Google Cloud Al, Azure Machine Learning, Docker, and Kubernetes), and general -purpose programming platforms and integrated development environments (VS Code, PyCharm, Eclipse, JetB rains Intelli J, Xcode, Visual Studio, and Emacs/Vim).

171. The method of any one of embodiments 59-170, wherein a module contains kernel services handle memory, scheduling, real-time neural event processing, and logging.

172. The method of any one of embodiments 59-171, wherein a module authenticates hardware reconfiguration commands and uses blockchain or tamper-proof logs.

173. The method of any one of embodiments 59-172, wherein a high-level programming language like python automatically generates Verilog code with proper ports and timing constraints.

174. The method of any one of embodiments 59-173, wherein a module generates a netlist, mapping logic gates, memory blocks, and connections to create FPGA configurations or ASIC layouts.

175. The method of any one of embodiments 59-174, wherein a module automates place- and-route (P&R), optimizing the physical arrangement of components and interconnections in a biocomputation context.

176. The method of any one of embodiments 59-175, wherein a module performs timing analysis to ensure signals meet required clock speeds and operational deadlines.

177. The method of any one of embodiments 59-176, wherein a module captures both spike events and low-frequency local field potentials in an in-vitro biocomputation context.

178. The method of any one of embodiments 59-177, wherein a notch filter is used at 50 or 60 Hz to eliminate electrical interference.

179. The method of any one of embodiments 59-178, wherein FPGA-based modules monitor noise and signal variance in real time, dynamically updating filter coefficients. 180. The method of any one of embodiments 59-179, wherein neural signals are formatted into standardized data structures, including consistent sample rates, labeling, formatting, time-stamped frames, fixed-length buffers, and/or structured packet.

181. The method of any one of embodiments 59-180, wherein LFPs are isolated using low- pass filtering to remove high-frequency spikes and noise.

182. The method of any one of embodiments 59-181, wherein wavelet and Fourier transforms are then applied to decompose LFPs into distinct frequency bands (e.g., theta, alpha, gamma).

183. The method of any one of embodiments 59-182, wherein wavelet transforms decompose signals into small wave packets, revealing both frequency content and temporal variations.

184. The method of any one of embodiments 59-183, wherein the system continuously monitors gamma activity to flag excessive power levels as potential indicators of neural tissue stress.

185. The method of any one of embodiments 59-184, wherein RQA is used to measure the recurrence of neural signal patterns over time, distinguishing structured neural communication from random noise.

186. The method of any one of embodiments 59-185, wherein Hurst exponent evaluates the long-term memory and persistence of neural signals.

187. The method of any one of embodiments 59-186, wherein sample entropy quantifies the predictability of neural signals.

188. The method of any one of embodiments 59-187, wherein a module enables the quantification and analysis of connectivity metrics, including techniques such as coherence analysis, phase-locking value (PLV), and Granger causality.

189. The method of any one of embodiments 59-188, wherein a module employs a healthload-synergy metric that integrates RQA, Hurst exponent, sample entropy, and connectivity metrics into a composite score.

190. The method of any one of embodiments 59-189, wherein a module utilizes a communication protocol for on-board and inter-board communication, particularly for dynamic neural tissue workload management.

191. The method of any one of embodiments 59-190, wherein Low-Voltage Differential Signaling is used for high-speed, reliable communication between components in a biocomputation context. 192. The method of any one of embodiments 59-191, wherein Serializer/Deserializer links are used to convert parallel data into a high-speed serial stream and back in a biocomputation context.

193. The method of any one of embodiments 59-191, wherein an Advanced extensible Interface (AXI) Bus is used for high-throughput, information exchange between FPGA-based chips and neural tissue clusters, or directly from neural clusters to neural clusters.

194. The method of any one of embodiments 59-193, wherein a module selects configurable data widths (e.g., 32-bit, 64-bit, 128-bit) are used in a biocomputation context.

195. The method of any one of embodiments 59-194, wherein a module selects the appropriate communication protocol, such as AXI for on-chip interconnects or LVDS/SERDES for remote modules.

196. The method of any one of embodiments 59-195, wherein a module performs real-time task redistribution in neural tissue computation, wherein workload, health, and cohesion metrics are continuously assessed to dynamically reallocate tasks, optimizing resource utilization while maintaining tissue stability.

197. The method of any one of embodiments 59-196, wherein neural computation task migration is performed based on spike features, gamma levels, complexity scores, and connectivity measures.

198. The method of any one of embodiments 59-197, wherein finite state machines are utilized in a biocomputation context to control task migration.

199. The method of any one of embodiments 59-198, wherein a Task Table is made wherein active sub-tasks are tracked with cluster assignment, priority, progress state, and/or workload metrics.

200. The method of any one of embodiments 59-199, wherein overloaded clusters in a neural computation system are calculated based on computational demand and neural signal stability.

201. The method of any one of embodiments 59-200, wherein a module enables checkpointing partial computational states in neural cluster workloads.

202. The method of any one of embodiments 59-201, wherein an FPGA dynamically selects underutilized clusters based on real-time workload.

203. The method of any one of embodiments 59-202, wherein a high-speed neural cluster communication system integrating LVDS, SERDES, and AXI bus architectures is utilized.

204. The method of any one of embodiments 59-203, wherein a module dynamically adjusts neural tissue stimulation and nutrient delivery. 205. The method of any one of embodiments 59-204, wherein a module uses a completion signal to reassign new tasks to a biocomputation cluster.

206. The method of any one of embodiments 59-205, wherein a module utilizes multichemical delivery with closed-loop feedback to optimize biochemical environments and maintain neural tissue viability for adaptive computation.

207. The method of any one of embodiments 59-206, wherein a module employs tissue calibration that gradually introduces tasks while adjusting stimulation and microfluidic flow to optimize synaptic plasticity.

208. The method of any one of embodiments 59-207, wherein a module employs self- healing and/or automatic reassignment that isolates damaged clusters and reallocates tasks to healthy clusters, employing regenerative interventions to extend tissue lifespan.

209. The method of any one of embodiments 59-208, wherein a module integrates realtime sensory inputs into neural clusters, improving system performance for applications in robotics, AR/VR, and medical diagnostics.

210. The method of any one of embodiments 59-209, wherein a module employs a trusted execution environment, utilizing specialized routines to encrypt data exchanged between the digital controller and neural tissue.

211. The method of any one of embodiments 59-210, wherein a module utilizes a crosslayer safety mechanism with a three-tier failsafe strategy including electrical cutoff, chemical protection, software hibernation, and/or digital signatures to prevent unauthorized reconfiguration.

212. The method of any one of embodiments 59-211, wherein a module uses an electrical cutoff safety mechanism.

213. The method of any one of embodiments 59-212, wherein a module uses a digital signatures safety mechanism.

214. The method of any one of embodiments 59-213, wherein a module uses thermal management system utilizing thermo-responsive microfluidics and smart heat sinks with integrated sensors to dynamically adjust cooling paths.

215. The method of any one of embodiments 59-214, wherein a module uses bioprocess media for electronic temperature control.

216. The method of any one of embodiments 59-215, wherein a module adjusts synaptic weights based on discrepancies between expected and actual neural outputs. 217. The method of any one of embodiments 59-216, wherein a module uses a stimulus generation system that dynamically adjusts microfluidic and stimulation parameters to optimize task-specific neural processing.

218. The method of any one of embodiments 59-217, wherein neural cells or tissues are trained to produce desired responses through repeated stimulation patterns, including reinforcement learning.

219. The method of any one of embodiments 59-218, wherein neural cells or tissues are trained to produce desired responses through repeated stimulation patterns, including nonreinforcement learning techniques.

220. The method of any one of embodiments 59-219, wherein a spike timing-dependent plasticity system adjusts synaptic strength based on the precise timing of neuronal spikes.

221. The method of any one of embodiments 59-220, wherein a Hebbian learning mechanism is utilized.

222. The method of any one of embodiments 59-221, wherein a homeostatic plasticity system that adjusts synaptic strengths.

223. The method of any one of embodiments 59-222, wherein training and inference can simultaneously occur across silicon and biological substrates.

224. The method of any one of embodiments 59-223, wherein a module dynamically decides if training and/or inference will occur in silicon vs biological substrates.

225. The method of any one of embodiments 59-224, wherein neural cells or tissues are used to implement multi-layered functions modular programming.

226. The method of any one of embodiments 59-225, wherein a module automatically converts between alphanumeric formats and neural spike trains in one or two directions.

227. The method of any one of embodiments 59-226, wherein a module trains neural tissues to execute basic scripts by associating specific input patterns with desired output responses.

228. The method of any one of embodiments 59-227, wherein digital data is automatically converted into neural spike trains and other signals using encoding schemes compatible with neural processing.

229. The method of any one of embodiments 59-228, wherein neural spike trains and other neural signals are automatically converted into digital data.

230. The method of any one of embodiments 59-229, wherein a module can directly encode alternative data types, including Binary, Hexadecimal, or Unicode, by efficiently encoding them into neural -compatible formats. 231. The method of any one of embodiments 59-230, wherein a module can directly decode alternative data types, including Binary, Hexadecimal, or Unicode, by efficiently decoding them from neural-compatible formats.

232. The method of any one of embodiments 59-231, wherein data is sent to neural cells or tissue(s) in a compressed format.

233. The method of any one of embodiments 59-232, wherein data is decompressed in neural cells or tissue(s).

234. The method of any one of embodiments 59-233, wherein a biocomputer autonomously refines data encoding and processing algorithms.

235. The method of any one of embodiments 59-234, wherein a biocomputer autonomously refines data encoding and processing algorithms across both neural and silicon based systems.

236. The method of any one of embodiments 59-235, wherein a module implements stratified layers to manage information at varying levels of abstraction.

237. The method of any one of embodiments 59-236, wherein a module performs matrix operations.

238. The method of any one of embodiments 59-237, wherein a module performs matrix operations within neural tissues or non-neural systems, leveraging parallel processing capabilities for efficient handling of complex mathematical computations, including transformations, optimizations, and modeling.

239. The method of any one of embodiments 59-238, wherein a module performs matrix operations within non-neural systems.

240. The method of any one of embodiments 59-239, wherein a module performs matrix operations within neural tissues or non-neural systems, leveraging parallel processing capabilities for efficient handling of complex mathematical computations, including transformations, optimizations, and modeling.

241. The method of any one of embodiments 59-240, wherein a module utilizes matrix math to enable efficient representation and manipulation of data structures.

242. The method of any one of embodiments 59-241, wherein matrix math is used as the basis for language models.

243. The method of any one of embodiments 59-242, wherein neural tissues with greater than >1000 micron cross section are used to process and generate human -like text for natural language processing. 244. The method of any one of embodiments 59-243, wherein a hybrid AI/ML model that distributes computational loads between silicon processors and neural tissues to efficiently manage intricate problem-solving tasks, offloading computationally expensive reasoning portions to neural tissues.

245. The method of any one of embodiments 59-244, wherein a hybrid AI/ML functions as a reasoning model.

246. The method of any one of embodiments 59-245, wherein a tissue with greater than >1000 micron cross section AI/ML functions as a layer in a reasoning model.

247. The method of any one of embodiments 59-246, wherein a biocomputation system enhances reasoning capabilities of a silicon model by offloading computational tasks to neural tissues.

248. The method of any one of embodiments 59-247, wherein a biocomputer is used for automated theorem proving.

249. The method of any one of embodiments 59-248, wherein a biocomputer is used for prime factorization and/or cryptanalysis.

250. The method of any one of embodiments 59-249, wherein a biocomputer is used for Self-contained AI/ML applications.

251. The method of any one of embodiments 59-250, wherein a biocomputer is used for Physics and simulation based discovery.

252. The method of any one of embodiments 59-251, wherein a biocomputer is used for Complex optimization problem calculations.

253. The method of any one of embodiments 59-252, wherein a biocomputer is used for Compression, encoding, and knowledge extraction applications.

254. The method of any one of embodiments 59-253, wherein a biocomputer is used for sensory processing applications.

255. The method of any one of embodiments 59-254, wherein a biocomputer is used for motor control applications.

256. The method of any one of embodiments 59-255, wherein a biocomputer is used for cognitive modeling applications.

257. The method of any one of embodiments 59-256, wherein a biocomputer is used for autonomous biological/silicon drone applications.

258. The method of any one of embodiments 59-256, wherein engineered neural tissue(s) are used for automated theorem proving in engineered tissue(s) with a cross section of >1000 microns. 259. The method of any one of embodiments 59-257, wherein engineered neural tissue(s) are used for prime factorization and/or cryptanalysis in engineered tissue(s) with a cross section of >1000 microns.

260. The method of any one of embodiments 59-259, wherein engineered neural tissue(s) are used for self-contained AI/ML applications in engineered tissue(s) with a cross section of >1000 microns.

261. The method of any one of embodiments 59-260, wherein engineered neural tissue(s) are used for Physics and simulation based discovery in engineered tissue(s) with a cross section of >1000 microns.

262. The method of any one of embodiments 59-261, wherein engineered neural tissue(s) are used for Complex optimization problem calculations in engineered tissue(s) with a cross section of >1000 microns.

263. The method of any one of embodiments 59-262, wherein engineered neural tissue(s) are used for compression, encoding, and knowledge extraction applications in engineered tissue(s) with a cross section of >1000 microns.

264. The method of any one of embodiments 59-263, wherein engineered neural tissue(s) are used for sensory processing applications in engineered tissue(s) with a cross section of >1000 microns.

265. The method of any one of embodiments 59-264, wherein engineered neural tissue(s) are used for motor control applications in engineered tissue(s) with a cross section of >1000 microns.

266. The method of any one of embodiments 59-265, wherein engineered neural tissue(s) are used for cognitive modeling applications in engineered tissue(s) with a cross section of >1000 microns.

267. The method of any one of embodiments 59-266, wherein engineered neural tissue(s) are used for autonomous biological/silicon drone applications in engineered tissue(s) with a cross section of >1000 microns.

268. The method of any one of embodiments 59-267, wherein an interface system provides specific stimulation events that vary in shape, time, and voltage.

269. The method of any one of embodiments 59-268, wherein biocomputer modules are single use.

270. The method of any one of embodiments 59-269, wherein biocomputer modules can be washed and reused. 271. The method of any one of embodiments 59-270, wherein neural cells and/or tissues are segmented by memory functions, repetitive functions, delay functions, interpretations, translations.

272. The method of any one of embodiments 59-271, wherein an artificial life is created but divided across computation components to limit sentience.

273. The method of any one of embodiments 59-272, wherein electrode shapes are varied based on use.

274. The method of any one of embodiments 59-273, wherein electrodes have impedance matched back to circuits.

275. The method of any one of embodiments 59-274, wherein electrodes are bimetallic to allow thermal movements.

276. The method of any one of embodiments 59-275, wherein electrodes have specific electrochemical orientation.

277. The method of any one of embodiments 59-276, wherein the biocomputing system provides initial bias to prevent inadvertent stimulation or sensing or provide gating threshold function.

278. The method of any one of embodiments 59-277, wherein any electrode can be connected directly to any other to force specific neural paths in thinking.

279. The method of any one of embodiments 59-278, wherein non-functioning, non-active neurons can be disconnected from the network.

280. The method of any one of embodiments 59-279, wherein non active neurons can be connected to specialized stimulation for regeneration.

281. The method of any one of embodiments 59-280, wherein a passive switching system provides neuron corrected speed and signaling.

282. The method of any one of embodiments 59-281, wherein neurons can control a switch to provide new self-generated hybrid thinking routes.

283. The method of any one of embodiments 59-282, wherein neurons can control functions outside of their specific well, outside a given module, and/or outside the system.

284. The method of any one of embodiments 59-283, wherein a passive switch is enhanced with active switch.

285. The method of any one of embodiments 59-284, wherein buffer signals assure capability to traverse electrical circuit paths switched in system.

286. The method of any one of embodiments 59-285, wherein the system can convert given action potentials into others - electronic equivalent of inverting a digital signal. 287. The method of any one of embodiments 59-286, wherein delays in action potentials is leveraged to achieve specific thinking mechanisms - circuit equivalent of delay line.

288. The method of any one of embodiments 59-287, wherein electronic circuits provide digital logic type functions - addition, subtraction, multiplication, etc. in a biocomputation context.

289. The method of any one of embodiments 59-288, wherein any electrode can be connected to the sensing or stimulation circuit or can be left open.

290. The method of any one of embodiments 59-289, wherein the system can turn off specific activity by ‘grounding’ neural paths, with specific stimulation or biasing.

291. The method of any one of embodiments 59-290, wherein the configuration forces axons and other connections through tunnels to align neural circuits to electrode paths.

292. The method of any one of embodiments 59-291, wherein the system stimulates along the length of an axon in multiple places per axon.

293. The method of any one of embodiments 59-292, wherein a through silicon via is used to guide dendrite or neural growth paths.

294. The method of any one of embodiments 59-293, wherein a through silicon via is used to connect multiple tissues.

295. The method of any one of embodiments 59-294, wherein a through silicon via is used to guide specific neural connections.

296. The method of any one of embodiments 59-295, wherein a through silicon via is used to deliver media to neural cultures.

297. The method of any one of embodiments 59-296, wherein electrode arrays are vertically stacked.

298. The method of any one of embodiments 59-297, wherein electrode arrays and tissue cultures are alternatively stacked.

299. The method of any one of embodiments 59-298, wherein media is delivered through micromachined channels in the bioreactor.

300. The method of any one of embodiments 59-299, wherein at least one through wafer vias (TSV) and electrodes are built in vertical stacks to provide compact connections.

301. The method of embodiment 300, wherein the at least one TSVs connect to a circuit board to eliminate wirebonds and accommodate bumping and board mount.

302. The method of any one of embodiments 59-301, wherein the system is sealed to prevent contamination. 303. The method of any one of embodiments 59-302, wherein biological sensing is accomplished through developed cell systems for sight, smell, and/or touch sensing.

304. The method of any one of embodiments 59-303, wherein neurons directly connect to digital camera chip.

305. The method of any one of embodiments 59-304, wherein neurons directly connect to an inertial sensing system(s).

306. The method of any one of embodiments 59-305, wherein each electrode has its own communication channel in a biocomputation context.

307. The method of any one of embodiments 59-306, wherein multiplexing is used to connect multiple electrodes through a single digital channel in a biocomputation context.

308. The method of any one of embodiments 59-307, wherein cultures are printed or assembled directly onto electrode panels.

309. The method of any one of embodiments 59-308, wherein panels containing electrodes or interfaced with electrodes are coated with biogrowth materials.

310. The method of any one of embodiments 59-309, wherein panels are made of paper.

311. The method of any one of embodiments 59-310, wherein stimulation and sensing electrodes are separate circuit paths.

312. The method of any one of embodiments 59-311, wherein stimulation and sensing electrodes or connected via the same electrical paths and are switched on and off.

313. The method of any one of embodiments 59-312, wherein stimulation and sensing electrodes are interspersed on same panel, alternating every other electrode, or 10: 1 or 100:1 or other necessary connection.

314. The method of any one of embodiments 59-313, wherein stimulation can be achieved with a non-DAC circuit.

315. The method of any one of embodiments 59-314, wherein DAC stimulation can be tailored to mimic signals or create new stimulation patterns that provide new thinking functions.

316. The method of any one of embodiments 59-315, wherein sensing of action potentials are used as a signal.

317. The method of any one of embodiments 59-316, wherein modulation is used to create new forms of stimulation that use electrochemical resonance.

318. A method for converting annotated neural recordings into biocomputational inputs, the method comprising: a) capturing an annotated neuroscience dataset during subject cognitive activities using brain recording devices; b) processing the annotated neuroscience dataset to extract neural data comprise a plurality of neural spike trains; and c) introducing the plurality of neural spike trains into an engineered tissue via a brainmachine interface.

319. The method of embodiment 318, wherein the annotated neuroscience dataset are derived from in vivo measurements.

320. The method of embodiment 318 or 319, wherein the annotated neuroscience dataset comprises emergent data generated via biocomputation.

321. The method of any one of embodiments 318-320, further comprising a tokenization module configured to convert the neural data into discrete tokens representing distinct neural activity patterns, raw data, or concepts.

322. The method of embodiment 321, wherein the stimulation inputs applied to the biocomputing tissue comprise a combination of tokenized waveforms and/or spike trains and simple spike signals that encode values such as numbers or letters.

323. A method for encoding and/or dual encoding of neural data, comprising: a) tokenizing neural data comprising a plurality of neural spike trains derived from annotated neuroscience datasets into discrete symbols; and/or b) forming a dual stimulation scheme in which both tokenized waveform representations and simple numerical spike signals are delivered to biocomputational tissue.

324. The method of embodiment 323, further comprising an encoding module that converts neural waveform signals into ASCII and/or binary representations, thereby facilitating digital -to-bi ol ogi cal conversi on .

325. The method of embodiment 324, wherein the converted waveform representations are treated as multi-dimensional vectors that serve as embeddings for conceptual information.

326. A method for processing and manipulating conceptual information, the method comprising: a) converting neural spike train-derived waveforms into multi-dimensional vectors; b) tokenizing said vectors to form literal and/or conceptual embeddings; and c) combining embeddings such that, for example, the addition of an embedding for “king” with a component representing “woman” yields an embedding representing “queen.” 327. The method of any one of embodiments 318-326, further comprising processing nodes operable in both a biological tissue layer and a digital computing layer to allow dynamic conversion, transmission, and processing of tokenized neural data.

328. A method for training and inference in biocomputation networks, comprising the steps of: a) Preserving the temporal and quantitative attributes of the original annotated neural recordings during conversion into spike trains; b) Delivering the neural information including but not limited to spike trains and, in some embodiments, their tokenized representations to biocomputational tissues; and c) Processing these signals to enable language-related and concept-based tasks in a manner inspired by biological neural networks.

329. The method of any one of embodiments 318-328, wherein the annotated neural data is further processed to generate emergent properties that can be used to enhance biocomputational performance.

330. A method for comprehensive encoding and decoding of neural information, comprising integrating layered conversion techniques that translate biological neural signals into digital representations (including ASCII or binary), forming multi-dimensional embedding vectors that preserve semantic integrity.

331. The method of any one of embodiments 318-330, wherein the tokenized waveform representations and simple spike signals are used to modulate and control biocomputational tissue dynamics in real time.

332. The method of any one of embodiments 318-331, further comprising a control mechanism that dynamically reconfigures neural encoding parameters to support scalable, adaptable biocomputation across a parallel, tissue-based computational network.

333. The method of any one of embodiments 318-332, wherein the annotated neuroscience dataset is employed to correlate conceptual data with specific neural spike patterns for enhanced biocomputational processing.

334. One or more non-transitory computer-readable media comprising computerexecutable instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any one of embodiments 59-333. V. EXAMPLES

[0291] The following examples are included for illustrative purposes only and are not intended to limit the scope of the invention.

Example 1: Tissue Engineering quality control

[0292] Quality by design is the concept that ‘quality’ should be designed into each component of product manufacturing rather than tested in. This is done through defining product attributes then conducting multi -factorial experiments to see how deviation in process inputs impact process outputs. An example is depicted in FIGS. 11A-11E.

[0293] Preliminary quality targets include cell viability, viable cell density and tissue electrical pulse quality as measured by a proprietary multi -el ectrode array score (FIG. 11 A). [0294] Inputs for an engineered brain tissue include neuron concentration, astrocyte concentration, oligodendrocyte concentration, undifferentiated stem cell concentration, primary tissue concentration, and T-cell concentration(FIG. 11B).

[0295] A secondary design of experiment looks at tissue vascularization washes including a non-disclosed, iterative mixture of endothelial cells, angioblast, muscle-like cells, other cell types, and categorical vascularization methods including vascularization by seeding open cavities within tissue, vascularization by seeding into closed cavities in the tissue, and a ‘direct’ vascularization approach by which small channels attempt to directly mimic vascular networks in lumen size and function (as opposed to merely a seed point from which cells can grow) (FIG. 11C). Various multifactorial experiments which look at process and additive manufacturing parameters are underway (FIG. 11D). Simulated data shows how various concentrations of cells may impact quality attributes as measured by validated analytical methods (FIG. HE).

Example 2: Bioprinted tissues

[0296] Examples of bioprinted tissues are depicted in FIGS. 12A-12D. three-dimensional bioprinted and injection molded, cell-based tissues were manufactured using biocompatible and semipermeable scaffolds described herein. The tissue was cultured in the two-bioreactor system with growth factors and a custom-made three-dimensional multi-electrode array to promote cell-to-cell connections. By creating an intact tissue where cells are fully integrated into their environment, cell survivability and the duration of therapeutic effects was extended. [0297] FIG. 12A depicts a recently printed tissue with minimal processing (bottom) compared to tissue incubated within tissue bioreactor (top). To test a primary tissue’s ability to integrate into secondary tissue, a small rectangle was vivisected from the secondary tissue. The cavity was then filled with a biocompatible gel. Results are depicted in FIG. 12B. A portion of the primary engineered tissue was implanted into the biocompatible gel in FIG. 12C. This tissue-in-tissue modality demonstrates a potential therapeutic modality as well as a non-clinical investigation modality.

[0298] Tests of tendon style cells were also conducted using multifactorial experimentation. Here, tissue was printed directly with plastic clips using a Cellink BioX thermoplastic printhead. Following incubation the clips were anchored to a strength testing device to test the impact of various cell, scaffolding, and process parameters. Tissues are shown in FIG. 12D.

Example 3: Cellular Agriculture

[0299] Early tests which looked at large-scale tissue manufacturing were conducted using fish cells. The process is shown in FIGS. 13A-13E.

[0300] Using Computer Assisted Design, 3-dimensional models of fat, meat, and vascular components for a salmon fillet were created. Vascular components are absent from the image presented. An example is depicted in FIG. 13A. By growing cells in-vitro, mixing them with scaffold material, and printing the CAD models, a salmon fillet in the BioX three- dimensional printer was printed as shown in FIG. 13B.

[0301] The vascularized fillet was cultured with electrical stimulation until cells align. Post incubation changes in tissue physiology was detectable in FIG. 13C. Because the cells have aligned the product resembled intact animal tissue (representative example in FIG. 3D) and can be cooked. Shown is a 5.5 cm long, 15.5-gram cellular agriculture chinook salmon fillet in FIG. 13E.

Example 4: Manufacturing tissue for biocomputing

[0302] In addition to regenerative medicine and cellular agriculture approaches the combination of a vascularized three-dimensional brain-like tissue and three-dimensional multi-electrode arrays allows for biocomputing with a combination of both traditional silicon and living neuronal computation as depicted in FIG. 14. A traditional artificial neural network with multiple computational nodes feeds electrical stimulus to the multi-electrode array, the living neurons then serve as a hidden layer, in which computation is performed. Cell derived electrical stimulus is recorded by the multi-electrode array and then deciphered by a secondary silicon artificial neural network. [0303] In addition to a biocomputer where information follows a linear path, from silicon neural network model a, to biological components, to silicon neural network model b, information may back-propagate through the network as represented by the arrows (FIG. 15A). The three-dimensional arrangement of probes and neural information travelling in a three-dimensional space may also be important, as represented by the two multi -electrode array shafts(FIG. 15B). Neural networks need not always compute every combination. Shown is a variety of combinations, including single electrode nodes, pairs, trifectas, etc., with a hidden living neural layer (FIG. 15C). Neuron network size can change as represented by the 3 node x 2 level input layers. The surface array also takes part in biocomputing (FIG. 15D)

[0304] A central data repository is used to link process control in an Arduino environment and stimulation/recording in a python environment with or without TensorFlow and Juypter notebook integration for machine based neural network integration. An example is depicted in FIG. 16.

[0305] Intan RHX is a free, powerful data acquisition software that displays and records electrophysiological signals from the Intan Stim/Recording controller. The screenshot shows standard pulse settings (FIG. 18A). Recording of individual neural spikes across a multielectrode array is shown in FIG. 18B.

[0306] Preliminary experiments used the Weighted Neural Spike Score to determine neural spike quality. In these studies brain-like tissue is stimulated with standard settings during a training phase.

[0307] During the training phase, the process is sometimes controlled through automated systems. Each time the process is controlled, concurrent electrophysiological stimulation is provided at the center of the range for neural spike pulse train and waveform metrics. For example, pH is controlled by pumping in small quantities of sodium bicarbonate to raise the pH each time required by the bioprocess. During the training phase each time the pH pump is engaged, the Intan system pulses at median spike values shown in FIG. 18A. A simplified version of these neural spike values in FIG. 19 represents a two spike pulse of strength 5, and interval duration i.

[0308] During the testing phase, the process is sometimes controlled by a weighted pulse score. This can be derived from artificial neural networks in such as those in the TensorFlow system or by more simple arithmetic means. Using direct means and the training example above, a two-spike pulse of strength 5 which repeatedly fired at interval i, would result in the pH pump moving for the same time and same speed as during the training phase. A weaker or less frequent spike would mean less pH control, a stronger spike would equate to stronger control. During either phase pH is not allowed to leave its control range of 7.05-7.45 as this would result in a non-viable tissue.

[0309] This system tests the brain’s ability to synthesize information and potentially alter control of its own environment.

[0310] Initial experiments tested biocomputing power and capability. Three initial experiments were conducted, in which tissues were bioprinted and incubated within the two- bioreactor system tissue reactor. Artificial neural spikes were provided through 2 electrodes for 12 hours/day, until 7 days after neural spikes consistently exceeded background noise levels. Neural spike activity was recorded.

[0311] Experiment FIG. 20A served as a control. In this control training spikes were not correlated to process control, nor did neural spikes control process parameters. Neural activity picks up around 6 days after initial spikes are recorded above the background noise cutoff. These spikes consistently scored lower weighted neural spike scores (less strong and frequent spikes) than the training spikes.

[0312] In experiment FIG. 20B, training pulses coincided with bioprocess control for a given parameter as described in FIG. 15, except that the parameter under control was not necessarily pH. Here, the neural spikes gain intensity and frequency immediately upon recording over the background threshold, as shown by the weighted neural spike score. Once the training pulses were disengaged, there was an initial drop off weighted neural spike score, suggesting there may have been a correlation between training fires and neural spikes. On the subsequent day, neural spikes were stronger than the training spikes. The associated process control in turn increased beyond setpoint in the provided training regimen and the associated process parameter shifted to a higher point in its control range. This suggests the neurons may have found it beneficial to alter the process control points. Similar experiments with different process parameters did not all provide comparable results.

[0313] In experiment FIG. 20C., the training pulses did not correlate with process control. However, the process parameter in question was only controlled when neurons fired. We see a significant lag phase lasting 15 days after the first recording above baseline. When compared to experiment FIG. 20A or 20B, the lag phase was significantly longer. This is likely due to poor process control and thus lower quality tissue. However, after 15 days, neural activity increased as demonstrated by the increased weighted neural spike scores. These scores were higher than the uncontrolled experiment A, again suggesting that the neurons may be responding to the process parameter in play and driving them to a higher level of control. In either case, both experiment b and c demonstrate the neuron’s ability to control an electronic system.

Example 5: Additional software, hardware, and wetware considerations

[0314] FIG. 22. depicts a modular system with most likely communication paths. In some embodiments, the system will be composed of these elements. Module 1 comprises a neural tissue growth and microfluidics system. Module 1 may comprise CMOS neuromorphic probes. Module 1 may communicate with Module 2, Module 6, Module 7, Module 8, and/or Module 9. Module 2 may comprise a neural tissue processing unit and/or a CMOS neuromorphic chip. Module 2 may communicate with Module 1, Module 3, Module 4, Module 5, Module 6, and/or Module 8. Module 3 may comprise adaptive workflow balancing. Module 3 may communicate with Module 1, Module 2, and/or Module 8. Module 4 may comprise a biological-to-digital decoder. Module 4 may communicate with Module 2, Module 5, and/or Module 6. Module 5 may comprise a based pre-processing unit. Module 5 may communicate with Module 2, Module 4, Module 6, Module 7, and/or Module 8. Module 6 may comprise a Digital-to-Biological encoder and a CMOS neuromorphic chip. Module 6 may communicate with Module 1, Module 2, Module 4, Module 5, Module 8, and/or Module 9. Module 7 may comprise an AI=assisted post-processing unit. Module 7 may communicate with Module 1, Module 5, and/or Module 8. Module 8 may comprise an ASIC & CMOS module. Module 8 may communicate with Module 1, Module 6, Module 7, and/or Module 9. Module 9 may comprise a connectivity module. Module 9 may communicate with Module 1, Module 2, Module 3, Module 6, Module 7 and/or Module 8. Module 10 may comprise cloud and local storage. Module 10 may communicate with Module 9 and/or Module 11. Module 11 may comprise cloud and local storage. Module 11 may communicate with Module 0 and/or Module 10. Not shown, in some embodiments, information can travel from every module to every other module. This is particularly valuable for retaining information and building models.

[0315] The Bioreactor/Biocomputer described herein is a modular system with many potential configurations. FIG. 23 illustrates an alternative iteration of our modular system, divided into three blocks: the MEA/biocomputing core 2301, the interface system 2302, and the analysis system 2303. In some embodiments, the MEA system contains tissue samples, neurons, and electrodes. In some embodiments the MEA system 2301 comprises a well 2304, fluid 2305, a multi-electrode array 2306 and tissue sample 2307. In some embodiments, the interface system 2302, which includes analog-to-digital converters (ADCs) 2309 and digital- to-analog converters (DACs) 2312, connects the MEA system 2301 to the digital computing components. In some embodiments, the interface system 2302 comprises an amplifier (AMP) 2308 that amplifies the signal sent to the ADC 2309. In some embodiments, the interface system 2302 comprises an amplifier 2311 that amplifies the signal sent to the DAC. In some embodiments, the interface system 2302 comprises a FPGA 2310 that receives the input from the ADC 2309 and the DAC 2312. In some embodiments, the analysis system 2303 comprises an FPGA 2313 and a neuromorphic chip 2315. In some embodiments, the analysis system 2303 comprises a memory 2314 and an interface 2316. The modular design supports bidirectional communication between the neural tissue and digital processors, with fluid control and temperature regulation for cell viability. In some embodiments, a bus 2317 connects the MEA system 2301 and the interface system 2302 and allows for electronic communication between the systems. In some embodiments, a bus 2319 connects the analysis system 2303 and the interface system 2302 and allows for electronic communication between the systems. In some configurations, 2D and 3D electrode arrays surround and penetrate the tissues from all sides to maximize tissue connectivity. In some embodiments, this scalable system captures neural signals, processes them through the FPGA and neuromorphic chip, and delivers real-time stimulation to the tissue. In some embodiments, this modular, stackable design mirrors modem microprocessor integrated circuits, forming a scalable, parallel computing architecture. This closed-loop process enables real-time interaction with the neural culture, forming the foundation of the biocomputing system.

[0316] FIG. 24 illustrates another biocomputer concept module. In some configurations, the module features a single engineered tissue centrally positioned (upper left image), interfaced with integrated vertically stacked circuits. In some embodiments, the integrated circuit, mounted on a printed circuit board, contains electrodes that interface directly with the cell culture. In some embodiments, the system is sealed to form a liquid chamber, with fluid entry and exit ports visible (upper right image). In some embodiments, the rack-mount diagram (below) illustrates multiple stacked biocomputer modules, each with electrical and fluidic connectors at the top for scalable integration. In some embodiments, the design envisions thousands of these modules operating within a master system. In some embodiments, each module is designed to support an approximately 10 cm * 10 cm x 3 cm engineered tissue.. In some embodiments, they include fluid circulation and filtration, ensuring sustained operation. Modules are designed to be disposable after their operational lifespan. [0317] FIG. 25 presents a top and side view of the integrated circuit (IC) concept for a multi-electrode array (MEA) system. In some embodiments, the IC is fabricated using standard complementary metal-oxide-semiconductor (CMOS) techniques, incorporating tungsten electrodes and 2D or 3D structures on the die’s top surface. In some embodiments, the side view (left) illustrates an oxide layer with electrodes that may either protrude beyond or remain planar with the oxide surface. In some embodiments, the active area, indicated by cross-hatching, includes a through-silicon via (TSV) for connectivity to the opposite side of the die, allowing for attachment to a printed circuit board via bump connections. In some embodiments, electrode shapes are not constrained to a Manhattan geometry. While circular electrodes are common, alternative shapes include elliptical, bean-shaped, or configurations combining small circular electrodes with larger surrounding structures. In some embodiments, electrodes may also feature 3D geometries, such as protruding tips, sharp or concave formations, or other micromachined structures, tailored to system requirements. Combined with various 3D electrode array geometries, we can leverage essentially any shape required for biocomputation.

[0318] FIG. 26 illustrates signal flow between the MEA and the bioculture, emphasizing a local, largely passive switching system for information transfer. Typically, digital signals are converted to analog for stimulation electrodes. The diagram depicts two MEA bioculture networks, either as separate culture wells, or within a single well system. In some embodiments, this setup enables direct electrode-to-electrode signal transfer without routing through the digital computing network. In some embodiments, the switching network facilitates signal redirection between electrodes or reintroduces signals into the bioculture as feedback, with or without modification (in amplitude or phase, etc.) stabilizing network activity and enhancing signaling control, similar to feedback mechanisms in electronic circuits. In some embodiments, this switching can be quasi-static or dynamically adjusted based on control system input. A key advantage is eliminating the need for full digitization, which introduces latency that may be too slow for real-time control of the culture system. In some embodiments, this approach bypasses the need for neural signal interpretation, simply redirecting pulses to the appropriate targets.

[0319] FIG. 27 shows a sensed signal coming from the upper left that is buffered and then directed into a switching circuit. In some embodiments, a switch controls or is controlled by a larger computing network or the local MEA subsystem/network. In some embodiments, this sensed signal could be switched to another stimulation electrode on the same or on another MEA network. In some embodiments, the communication can be handled locally between one MEA network to another. In some embodiments, the action potential can be transformed into a prescribed stimulation signal to optimize communication. In some embodiments, delays and/or transformations can be introduced, allowing the computing network to stabilize before signals propagate to other electrodes or cultures, enhancing controlled signal processing.

[0320] FIG. 28 shows a multiplexing block. In some embodiments, multiplexing enables dynamic electrode connectivity while maintaining parallel pathways into the computing network. In some embodiments, each electrode connects to a dedicated SPST switch, allowing signals to be routed, fanned out, or switched to other electrodes. In some embodiments, a passive block may incorporate a cross-connect network to support multiple switching mechanisms. In some embodiments, this configuration enables adaptive control over neural information transfer, allowing initial connectivity to evolve as the culture differentiates and trains. In some embodiments, direct module-to-module switching bypasses the digital computing network when beneficial, optimizing signal fidelity and processing efficiency.

[0321] FIG. 29 shows neural cultures grown on a substrate (cross-hatched). In some embodiments, the substrate has small channels that facilitate neuron, or dendrite growth. In some embodiments, electrodes in the channel allow communication with neurons or dendrites at specific locations. The benefit is the sensing and stimulation on the surface of the axon rather than the surface of the larger culture. In some embodiments, the dendrites could grow from individual neurons, or from the larger tissues. In some embodiments, micro machined channels within the substrate can be created, with or without breathable membranes. In some embodiments, these tunnels can be used for electrode positioning or microfluidics.

[0322] FIG. 30 shows a cross section of a silicon chip (SI) with a protective oxide layer (OXIDE). In some embodiments, a micromachined channel traverses through the oxide and silicon layers. In some embodiments, these channels provide a series of very tiny openings in the silicon wafer on the micron level. In some embodiments, a cell/tissue culture would exist above and below this chip and individual cells or axons would connect in predetermined locations. In some embodiments, this ensures connective reproducibility. In this way the neurons behave in a way reminiscent of a through-silicon via (TSV) that passes completely through a silicon wafer or die.

[0323] FIG. 31 shows an integrated MEA/Well concept diagram. In some embodiments, a stack of several tissue structures layered on MEA structures all within a well system. In some embodiments, these MEA Structures are shown entering and exiting electrically from the sides of the well. In some embodiments, media enters the system to the left and exits to the right. In some embodiments, the system would have a cover, either with or without a penetrating 3D electrode array. In some embodiments, the tissue could continue to compute in 3D and function as a single system by virtue of the axon connections that bridge through the MEA as shown in the previous image. In some embodiments, STIM and SENSE electrodes enter and exit from different sides but this need not be the case.

[0324] FIG. 32 illustrates a counter approach to conventional integrated circuit (IC) electrode systems. In standard designs, the stimulation and sensing mechanisms are connected to a single electrode, with the sensing mechanism typically AC-coupled. While AC coupling prevents damaging DC voltage levels, it also eliminates potentially valuable DC information. In some embodiments, of media-based systems, the stimulation signal conducts through the medium, directly influencing the sensing mechanism. Suppliers have indicated that direct DC coupling may introduce problematic voltage levels. However, in some embodiments, with sufficient dynamic range, DC signals could provide useful insights, such as pH levels, media composition, or overall tissue health. In some embodiments, the system integrates a variety of sensing mechanisms, including interdigitated electrodes, to monitor media conditions, tissue viability, and circuit health. In some embodiments, to optimize functionality, the design incorporates approximately 1,200 stimulation inputs while maintaining a smaller number of sensing channels. The methodology separates stimulation and sensing into distinct electronic subsystems, achieving an optimal ratio of 128 sensing channels while maximizing stimulation input capacity.

[0325] FIG. 33 illustrates a method for guiding the growth of neural cultures into defined positions to biomechanically structure their organization, thereby influencing and enhancing their predetermined bio-function. The fundamental concept is that form dictates function, allowing for controlled neural network development. In some embodiments, a simple implementation involves perforations in the MEA panel, facilitating gas exchange and media access for the cell culture. In some embodiments, beyond nutrient access, these openings enable axonal growth between separate tissue cultures. In some embodiments, the electrode interface is primarily formed at these openings, where electrodes interact directly with axons rather than with the tissue surface. In some embodiments, this configuration allows for precise signal communication between neural structures. In some embodiments, by stacking multiple layers, greater control over both function and system configuration can be achieved. In some embodiments, micro-machined channels within the substrate can be fabricated, with or without breathable membranes, to further refine electrode positioning or integrate microfluidic systems for enhanced tissue maintenance and interaction.

[0326] FIG. 34 shows an MEA from the top view of looking down on the electrodes in a Manhattan geometry view. The diagram shows sense electrodes (open circle) and the stimulation electrodes (filled circle) interspersed in a pattern. In some embodiments, stimulation and sense electrodes could be on the same panels or they could be on different panels, or in different ratios than shown. In some embodiments, electrode paths can be hardwired or multiplexed. In some embodiments, we can control any input and output to be physically located where the tissue requires. Presently, users of existing hardware grow the cell cultures on the electrode arrays and then program which inputs and outputs connect to the tissue based on where the cell culture has grown. In some embodiments, we do not need that mechanism because we have prescribed and dictated structure.

[0327] FIG. 35 illustrates the control and customization of stimulation pulse shapes within the system. In some embodiments, the number of channels is scalable, and stimulation (stim) and sensing (sense) functions can be independently assigned to each electrode. In some embodiments, stimulation pulses can be precisely shaped, and the sensing signals can be DC- coupled as needed. In some embodiments, stimulation signals can be generated without a DAC, as the fundamental signal can be produced directly from the FPGA. In some embodiments, once the required stimulation signal characteristics are determined, they are managed accordingly. In some embodiments, a DAC can be used to further refine the stimulation signal shape. A key innovation not previously described in the literature is the ability to generate asymmetric stimulation signals in both time and amplitude while maintaining a balanced charge. This capability may have neuromorphic significance, allowing for signal encoding that conveys specific information to the neural tissue. In some embodiments, tailored signal shapes could influence neural activity in a manner that facilitates functional communication and interpretation within the cultured neural network.

Example 6: An example scenario of dynamic neural tissue workload management is provided, in which a biocomputer is performing LLM and Protein Folding tasks.

[0328] LLM Task (High Priority): Cluster A is running partial matrix multiplications for a real-time LLM query. Its load score hits 0.85, and it shows slightly chaotic signals (sample entropy spiking). The global FPGA flags it as overloaded.

[0329] Protein Folding (Medium Priority): Cluster B is working on a background proteinfolding sub-task. Its load score is only 0.25. The global FPGA sees B is underutilized. [0330] Task Migration: The FPGA instructs Cluster A to checkpoint part of the LLM matrix computation. Cluster A’s partial state is transferred to Cluster B via LVDS/SERDES. Cluster B loads the matrix data and continues that portion of the LLM sub-task, temporarily pausing or sharing CPU cycles with the protein-folding sub-task.

[0331] Stimulation & Nutrient Tuning: Since LLM tasks are high priority, the FPGA slightly increases stimulation pulses to Cluster B, ensuring it handle the sudden additional load. If B’s signals later indicate rising stress, the FPGA might reduce pulses or deliver extra nutrients via the microfluidic system.

[0332] Completion: Once B completes the partial matrix operation, the result is sent back to the global FPGA or stored in a checkpoint. Cluster A can continue its portion of the LLM math with a lower load or focus on the next sub-task if it’s recovered.

Example 7: Software/Hardware co-design for the Hybrid Bio Operating System

[0333] This section details the software/hardware co-design process for the Hybrid Bio Operating System (HBOS), focusing on how it merges Verilog-like and Python-like scripting to dynamically program FPGAs or generate ASIC snippets.

[0334] By merging Verilog-like and Python-like scripting in HBOS, developers can dynamically configure FPGA modules, test Al or DSP routines in real time, and export stable designs to ASICs for mass production. This approach provides fast iteration, by allowing users to prototype new algorithms quickly on FPGA; efficient deployment, by locking in stable designs as ASIC blocks for lower power and cost; and a robust integration with BNPL through use of a user-friendly language for both software and hardware tasks, bridging biological signals and digital logic. This detailed process, from module definition to FPGA testing and final ASIC generation, is fundamental to our vision for an adaptive, scalable neuromorphic computing platform that unifies living neural tissue with modern hardware design.

[0335] The Hybrid Bio Operating System (HBOS) and Bio-Neural Programming Language (BNPL) Software-Oriented BNPL (Python-Like) code, focuses on high-level logic (e.g., wavelet transforms, routing decisions) without requiring hardware expertise. This is ideal for software engineers or data scientists. Meanwhile Hardware-Oriented BNPL (Verilog-Like) embeds hardware constructs (like modules, wires, parameters) for direct FPGA or ASIC design this is ideal for hardware engineers needing fine-grained control. [0336] In addition, HBOS includes an automatic conversion layer that translates the high- level Python-like BNPL code into Verilog-level code. This lets software developers write all functionality, ranging from ASIC wavelet transforms to dynamic neural signal ingestion, Al post-processing, biological-to-digital (ADC) conversion, digital-to-biological (DAC) conversion, and Al preprocessing, in a Python-like syntax, while HBOS automatically generates the low-level hardware descriptions required for programming FPGAs, ASICs, and the global FPGA controller.

[0337] This conversion abstracts all hardware-level details and makes it possible for software engineers to define how analog neural signals should be digitized using high-level code, without needing to write the low-level Verilog. The result is an efficient interface between biological sensors and digital processing, essential for accurate neuromorphic computation.

[0338] These protocols make sure that rapid, reliable, and secure data transfer occurs across all system layers, allowing real-time closed-loop feedback and dynamic hardware reconfiguration. By providing two distinct BNPL views, one in Python-like pseudocode for software engineers and another in Verilog-like syntax for hardware engineers, HBOS allows the same functionality to be developed at different levels of abstraction. Software engineers can write high-level BNPL code without worrying about low-level hardware details, as HBOS automatically converts this code into Verilog for FPGAs, biological-to-digital encoders, digital-to-biological encoders, and ASIC programming, including the global FPGA controller.

[0339] While preferred embodiments of the present invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions will now occur to those skilled in the art without departing from the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention. It is intended that the following claims define the scope of the invention and that methods and structures within the scope of these claims and their equivalents be covered thereby.

Claims

1. A bioreactor system comprising:

(a) a first module comprising a plurality of bioprocess controls; and

(b) a second module comprising a system for electrical stimulation and electrical recording, wherein the first module is isolated from electrical stimulation.

2. The bioreactor system of claim 1, wherein each electrical component is grounded.

3. The bioreactor system of claim 1 or 2, wherein the electrical component comprises at least the pump, for the addition and removal of liquids, and pressure control.

4. The bioreactor system of any one of claims 1-3, wherein the first module is physically isolated from the second module.

5. The bioreactor system of claim 1-4, wherein the first module comprises a first chamber and the second module comprises a second chamber.

6. The bioreactor system of claim 1-5, wherein the first module comprises an incubator and the second module comprises a bioreactor placed within the incubator.

7. The bioreactor system of claim 1-6, wherein the second module comprises a cell culture or a tissue culture plate.

8. The bioreactor system of any one of claims 1-7, wherein the plurality of bioprocess controls comprises at least one of a pH control, a dissolved oxygen control, a temperature control, a pressure control, and control of gas, liquid, and solid components added and removed from each reactor.

9. The bioreactor system of any one of claims 1-8, further comprising a system for additive manufacturing, subtractive manufacturing, or a combination thereof.

10. The method of claim 9, wherein additive manufacturing comprises bioprinting and injection molding of biomaterials.

11. The bioreactor system of any one of claims 9 or 10, wherein the system for additive and subtractive manufacturing comprises a scaffold material and a plurality of cells.

12. The bioreactor system of claim 11, wherein the scaffold material is biocompatible.

13. The bioreactor system of claim 11 or 12, wherein the plurality of cells comprise a neural progenitor cell, a stem cell, a primary tissue cell, a differentiated neuron, an astrocyte, an oligodendrocyte, a t-cell, a vascular cell, or a combination thereof.

14. The bioreactor system of any one of claims 1-13, wherein the second module comprises a system assembly, growth, and control of tissues by additive or subtractive manufacturing.

15. The bioreactor system of any one of claims 1-14, wherein the second module comprises an electrode array.

16. The bioreactor system of claim 15, comprising both a three-dimensional electrode or microelectrode array for recording and stimulation of engineered tissues.

17. The bioreactor system of any one of claims 15-16, comprising both a surface and three-dimensional electrode or microelectrode array for recording and stimulation of engineered tissues.

18. The bioreactor system of any one of claims 15-17, wherein each electrode is configured to stimulate and record electronic cellular messaging.

19. The bioreactor system of any one of claims 1-18, further comprising at least 1 port for flow of liquid into the bioreactor.

20. The bioreactor system of claim 19, wherein the liquid comprises growth factors, nutrients, metabolites, stabilizers, pH indicators and controllers, living and non-living components .

21. The bioreactor system of any one of claims 1-20, wherein the system for electrical stimulation comprises at least one electrical component comprising at least one of a pump, a microcontroller, or controller, a probe, and a microelectrode array.

22. The bioreactor system of any one of claims 1-21, further comprising: a) a module comprising a data processing and control center; b) a module comprising a biological to digital decoder; c) a module comprising a neural signal preprocessing block; d) a module comprising a digital-to-biological encoder; e) a module comprising AI/ML assisted post-processing; f) a module for additional ML training and offloading; g) a module for local data storage; h) a module for cloud data storage; i) a module for CMOS-based biocomputation; or j) a combination thereof.

23. A method of using the bioreactor system of any one of claims 1-22, wherein additive manufacturing is used to directly biomanufacture tissues into said reactor.

- no -

24. The method of claim 23, wherein the bioreactor system allows stimulation and recording of engineered tissue electrophysiochemistry signals through isolation and reduction of background electrical noise.

25. The method of claim 23 or 24, wherein engineered tissue viability is maintained by way of active perfusion, or passive perfusion in which avascular engineered tissues of diameter greater than 1000 microns, are maintained through the use of porous biomaterials.

26. The method of claim 25, wherein engineered tissue viability is maintained by way of active perfusion, or passive perfusion in which avascular engineered tissues of diameter greater than 1000 microns, are maintained through the use of tissue-free spaces, such as avascular channels for media transport, by way of additive or subtractive manufacturing perfusion.

27. The method of any one of claims 23-26, wherein one or more digital modules serve as a digital twin to a plurality of biocomputation parameters, allowing for the biological information to be stored in a non-biological substrate.

28. The method of claim 27, wherein the digital twin is used to train novel biological components in a biocomputation context.

29. The method of any one of claims 23-28, wherein model features are directly transferred from one biological module to another without routing information back to through digital components.

30. The method of any one of claims 23-29, wherein the biocomputing system comprises a neuromorphic platform that integrates living biological components with reconfigurable digital hardware.

31. The method of any one of claims 23-30, wherein an automatic conversion layer translates high-level, python-like code to perform neuromorphic tasks that comprise: a) in-silico functionality; b) in-vitro functionality; c) in-vivo functionality; d) ASIC wavelet transforms; e) dynamic neural signal ingestion; f) Al post-processing; g) biological-to-digital (ADC) conversion; h) digital-to-biological (DAC) conversion; i) Al preprocessing; or

- I l l - j) a combination thereof.

32. The method of any one of claims 23-31, wherein a module stores biological data including spike trains and synaptic weights, and runs on-chip ML models to extract neural data features and perform predictive analysis while reducing latency and energy consumption.

33. The method of any one of claims 23-32, wherein animal or human neural tissues are decellularized, recellularized with new cells, and then used for biocomputation.

34. The method of any one of claims 23-33, wherein a module performs real-time task redistribution in neural tissue computation, wherein workload, health, and cohesion metrics are continuously assessed to dynamically reallocate tasks, optimizing resource utilization while maintaining tissue stability.

35. The method of any one of claims 23-34, wherein neural computation task migration is performed based on spike features, gamma levels, complexity scores, and connectivity measures.

36. The method of any one of claims 23-35, wherein a Task Table is made wherein active sub-tasks are tracked with cluster assignment, priority, progress state, and/or workload metrics.

37. The method of any one of claims 23-36, wherein training and inference can simultaneously occur across silicon and biological substrates.

38. The method of any one of claims 23-37, wherein a module automatically converts between alphanumeric formats and neural signals in one or two directions.

39. The method of any one of claims 23-38, wherein a module trains biological components to execute basic scripts by associating specific input patterns with desired output responses.

40. The method of any one of claims 23-39, wherein digital data is automatically converted into neural spike trains and other signals using encoding schemes compatible with neural processing.

41. The method of any one of claims 23-40, wherein neural spike trains and other neural signals are automatically converted into digital data.

42. The method of any one of claims 23-41, wherein data is sent to biological components such as neural cells or tissue(s) in a compressed format.

43. The method of any one of claims 23-42, wherein data is decompressed in biological components such as neural cells or tissue(s).

44. The method of any one of claims 23-43, wherein a module performs matrix operations.

45. The method of any one of claims 23-44, wherein a module performs matrix operations within biological components such as neural cells or tissue(s).

46. The method of any one of claims 23-45, wherein a biocomputer is used for autonomous biological/silicon drone applications.

47. The method of any one of claims 23-46, wherein the biocomputing system provides initial bias to prevent inadvertent stimulation or sensing or provide gating threshold function.

48. The method of any one of claims 23-47, wherein any electrode can be connected directly to any other to force specific neural paths in thinking.

49. The method of any one of claims 23-48, wherein neurons can control a switch to provide new self-generated hybrid thinking routes.

50. The method of any one of claims 23-49, wherein neurons can control functions outside of their specific well, outside a given module, and/or outside the system.

51. The method of any one of claims 23-50, wherein delays in action potentials is leveraged to achieve specific thinking mechanisms - circuit equivalent of delay line.

52. The method of any one of claims 23-51, wherein electronic circuits provide digital logic type functions in a biocomputation context.

53. The method of any one of claims 23-52, wherein a through silicon via is used to guide biological component growth, such as dendrite or neural growth paths.

54. The method of any one of claims 23-53, wherein electrode arrays and biological components such as tissue cultures are alternatively stacked.

55. The method of any one of claims 23-54, wherein at least one through wafer vias (TSV) and electrodes are built in vertical stacks to provide compact connections.

56. The method of any one of claims 23-55, wherein biological sensing is accomplished through developed biological systems for sight, smell, touch sensing, or a combination thereof.

57. The method of any one of claims 23-56, wherein biological components directly connect to digital camera chip.

58. The method of any one of claims 23-57, wherein biological components directly connect to an inertial sensing system(s).

59. A method of manufacturing a three-dimensional tissue, comprising manufacturing the three-dimensional tissue directly into a bioreactor; wherein the tissue is greater than 1000 gm³.

60. The method of claim 59, wherein the three-dimensional tissue is integrated with a three-dimensional multi el ectrode array.

61. The method of claim 60, wherein the bioreactor comprises a first module comprising bioprocess controls and a second module comprising, the three-dimensional electrode array, wherein the first module is isolated from electrical stimulation.

62. The method of any one of claims 59-61, wherein the three-dimensional tissue comprises a three-dimensional tissue for use in regenerative medicine.

63. The method of claim 62, wherein the three-dimensional tissue comprises neural tissue or muscle tissue.

64. The method of claim 62, wherein the three-dimensional tissue is for use in a non- clinical trial.

65. The method of any one of claims 59-61, wherein the three-dimensional tissue is for use in cellular agriculture.

66. The method of any one of claims 59-61, wherein the three-dimensional tissue is for use in biocomputing.

67. A method of analyzing a tissue comprising a three-dimensional electrode array, comprising:

(a) additive and subtractive manufacture of the tissue enmeshed with three-dimensional electrode array in a bioreactor, wherein the first module comprises bioprocess controls and the second module comprises the three-dimensional electrode array, wherein the first module is isolated from electrical stimulation;

(b) sending training signals to the tissue in the second module; and

(c) receiving signals from the three-dimensional electrode array.

68. The method of claim 67, wherein the engineered tissue and the three-dimensional electrode array comprise a three-dimensional biocomputing system.

69. The method of claim 68, further comprising a brain machine interface.

70. The method of any one of claims 67-69, wherein the engineered tissue is greater than 1000 gm³.

71. The method of any one of claims 59-70, wherein manufacturing comprises at least one method of additive manufacturing.

72. The method of any one of claims 59-71, wherein the tissue is vascularized.

73. The method of any one of claims 59-72, wherein the tissue comprises at least one cell type, at least two cell types, or at least three cell types.

74. The method of claim 72, wherein at least one cell type comprises a neural progenitor cell, a stem cell, a primary tissue cell, a differentiated neuron, an astrocyte, an oligodendrocyte, a T cell, or a vascular cell.

75. The method of any one of claims 60-74, wherein the three-dimensional microelectrode array is embedded into engineered tissues.

76. The method of any one of claims 60-75, wherein at least one electrode in the three- dimensional electrode array can stimulate and record electronic cellular messaging.

77. The method of any one of claims 60-76, wherein a plurality of bioprocess controls regulate a plurality of bioprocess parameters simultaneously with at least one electrode in the three dimensional electrode array stimulating and recording electronic cellular messaging.

78. The method of any one of claims 60-77, wherein the three-dimensional microelectrode array comprises read and write capabilities.

79. The method of any one of claims 60-78, wherein the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues.

80. The method of any one of claims 60-79, wherein the three-dimensional microelectrode array with read and write capabilities is embedded into engineered tissues supported by engineered vascular networks.

81. The method of any one of claims 60-80, wherein the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues, supported by vascular networks.

82. The method of any one of claims 60-81, wherein the three-dimensional microelectrode array with read and write capabilities is embedded into engineered tissues supported by engineered avascular, active perfusion networks.

83. The method of any one of claims 60-82, wherein the three-dimensional microelectrode array and the surface grid microelectrode array with read and write capabilities are embedded into engineered tissues, supported by avascular passive perfusion networks through the use of negative space in engineered tissues.

84. The method of any one of claims 60-83, comprising manufacturing of the engineered tissue comprising a plurality of cells with the three-dimensional electrode array in a bioreactor system, with a three-dimensional electrode array capable of: a) sending relevant electrophysiology signals to the tissue; and b) receiving signals from the engineered tissue cells.

85. The method of claim 84, further comprising receiving at least one signal from a plurality of cells in the engineered tissue, wherein the signal is received by the three- dimensional electrode array.

86. The method of claim 84, further comprising receiving at least one signal from a plurality of cells in the engineered tissue, wherein the signal is received by an external receiver.

87. The method of any one of claims 60-86, further comprising training of cells within an engineered tissue using microelectrode pulses at physiologically relevant ranges embedded within engineered tissues.

88. The method of any one of claims 60-87, further comprising training of cells within an engineered tissue using microelectrode pulses at physiologically relevant ranges embedded within the tissues through the use of the three-dimensional arrangement of microelectrode arrays.

89. The method of any one of claims 60-88, wherein the three-dimensional electrode array emits a three-dimensional arrangement of electrical pulses to map networks of engineered tissues.

90. The method of any one of claims 60-89, comprising reading data communication between a plurality of bioprocess controllers and a plurality of electric cell firings by multi-electrode arrays in a biomanufactured three-dimensional environment.

91. The method of any one of claims 60-90, wherein the use of training and test phases with three-dimensional multi-electrode arrays and which train and evaluate the quality of a neural network.

92. The method of any one of claims 60-91, further comprising controlling a bioprocess with a plurality of neural spikes.

93. The method of any one of claims 60-91, further comprising controlling a bioprocess with a neural spike pattern.

94. The method of any one of claims 60-93, further comprising storing information via in- silico and in vitro neural networks within a manufactured biological three- dimensional environment exceeding >1000 pm cross-section.

95. The method of any one of claims 60-94, further comprising writing, reading, and interpreting cell electrical firings with an artificial neural network within a manufactured biological three-dimensional environment exceeding >1000 pm crosssection.

96. The method of claim 60-95, wherein multiple organoids or tissues are used together in biocomputation and information storage.

97. The method of any one of claims 60-96, wherein the pulse train, strength, frequency, pattern, duration, waveform, amplitude, shape, and physical location are used for read, write, and storage of information in a biocomputation environment.

98. The method of any one of claims 60-97, comprising correlating gene expression to neuronal activity over time and three-dimensional space to indicate learning patterns and mechanisms.

99. The method of any one of claims 60-98, further comprising: a) the use of electrical, chemical, and/or physical stimulation to engineered tissue interior for use in biocomputation; b) the use of living neurons as perceptrons in an artificial neural network; c) the use of artificial neurons and biological neurons as enmeshed nodes in a neural network; or d) a combination thereof.

100. The method of any one of claims 60-99, wherein biological neuron firings from the three-dimensional multi el ectrode embedded in engineered tissue populates a data table which is then interpreted by artificial intelligence.

101. The method of any one of claims 60-100, wherein biological neuron firings from 3D multi el ectrode embedded in engineered tissue is directly interpreted by artificial intelligence in real-time.

102. A method for converting annotated neural recordings into biocomputational inputs, the method comprising: a) capturing an annotated neuroscience dataset during subject cognitive activities using brain recording devices; b) processing the annotated neuroscience dataset to extract neural data comprise a plurality of neural spike trains; and c) introducing the plurality of neural spike trains into an engineered tissue via a brain-machine interface.

103. The method of claim 102, wherein the annotated neuroscience dataset are derived from in vivo measurements.

104. The method of claim 102 or 103, wherein the annotated neuroscience dataset comprises emergent data generated via biocomputation.

105. The method of any one of claims 102-104, further comprising a tokenization module configured to convert the neural data into discrete tokens representing distinct neural activity patterns, raw data, or concepts.

106. The method of claim 105, wherein the stimulation inputs applied to the biocomputing tissue comprise a combination of tokenized waveforms and/or spike trains and simple spike signals that encode values such as numbers or letters.

107. The method of any one of claims 102-106, further comprising processing nodes operable in both a biological tissue layer and a digital computing layer to allow dynamic conversion, transmission, and processing of tokenized neural data.

108. The method of any one of claims 102-107, wherein the annotated neural data is further processed to generate emergent properties that can be used to enhance performance of the biocomputation network.

109. The method of claim 102-108, further comprising an encoding module that facilitates digital-to-biological conversion by converting biological data such as neural waveform signals into ASCII representations, binary representations, or a combination thereof.

110. The method of claim 102, wherein data are treated as multi-dimensional vectors that serve as embeddings for conceptual information.

111. The method of claim 110, wherein the biological data is further processed to generate emergent properties that can be used to enhance biocomputational performance.

112. A method for encoding and/or dual encoding of neural data, comprising: a) tokenizing neural data comprising a plurality of neural spike trains derived from annotated neuroscience datasets into discrete symbols; and/or b) forming a dual stimulation scheme in which both tokenized waveform representations and simple numerical spike signals are delivered to biocomputational tissue.

113. A method for processing and manipulating conceptual information, the method comprising: a) converting neural spike train-derived waveforms into multi-dimensional vectors; b) tokenizing said vectors to form a plurality of embeddings; and c) combining the plurality of embeddings to deliver an output.

114. A method for training and inference in a biocomputation network comprising biocomputational tissues, comprising the steps of: a) preserving the temporal and quantitative attributes of a plurality of original annotated neural recordings during conversion into a plurality of spike trains; b) delivering the plurality of spike trains to the biocomputational tissues; and c) processing the plurality of spike trains to enable language-related and conceptbased tasks in a manner inspired by biological neural networks.

115. A method for comprehensive encoding and decoding of neural information, comprising integrating layered conversion techniques that translate biological neural signals into digital representations, forming multi-dimensional embedding vectors that preserve semantic integrity.

116. One or more non-transitory computer-readable media comprising computerexecutable instructions that, when executed by at least one processor, cause the at least one processor to perform the method of any one of claims 60-115.