WO2025054548A1 - Systems and methods for microbiome analysis - Google Patents

Abstract

A treatment method includes collecting biological samples from an oral cavity of a subject to provide a biome dataset. A risk of oral disease may be identified based on a risk score. The risk score may be generated from the relative abundance data of bacteria in the biome dataset. An AI model that traverses a weighted tree data structure may be used to analyze the relative abundance of bacteria in the biome dataset to provide the risk score. A plan for an antibiotic regimen may be generated for the subject at the first risk of disease.

Description

SYSTEMS AND METHODS FOR MICROBIOME ANALYSIS

CLAIM TO PRIORITY

[0001] This application claims the benefit of the following provisional application, which is hereby incorporated by reference in its entirety: U.S. Serial No. 63/581,473, filed September 8, 2023 (IDNT-0001-P01).

BACKGROUND

[0002] Traditional diagnostic methods for oral diseases often rely on clinical observations. While these approaches are valuable, they may not provide early detection of disease onset or offer insights into the underlying factors contributing to the disease process. As a result, there is a growing interest in developing more sophisticated diagnostic tools that can leverage the information contained within the oral microbiome.

SUMMARY

[0003] In some aspects, the techniques described herein relate to a post-dental implant treatment method including: collecting biological samples from an oral cavity of a subject to provide a biome dataset; identifying the subject as at a first risk of oral disease based on a risk score that is generated from relative abundance data of bacteria in the biome dataset by an Al model that traverses a weighted tree data structure to analyze the relative abundance of bacteria in the biome dataset to provide the risk score; and generating a plan for an antibiotic regimen for the subject at the first risk of oral disease after dental implant surgery, wherein the antibiotic regimen includes a plurality of antibiotics selected based on the relative abundance of the bacteria and antibiotic resistance of the bacteria.

[0004] In some aspects, the techniques described herein relate to a method, wherein the Al model is a random forest model or a decision tree model.

[0005] In some aspects, the techniques described herein relate to a method, wherein the oral disease includes at least one of gingivitis, periodontitis, peri-implant mucositis, or peri-implantitis.

[0006] In some aspects, the techniques described herein relate to a method, further including processing the biological samples to provide the biome dataset by: extracting and purifying DNA from the biological samples; performing DNA sequencing on the purified DNA; and analyzing the DNA sequencing to generate the biome dataset.

[0007] In some aspects, the techniques described herein relate to a method, wherein the relative abundance is at a family level.

[0008] In some aspects, the techniques described herein relate to a method, wherein the relative abundance is at a genus level. [0009] In some aspects, the techniques described herein relate to a method, wherein the first risk of oral disease is above a threshold risk score value.

[0010] In some aspects, the techniques described herein relate to a method, further including administering the plan for the antibiotic regimen to the subject.

[0011] In some aspects, the techniques described herein relate to a method, further including: collecting second biological samples from the oral cavity of the subject to provide a second biome dataset during the antibiotic regimen; identifying the subject as at a second risk of oral disease based on a risk score that is generated from relative abundance data in the second biome dataset by the Al model that traverses the weighted tree data structure to analyze the relative abundance of bacteria in the second biome dataset to provide the risk score; and modifying the plan for the antibiotic regimen. [0012] In some aspects, the techniques described herein relate to a method, wherein modifying the plan for the antibiotic regimen includes pausing the antibiotic regimen.

[0013] In some aspects, the techniques described herein relate to a method, wherein the second risk of oral disease is below a threshold risk score value.

[0014] In some aspects, the techniques described herein relate to a method, further including: collecting third biological samples from the oral cavity of the subject to provide a third biome dataset during the antibiotic regimen; identifying the subject as at a third risk of oral disease based on the risk score that is generated from relative abundance data in the third biome dataset by the Al model that traverses the weighted tree data structure to analyze the relative abundance of bacteria in the third biome dataset to provide the risk score; and modifying the plan for the antibiotic regimen and implementing a different treatment.

[0015] In some aspects, the techniques described herein relate to a method, wherein the third risk of oral disease is above a second threshold risk score value.

[0016] In some aspects, the techniques described herein relate to a method, wherein the different treatment includes at least one treatment selected from: saliva increasing chewing gum, xylitol chewing gum, surgical intervention, gum grafting, implant removal, partial implant removal, chemical debridement, mechanical debridement, acid etchant cleaning, custom tray delivery of antimicrobial or peroxide medicaments, implant surface polishing, hard tissue grafting, soft tissue grafting, surgical soft tissue resection, or gingival pocket irrigation or disinfection.

[0017] In some aspects, the techniques described herein relate to a method, wherein the relative abundance of bacteria includes a relative abundance measure of early bacteria, bridge bacteria, and late pathogenic bacteria.

[0018] In some aspects, the techniques described herein relate to a method, further including: determining a peri-implant disease-associated taxa percentage from the relative abundance of bacteria and wherein the Al model is trained to provide a risk score based on an input that includes the peri-implant disease-associated taxa percentage.

[0019] In some aspects, the techniques described herein relate to a method, further including: determining an anaerobe score from the relative abundance of bacteria and wherein the Al model is trained to provide a risk score based on an input that includes the anaerobe score.

[0020] In some aspects, the techniques described herein relate to a method, further including: determining a gram stain profile from the relative abundance of bacteria and wherein the Al model is trained to provide a risk score based on an input that includes the gram stain profile.

[0021] In some aspects, the techniques described herein relate to a method, further including: determining an alpha diversity from the relative abundance of bacteria and wherein the Al model is trained to provide a risk score based on an input that includes the alpha diversity.

[0022] In some aspects, the techniques described herein relate to a treatment method including: collecting biological samples from an oral cavity of a subject to provide a biome dataset; identifying the subject as having a first risk score for a disease based on a risk score that is generated from relative abundance data in the biome dataset by a trained Al model that predicts scores based on training examples; and generating a treatment plan for the subject having the first risk score for disease, wherein the treatment plan is based on the first risk score and the relative abundance data. [0023] In some aspects, the techniques described herein relate to a method, wherein the Al model is a random forest model or a neural network model.

[0024] In some aspects, the techniques described herein relate to a method, further including administering the treatment plan to the subject.

[0025] In some aspects, the techniques described herein relate to a method, wherein the disease includes at least one of an oral disease, cancer, cognitive decline, rheumatoid arthritis, or Alzheimer's.

[0026] In some aspects, the techniques described herein relate to a method, further including processing the biological samples to provide the biome dataset by: extracting and purifying DNA from the biological samples; performing DNA sequencing on the purified DNA; and analyzing the DNA sequencing to generate the biome dataset.

[0027] In some aspects, the techniques described herein relate to a method, wherein the relative abundance is at a family level.

[0028] In some aspects, the techniques described herein relate to a method, wherein the relative abundance is at a genus level.

[0029] In some aspects, the techniques described herein relate to a system for determining a postdental implant treatment method, including: a collection device configured to be introduced into an oral cavity and to collect biological samples from surfaces; a storage unit, containing a storage medium that preserves the collected sample by stabilizing the sample for subsequent analysis; a processing module configured to extract biological material from the collected sample to isolate bacteria; an analysis unit, configured to analyze the isolated bacteria material to identify and quantify microbial species; a data recording module integrated with the analysis unit, configured to record a relative abundance of the microbial species; a model module configured to correlate the recorded relative abundance with an oral health score; a treatment identification unit configured to identify one or more treatment regimens based on the oral health score and the relative abundance of bacteria; and a user interface configured to provide a platform for users to access the recorded relative abundance and the one or more treatment regimens.

[0030] In some aspects, the techniques described herein relate to a system, wherein the oral health score includes at least one of gingivitis, periodontitis, peri-implant mucositis, or peri-implantitis risk score.

[0031] In some aspects, the techniques described herein relate to a sy stem, wherein the model module includes a random forest model.

[0032] In some aspects, the techniques described herein relate to a system, wherein the analysis unit is further configured to: perform DNA sequencing on the isolated bacteria material; and analyze the DNA sequencing to quantify microbial species.

[0033] In some aspects, the techniques described herein relate to a system, wherein the relative abundance is at a family level.

[0034] In some aspects, the techniques described herein relate to a system, wherein the relative abundance is at a genus level.

[0035] In some aspects, the techniques described herein relate to a system, wherein the model module is further configured to: identify a high risk of oral disease based on the oral health score generated from relative abundance data by an Al model traversing a weighted tree data structure. [0036] In some aspects, the techniques described herein relate to a system, wherein the one or more treatment regimens includes at least one treatment selected from: saliva increasing chewing gum, xylitol chewing gum, surgical intervention, gum grafting, implant removal, partial implant removal, chemical debridement, mechanical debridement, acid etchant cleaning, custom tray delivery of antimicrobial or peroxide medicaments, implant surface polishing, hard tissue grafting, soft tissue grafting, surgical soft tissue resection, or gingival pocket irrigation or disinfection.

[0037] In some aspects, the techniques described herein relate to a system, wherein the relative abundance of the microbial species includes a relative abundance measure of early bacteria, bridge bacteria, and late pathogenic bacteria. [0038] In some aspects, the techniques described herein relate to a system, wherein; the data recording module is further configured to determine a peri-implant disease-associated taxa percentage from the relative abundance of microbial species; and the model module is further configured to provide the oral health score based on an input that includes the peri-implant disease- associated taxa percentage.

[0039] In some aspects, the techniques described herein relate to a system, wherein; the data recording module is further configured to determine an anaerobe score from the relative abundance of microbial species; and the model module is further configured to provide the oral health score based on an input that includes the anaerobe score.

[0040] In some aspects, the techniques described herein relate to a system, wherein; the data recording module is further configured to determine a gram stain profile from the relative abundance of microbial species; and the model module is further configured to provide the oral health score based on an input that includes the gram stain profile.

[0041] In some aspects, the techniques described herein relate to a system, wherein; the data recording module is further configured to determine an alpha diversity from the relative abundance of microbial species; and the model module is further configured to provide the oral health score based on an input that includes the alpha diversity.

[0042] In some aspects, the techniques described herein relate to a method for analysis of oral health, the method including: receiving a sample including bacteria from an oral cavity of a subject; analyzing the sample to generate an analysis including a relative abundance of bacteria in the sample; determining, using a trained machine model, a classification of the relative abundance indicative of an oral health issue; and generating a report based on the classification, wherein the report includes personalized intervention recommendations.

[0043] In some aspects, the techniques described herein relate to a method, wherein the sample includes a saliva sample.

[0044] In some aspects, the techniques described herein relate to a method, wherein the sample includes a plaque sample.

[0045] In some aspects, the techniques described herein relate to a method, wherein the relative abundance is at a family level.

[0046] In some aspects, the techniques described herein relate to a method, wherein the relative abundance is at a genus level.

[0047] In some aspects, the techniques described herein relate to a method, wherein the trained machine model includes a random forest structure. [0048] In some aspects, the techniques described herein relate to a method, wherein the trained machine model includes a decision tree structure.

[0049] In some aspects, the techniques described herein relate to a method, further including: determining a current state of oral health based on the classification; and determining a probability of transitioning into a second stage of oral health based on the classification.

[0050] In some aspects, the techniques described herein relate to a method, further including: applying a qualifier to the analysis prior to the use of the trained machine model, wherein the qualifier is configured to screen for high-risk pathogenic profiles.

[0051] In some aspects, the techniques described herein relate to a method, further including: adjusting the probability of transitioning based on subject demographic data.

[0052] In some aspects, the techniques described herein relate to a method of using a trained machine learning model to identify a treatment based on oral biome composition, including: training, by a computer, a machine model based on input data and a selected training algorithm to generate a trained machine model, wherein the selected training algorithm includes a computation of a loss function; detecting one or more health-adverse biome compositions using the trained machine model; determining that the one or more health-adverse biome compositions are associated with one or more oral diseases; identifying, based on the detected health-adverse biome compositions, a ty pe of treatment for the one or more oral diseases; and generating, based on the biome composition, a regimen for the type of treatment.

[0053] In some aspects, the techniques described herein relate to a method, wherein the regimen includes antibiotic dosing.

[0054] In some aspects, the techniques described herein relate to a method, wherein the regimen includes a type of antibiotic.

[0055] In some aspects, the techniques described herein relate to a method, wherein the biome composition includes a relative abundance of bacteria.

[0056] In some aspects, the techniques described herein relate to a method, wherein the relative abundance is at a family level.

[0057] In some aspects, the techniques described herein relate to a method, wherein the relative abundance is at a genus level.

[0058] In some aspects, the techniques described herein relate to a method, wherein the relative abundance includes a relative abundance measure of early bacteria, bridge bacteria, and late pathogenic bacteria.

[0059] In some aspects, the techniques described herein relate to a method, wherein the trained machine model is a random forest model. [0060] In some aspects, the techniques described herein relate to a method, wherein the trained machine model is a neural network model.

[0061] In some aspects, the techniques described herein relate to a method, wherein training further includes: selecting a random subset of the input data; and generating a tree structure based on the subset of the input data.

[0062] In some aspects, the techniques described herein relate to a method, wherein training further includes: generating a plurality of tree structures based on different random subsets of the input data. [0063] In some aspects, the techniques described herein relate to a method, wherein detecting one or more health-adverse biome compositions using the trained machine model includes at least one of voting or averaging outputs of each tree.

[0064] In some aspects, the techniques described herein relate to a method, wherein the loss function is a mean absolute error including a measure of an average absolute difference between observed and predicted values.

[0065] In some aspects, the techniques described herein relate to a computer-implemented method of training a machine learning model for biome analysis including: collecting a set of biome data for a set of patients from a database; applying one or more transformations to each biome data to create a modified set of biome data; creating a first training set including the modified set of biome data and a set of patient data of the set of patients; training the machine model in a first stage using the first training set; creating a second training set for a second stage of training including the first training set. the modified set of biome data, and biome classification data from the machine learning model after the first stage of training; and training the machine learning model in the second stage using the second training set.

[0066] In some aspects, the techniques described herein relate to a method, wherein training the machine model in the first stage includes training random forest machine learning methods.

[0067] In some aspects, the techniques described herein relate to a method, wherein training the machine model in the second stage includes training decision tree categorization methods.

[0068] In some aspects, the techniques described herein relate to a method, wherein training the machine model in the first stage includes training a neural network model.

[0069] In some aspects, the techniques described herein relate to a method, wherein training the machine model in the second stage includes fine-tuning the neural network model.

[0070] In some aspects, the techniques described herein relate to a method, wherein the machine learning model is trained to identify one or more health-adverse biome compositions using the trained machine model. [0071] In some aspects, the techniques described herein relate to a method, wherein the machine learning model is trained to identify one or more health-adverse biome compositions using the trained machine model.

[0072] In some aspects, the techniques described herein relate to a method, wherein applying one or more transformations includes determining a relative abundance of bacteria at a family level or a genus level.

[0073] In some aspects, the techniques described herein relate to a computer-implemented method of training a machine learning model for predicting progression of an oral disease using biome analysis, the method including: collecting a set of biome data for a set of patients from a database; applying one or more transformations to each biome data to create a modified set of biome data; creating a first training set including the modified set of biome data and a set of patient data of the set of patients; training the machine model in a first stage using the first training set; creating a second training set for a second stage of training including the first training set, the modified set of biome data, and biome classification data from the machine learning model after the first stage of training; and training the machine learning model in the second stage using the second training set.

[0074] In some aspects, the techniques described herein relate to a method for processing biome data and generating personalized user reports, including: receiving, by a computer system, raw biome data collected from oral biological samples; preprocessing the raw biome data to remove artifacts and normalize the data; storing the preprocessed biome data in a centralized database; extracting features from the preprocessed biome data; analyzing the extracted features using one or more machine learning models to generate biome classifications; storing the biome classifications in the centralized database; receiving a request for biome analysis for a subject; retrieving relevant biome data and classifications for the subject from the centralized database; retrieving personal data of the subject, wherein the personal data includes at least one of demographic information, medical history, lifestyle factors, or dental care habits; generating a personalized biome report for the subject based on the retrieved data, classifications, and personal data; and transmitting the personalized biome report to a user device.

[0075] In some aspects, the techniques described herein relate to a method wherein the personalized biome report includes an identification of an oral disease for the subject.

[0076] In some aspects, the techniques described herein relate to a method wherein the personalized biome report includes a treatment plan for an oral disease.

[0077] In some aspects, the techniques described herein relate to a method wherein the request for the biome analysis is received from the user device and the personalized biome report is generated in real time based on the request. [0078] In some aspects, the techniques described herein relate to a system for multi-site biome sampling and analysis, including: a first collection device configured to be introduced into an oral cavity and to collect a first biological sample at a first site of a subject; a second collection device configured to be introduced into the oral cavity and to collect a second biological sample at a second site of the subject; a processing module configured to extract biological material from the collected biological samples to isolate bacteria; an analysis unit, configured to analyze the isolated bacteria material to identify and quantify microbial species at the first site and the second site; a data recording module integrated with the analysis unit, configured to record a relative abundance of the microbial species at the first site and the second site; a model module configured to: correlate, using a first model, the recorded relative abundance at the first site with a first oral health score; correlate, using a second model, the recorded relative abundance at the second site with a second oral health score; analyze, using a third model, the first oral health score and the second oral health score to determine a third oral health score; a treatment identification unit configured to identify one or more treatment regimens based on the third oral health score; and a user interface configured to provide a platform for users to access the one or more treatment regimens.

[0079] In some aspects, the techniques described herein relate to a system, wherein the first model and the second model include a random forest model.

[0080] In some aspects, the techniques described herein relate to a system, wherein the third oral health score includes at least one of gingivitis, periodontitis, peri-implant mucositis, or peri- implantitis risk score.

[0081] In some aspects, the techniques described herein relate to a system, wherein the third model includes a neural network model.

[0082] In some aspects, the techniques described herein relate to a system, wherein the first oral health score includes at least one of gingivitis or periodontitis risk score and the second oral health score includes a score for peri-implant mucositis or peri-implantitis risk.

[0083] In some aspects, the techniques described herein relate to a system, wherein the first model module is further configured to: identify a first oral health score based on a relative abundance data by an Al model traversing a weighted tree data structure.

[0084] In some aspects, the techniques described herein relate to a system, wherein the one or more treatment regimens includes at least one treatment selected from: saliva increasing chewing gum, xylitol chewing gum, surgical intervention, gum grafting, implant removal, partial implant removal, chemical debridement, mechanical debridement, acid etchant cleaning, custom tray delivery of antimicrobial or peroxide medicaments, implant surface polishing, hard tissue grafting, soft tissue grafting, surgical soft tissue resection, or gingival pocket irrigation or disinfection. [0085] In some aspects, the techniques described herein relate to a method for multi-site biome sampling and analysis, including: introducing one or more collection devices into an oral cavity and collecting biological samples from a first site and a second site within the oral cavity of a subject; extracting biological material from the collected samples to isolate bacteria; analyzing the isolated bacteria to identify and quantify microbial species at the first site and the second site; recording a relative abundance of the microbial species at the first site and the second site; correlating the recorded relative abundance at the first site with a first oral health score using a first model; correlating the recorded relative abundance at the second site with a second oral health score using a second model; analyzing the first and second oral health scores using a third model to determine a third oral health score; identifying one or more treatment regimens based on the third oral health score; and providing a platform through a user interface for users to access the identified treatment regimens.

[0086] In some aspects, the techniques described herein relate to a post-dental implant treatment method including: collecting a first set of biological samples, at a first time, from an oral cavity of a subject to provide a first biome dataset; collecting a second set of biological samples, at a second time, from the oral cavity of the subject to provide a second biome dataset; determining a risk level of the subject for oral disease based on a risk score that is generated from changes in a relative abundance of bacteria between the first biome dataset and the second biome dataset by an Al model that traverses a weighted tree data structure to analyze the changes in the relative abundance of bacteria to provide the risk score; and determining a first treatment plan for the subject, wherein the first treatment is based on the changes of the relative abundance of the bacteria, the risk level, and a relative abundance of bacteria in the second biome dataset.

[0087] In some aspects, the techniques described herein relate to a method, wherein administering the first treatment includes: administering an antibiotic regimen to the subject according to the first treatment plan of oral disease after dental implant surgery, wherein the antibiotic regimen includes a plurality of antibiotics selected based on the relative abundance of the bacteria in the second biome dataset and antibiotic resistance of the bacteria.

[0088] In some aspects, the techniques described herein relate to a method, wherein the changes in relative abundance include changes in the relative abundance of early bacteria, bridge bacteria, and late pathogenic bacteria.

[0089] In some aspects, the techniques described herein relate to a method, wherein the second set of biological samples is collected after a second treatment, and wherein the first treatment is different from the second treatment. [0090] In some aspects, the techniques described herein relate to a method, further including: determining adverse effects of the second treatment based on the changes in the relative abundance of bacteria.

[0091] In some aspects, the techniques described herein relate to a method, further including: determining an effectiveness of the second treatment based on the changes in the relative abundance of bacteria.

[0092] In some aspects, the techniques described herein relate to a system for post-dental implant treatment, including: a first module configured to collect a first set of biological samples at a first time from an oral cavity of a subject, and to generate a first biome dataset from these samples; a second module configured to collect a second set of biological samples at a second time from the oral cavity of the subject, and to generate a second biome dataset from these samples; an analysis module including an Al model that traverses a weighted tree data structure to analyze changes in a relative abundance of bacteria between the first biome dataset and the second biome dataset, and to generate a risk score for identifying the subject as being at a first risk level of oral disease; and a treatment module configured to determine a treatment plan for the subject identified as being at the first risk level of oral disease after dental implant surgery, wherein the treatment is determined based on the changes in the relative abundance of bacteria between the first and second biome datasets, as well as the relative abundance of bacteria in the second biome dataset.

[0093] The methods and systems described herein may be implemented in various forms. It should be understood that each method described may have a companion system configured to perform the steps of the method, and each system described may have a companion method comprising the steps performed by the system. Therefore, any reference to a method in this disclosure may be understood to include the companion system for performing that method, and any reference to a system may be understood to include the companion method performed by that system.

BRIEF DESCRIPTION OF THE FIGURES

[0094] The invention and the following detailed description of certain embodiments thereof may be understood by reference to the following figures:

[0095] Fig. 1 depicts one example of the steps of a process.

[0096] Fig. 2 is a graphical depiction of a sample of a generated dataset.

[0097] Fig. 3 depicts a graphical view of the output data.

[0098] Fig. 4 depicts aspects of machine learning.

[0099] Fig. 5 depicts aspects of a structure for machine learning categorization.

[00100] Fig. 6 depicts aspects of a machine learning example structure.

[00101] Fig. 7 depicts aspects of a machine learning example structure. [00102] Fig. 8 depicts aspects of a machine learning example structure.

[00103] Fig. 9 depicts aspects of a machine learning example structure.

[00104] Fig. 10 shows aspects of a process of patient assessment.

[00105] Fig. 11 depicts aspects of sample collection.

[00106] Fig. 12 depicts aspects of saliva pre-screening.

[00107] Fig. 13 depicts aspects of generating libraries with cascading sequencing depth.

[00108] Fig. 14 depicts aspects of processing saliva and plaque combination.

[00109] Fig. 15 depicts aspects of a report scope.

[00110] Fig. 16 depicts aspects of a use case related to peri-implantitis.

[00111] Fig. 17 depicts aspects of a use case of the systems and methods described herein.

[00112] Fig. 18 depicts aspects of another use case of the systems and methods described herein.

[00113] Fig. 19 depicts a flowchart for a DNA extraction and purification lab process flow.

[00114] Fig. 20 depicts a workflow for DNA library preparation and sequencing.

[00115] Fig. 21 depicts a flowchart for a DNA sequencing process.

[00116] Fig. 22 depicts a flowchart of an example method for generating a post-dental implant treatment.

[00117] Fig. 23 depicts a flowchart of an example method for generating a treatment.

[00118] Fig. 24 depicts a flowchart of another example method for generating a treatment.

[00119] Fig. 25 depicts a schematic of an example system for determining a post-dental implant treatment.

[00120] Fig. 26 depicts a flowchart of an example method for using a trained machine learning model to identify a treatment based on oral biome composition.

[00121] Fig. 27 depicts a flowchart of an example method for training a machine learning model for biome analysis.

[00122] Fig. 28 depicts a flowchart of an example method for multi-site biome sampling and analysis.

[00123] Fig. 29 depicts a schematic of an example system for multi-site biome sampling and analysis.

[00124] Fig. 30 depicts a flowchart of an example method for diagnosis using characteristics of biome progression.

[00125] Fig. 31 depicts a flowchart of an example method for processing biome data and generating personalized user reports.

[00126] Fig. 32 depicts one example of a part of an output of a generated report that may be provided to a user. [00127] Fig. 33 depicts elements of an example user interface for a risk assessment which may be part of a report.

[00128] Fig. 34 depicts elements of an example user interface for intervention guidelines that may be part of a report.

[00129] Fig. 35 depicts a table showing the effectiveness of various antibiotics against different bacterial species that may be part of a report.

DETAILED DESCRIPTION

[00130] Oral microbiome refers to the community of microorganisms, including bacteria, fungi, viruses, and other microbes, that inhabit the oral cavity or mouth. This community’ is immensely diverse, and the oral microbiome plays an important part in oral health and can influence overall health. The community of microbes that reside within the oral cavity can directly impact oral and/or systemic health. Some microorganisms of the microbiome (for example, some bacteria) play a role in the early stages of tooth decay and gum diseases like gingivitis and periodontitis and have been linked to systemic health conditions like cardiovascular disease, rheumatoid arthritis, and many others.

[00131] Analyzing the composition, changes in the composition, and or behavior of these microorganisms can provide insights into oral health and systemic conditions. In one example, a microbiome may be analyzed to produce relative abundance signatures of microorganisms. Relative abundance signatures can be associated with oral and/or systemic health outcomes. The implication of oral microbes in health is often complex and network-driven, but increasingly discrete relative abundance signatures and associations have been identified with direct health conditions and risk states.

[00132] Publications are increasingly demonstrating associations of oral microbial signatures with disease and risk states. However, no method or approach currently exists to aggregate these isolated findings, assess consistencies/overlap, learn from, and refine findings, and ultimately support clinical evaluations.

[00133] Systems and methods described herein may be used to aggregate disparate oral microbiome relative abundance disease/risk state association data, apply machine learning and Bayesian statistical methods to optimize signal selection for sample disease/risk categorization, and then apply an algorithm to individual oral microbiome samples to inform patient disease-state classification and treatment recommendations. For instance, publicly available oral microbiome relative abundance datasets may be utilized to identity' significant disease-state signals and apply Bayesian and machine learning methods to classify patient peri-implantitis risk based on the relative abundance of key microbial genera. Systems and methods described may be used to provide a highly predictive clinical decision support tool to assist dentists and other healthcare providers in determining the personalized risk of peri-implantitis infection.

[00134] Systems and methods described herein provide for determining a patient-specific diagnosis or risk score of a dental or other medical condition (such as a physical or mental condition) based on disease-associated relative abundance signatures observed in patient-specific oral microbiome samples. The systems and methods described herein collect oral microbiome relative abundance data from a plurality of patients from public and private sources, normalize metadata categorizations, and create a training database from which risk-associated signatures are derived utilizing classical and Bayesian statistical approaches and machine learning. Patient-specific disease state/risk is assessed by applying the relative abundance thresholds determined by the machine learning environment, and the individually determined risk score is returned to the patient and/or their clinician to inform treatment and clinical decision-making. The supervised Bayesian/machine learning model is updated using individual patient relative abundance data and clinical outcomes.

[00135] In one example, a test may be applied to determine risk scores that reflect the current disease-state risk and/or the risk of progressing to higher risk profiles. The test may include collecting oral microbiome samples (e.g., swabs of implant plaque) from patients, and processing the sample into genomic libraries to perform taxonomic and relative abundance characterization (e g., 16S). The output of this characterization may then be processed through the categorization algorithm to produce a personalized risk score(s). The risk assessment may be delivered to the patient, their clinician, or both to inform clinical decision-making and offer suggestions/recommendations for intervention. In some implementations, risk scores may include a risk of progression between disease states (i.e., progression from healthy to mucositis). In some implementations, risk scores may also or instead include a risk of progression within a disease state (i.e., progression of mucositis state). In embodiments, the determination of risk assessment and/or risk score may include one or more traditional analyses, Bayesian clustering, network analysis, and the like.

[00136] In embodiments, methods and systems may include one or more steps for collecting data, training a model, collecting patient samples, generating patient assessment, and/or generating patient guidance/recommendations. Fig. 1 depicts one example of the steps of a process. In one example, a process may include a database build and cleanup step 102. a signal identification step 104, a machine learning step 106, a patient assessment step 108, and a personalized intervention and guidance step 110. It will be obvious to those skilled in the art that many changes and modifications may be made thereunto without departing from the spirit and scope of the present disclosure. Those skilled in the art will appreciate that the steps and operations may be combined, divided, re-ordered, added, distributed, or removed in a manner consistent with the disclosure herein. [00137] In embodiments, the database build and cleanup step 102 may include aggregating relative abundance of oral microbiome data from public and private sources. A supervised learning environment may be created and used to collect and aggregate samples with known disease state classifications, oral microbiome taxonomic abundance, and other demographics/bioindicators. Data may be aggregated by extracting data from publications (i.e., using natural language processing), medical databases, and any other suitable data source. In some implementations, data may be aggregated automatically or semi-automatically by trained models. Models, such as language models, may be trained to extract relevant data. In one example, trained extraction models may be trained using labeled examples of extracted data from articles that were extracted by an expert. Aggregated data may be normalized and processed. For example, normalization may include sample segmentation and addressing overlapping metadata. In some cases, aggregated data may not include relative abundance data and additional calculations may be performed to derive relative abundance data. Other normalizations may include adjusting null/zero to very small values (e.g., less than 1% or 0.1%).

[00138] In embodiments, database build and cleanup step 102 may include linking clinical diagnoses/measurements to sample data and/or transformation of data to magnify signals. [00139] Data elements of the aggregated data may be categorized. In one example, categorization may include categorization and analysis based on signals. Signals may include disease state association distributions. In another example, categorization may include categorization and analysis based on qualifiers. Qualifiers may include signal thresholds indicative of risk regardless of disease state categorization and may be used to “fast-track” samples with high-risk pathogens.

[00140] Examples of signals and qualifiers may include microbial relative abundance. Microbial relative abundance may be identified on multiple taxonomic levels and may include bacteria only, virus only, fungi only, phage only, and/or any combination thereof. Another example of signals and qualifiers may include microbial community diversity. Microbial community diversity may be identified on multiple taxonomic levels and may include Shannon index and Simpson index. Other examples of signals and qualifiers include host DNA, genetic markers, proteome, salivary biomarker/indicator concentration (e.g., cortisol), oral microbiome transcriptome/metabolome, host transcriptome/metabolome, sample pH, salivary viscosity, hydration level, clinical diagnoses/ assessment (e.g., for peri-implantitis, pocket depth, bleeding, etc.), and/or any combination thereof or signals and qualifiers herein.

[00141] In another example, categorization may include categorization and analysis based on amplifiers. Amplifiers may include indicators known to increase the risk of disease state progression (e.g., smoking). Amplifiers may include medical screening identifiers (e.g., diabetes, medication. periodontitis level, caries risk/history), smoking habits, longitudinal trends/shifts, demographics/metadata (e.g., age/gender, etc., tooth location, number of implants, number of teeth per implant, etc.).

[00142] In embodiments, the database build and cleanup step 102 may generate a dataset (e.g., a database, a list, a trained model, or the like) that includes the potential signals/metrics. Fig. 2 is a graphical depiction of a sample of a generated dataset. The figure shows a graphical representation of bacterial family -level relative abundance distributions by healthy, peri-mucositis, and peri- implantitis disease states.

[00143] In some embodiments, the database build and cleanup step 102 may include microbiome sampling from users along with their health status data. User sampling may include processing a sample to assess relative abundance and may include techniques such as:

[00144] Next Generation Sequencing

[00145] MassSpec Matrix-Assisted Laser Desorption/Ionization Time-Of-Flight Mass Spectrometry (MALDI-TOF MS): This technique is a tool for rapid, accurate, and cost-effective identification of bacterial isolates at the species level in the clinical microbiology lab.

[00146] Microscopic Counts: A sample is examined under a microscope and the number of microbial cells is physically counted. This can either be done directly on the sample, or after the cells have been cultured. A disadvantage of this method is that it is time-consuming and doesn't differentiate well between different species.

[00147] Flow Cytometry: This technique is used to measure physical and chemical characteristics of a population of cells or particles. In microbial ecology, it is used for enumerating and sorting microbes. It also has the abi 1 i ty to distinguish live from dead cells.

[00148] Culturing Methods: In this technique, microbes are cultured in the lab and the colonies that form are counted. This can give a rough idea of the relative abundance of different microbes, but it has a drawback that many microbes cannot be cultured in the lab.

[00149] Quantitative Real-Time PCR (qPCR): This technique is used to amplify and simultaneously quantify a targeted DNA molecule. It can be used to quantify the abundance of a particular species of microbe in a sample, by using primers that are specific to that species.

[00150] Fluorescent In Situ Hybridization (FISH): This method uses fluorescent probes that bind to specific parts of the microbial DNA or RNA, which can then be visualized under a microscope. The signal intensify can give an indication of the relative abundance of different microbes.

[00151] Phospholipid Fatty Acid Analysis (PLFA): This method analyzes the phospholipid profiles of microbial communities. It's a kind of biochemical method used to estimate the biomass and community composition of soil microorganisms. [00152] Biomarker Analysis: Certain molecules, such as lipids, proteins, or small metabolites, can be used as indicators (or biomarkers) of the presence and abundance of specific types of microbes. [00153] Denaturing Gradient Gel Electrophoresis (DGGE) and Temperature Gradient Gel Electrophoresis (TGGE): These techniques are used to separate DNA fragments according to their sequence. This can give an idea of the diversity of microbes in a sample.

[00154] Microarrays: These are used to detect the presence of known microbial species in a sample. Microarrays consist of small DNA or RNA spots attached to a solid surface, which can hybridize with the microbial DNA/RNA in the sample.

[00155] Stable Isotope Probing (SIP): SIP is a technique for linking the identity of microorganisms to their function. Microorganisms are incubated with a substrate labeled with a stable isotope (e.g., 13C). After an incubation period, the isotope becomes incorporated into the DNA, RNA, proteins, or metabolites of the microbes that utilized the substrate. These labeled molecules can then be separated from the unlabeled ones and identified by methods like DNA sequencing or mass spectrometry.

[00156] Single-Cell Techniques: Techniques such as microfluidics or flow cy tometry coupled with fluorescence-activated cell sorting (FACS) can isolate individual microbial cells. These cells can then be lysed, and their DNA can be amplified and sequenced, or their proteins can be analyzed by mass spectrometry. This provides information not only about the identity of the microbes but also about their individual metabolic activities.

[00157] Imaging Techniques: Techniques such as scanning electron microscopy (SEM), transmission electron microscopy (TEM). or fluorescence microscopy can provide visual information about microbial communities and their spatial arrangements.

[00158] Biolog Phenotype Micro Arrays: This high-throughput system can measure the rate at which a microbe consumes different carbon sources or its sensitivity to different chemicals, providing insights into its metabolic activities and potential ecological roles.

[00159] In embodiments, the step of signal ID/selection 104 may include analysis of the collected data to identify and prioritize signals within the dataset with disease-state or risk associations. Analysis may include classic and/or Bayesian analyses to identify features with significant differences. Analysis may further include cluster analyses to confirm categorization and identify disease transitioning profiles. In implementations, the step may include feature/signal prioritization. Feature/signal prioritization may include traditional and/or Bayesian Regression and/or Analysis of Variance (ANOVA) to assess signal strength and prioritize accordingly. In some implementations, only traditional analysis or only Bayesian analysis can be performed. In some implementations, only regression or only ANOVA analysis can be performed. In implementations, the step may include feature/signal consolidation. Feature/signal consolidation may include analysis across priority signals to identify correlation and combine highly correlated signals. In some implementations, this step maybe excluded or may include k-nearest neighbor analysis. In implementations, the step may include feature/signal selection. Feature/signal selection may include applying random forest machine learning to determine the optimal feature selection/ decision tree panel. In some implementations, alternative machine learning models may be used to assess feature importance to inform selection (decision tree, neural network, naive Bayesian, Bayesian Network Analysis, and the like). In implementations, the step may include setting qualifier thresholds. Setting qualifier thresholds may include setting the thresholds after which a given qualifier is activated. In implementations, the step may exclude qualifier thresholds and/or include or exclude Family/Genus/Species/Strain-level qualifier thresholds. In implementations, the step may include cluster/segment analysis. Cluster/segment analysis may include confirming selected features organized into expected clusters and/or identifying disease-state transitional clusters. In implementations, the step cluster confirmation/identification may be excluded. In some implementations, cluster validation may include one-on-one comparison (e.g., healthy (H) v peri-implantitis (PI); healthy (H) v peri-implant mucositis (PM); peri-implantitis (PI) v peri-implant mucositis (PM)) and/or determining subclusters based on longitudinal sampling accompanied by clinical assessment/indicators/biomarkers. In some implementations, signal selection may occur based on one-on-one analyses (e g., H v PI, H v PM, PI v PM), at each taxonomic level individually, across all taxonomic levels together, and/or based on the longitudinal changes/shifts in an individual's characterization over time.

[00160] In embodiments, the step of signal ID/selection 104 may generate a dataset (e.g., a database, a list, a trained model, and the like) that includes a prioritized list of features/parameters to utilize, feature/parameters correlation, and/or amplified signals. Fig. 3 depicts a graphical view of the output data of the step. The figure shows example feature distributions by segment (PI likely at low levels; toss-up at mid-levels: likely healthy at high levels) - this "signal" is layered with multiple others to determine sample categonzation. A ’‘signal” is derived from the likelihood a given metric value falls into a given disease state category based on the distribution of measured learning environment samples. Categorization accuracy is optimized/improved by applying multiple signal layers.

[00161] In embodiments, the step of machine learning 106 may include utilizing machine learning methods to map samples to disease states and subsegments. The step 106 may include applying machine learning methods to categorize disease-states and disease-state transitioning risk and/or applying selection/matching algorithms of microbial signatures to optimal interventions/treatments. In implementations, the step may include disease-state categorization. Disease-state categorization may include determining and/or selecting the optimal decision tree panel to categorize samples into disease-state classifications. In some implementations, disease-state categorization may include the use of decision trees with Random Forest machine learning structures and/or other machine learning categorization methodology. In some implementations, disease-state categorization may include multiple decision trees and using an average result across the decision trees. In some implementations, disease-state categorization may include conducting machine learning using a single population comprised of all (or >2) disease-state/risk categories. In some implementations, disease-state categorization may include conducting machine learning on multiple data subsets of the broader population, each of which is comprised of two unique categorizations (i.e., categorize by aggregating multiple one-on-one differences/signatures).

[00162] In implementations, step 106 may include sub-cluster categorization. Sub-cluster categorization may include determining/selecting the optimal decision tree panel to categorize samples at risk of disease state transitioning. In implementations, sub-cluster categorization may include the use of a decision tree with an alternative machine learning methodology (e.g., Random Forest). In some implementations, subclusters and transitioning profiles may be derived from passive longitudinal monitoring of representative population(s).

[00163] In implementations, step 106 may include categorization adjustments. Categorization adjustments may include optimizing qualifier/adjustor integration into the model/output. In implementations, qualifiers may be applied prior to machine learning categorization to reduce processing demand (i.e., screen the high-risk pathogenic profiles prior to machine learning categorization). In some implementations, qualifiers may not be used. In some implementations, transition risk assessment may be increased by a factor based on the presence/absence of patientspecific demographics (e.g., diabetes status). In implementations, the step may include microbial profiling. Microbial profiling may map dominant microbial actors to targeted/effective treatment options.

[00164] Fig. 4 depicts some aspects of machine learning examples. In embodiments, machine learning models may be trained to generate outputs such a disease state output 402, sub- cluster/transition risk output 404, adjustments outputs 406, and/or microbial profiling outputs 408. [00165] Fig. 5 depicts example aspects of a structure for machine learning categorization. In one example, machine learning categorization may include a decision tree. Fig. 5 depicts an example machine learning structure that may be trained on aggregated data to determine disease state transition risk. The trained structure may be used to determine disease state transition risk by first determining split determinants (labeled with the value "1" in Fig. 5). Splitting determinants may include grouping for a given taxa grouping. The algorithm identifies the optimal threshold at which to split/initially divide samples. Additional layers are then applied using the same methodology, either strengthening or weakening the classification and the result of the splitting is a disease state categorization. The trained structure may further be used to determine tree design and/or selection. Features (in this case the taxa relative abundance) and tree branches may be selected during the signal optimization process and/or refined during algorithm development (labeled with value "2" in Fig. 5). The leaf nodes of the structure (labeled with value "3" in Fig. 5) may indicate disease state transition risk. Disease-state categorizations are supplemented with the risk of transitioning to the next state identified in the signal ID phase.

[00166] Fig. 6 depicts aspects of another machine learning example structure. The example structure is configured for cascading taxonomic screening. In implementations, the structure of Fig. 6 may be used to perform categorization in sequence or in a cascading fashion. The structure may include a first stage for applying a family -level categorization 602. If samples show strong category associations in the first stage, the family categories may be assigned, and processing may stop, while inconclusive samples may proceed to a second stage for categorization/screening on the Genus level 604. Inconclusive samples from the second stage may proceed to the third stage for final categorization 606. In embodiments, the example structure may reduce processing demands and, thus, costs.

[00167] In another example, a machine learning example structure may include aggregation of outputs of multiple machine learning models (i.e., multiple decision trees) for performing categorization. In some implementations, one or more machine learning models (i.e., a plurality of models configured to execute in parallel and/or a model configured to execute iteratively) may be configured to perform categorization at different taxonomic levels (i.e.. Family, Genus, Species) and/or relationships (i.e., Healthy v Infection, Healthy vs Peri-Implant Mucositis, Healthy v Peri- Implantitis, PM v PI). The outputs of the model(s) may be aggregated by weighting each output (i.e., each Family. Genus, Species, etc.) by a factor (i.e., a probability calculated by the machine learning method for each categorization) to generate an overall categorization of samples.

[00168] Fig. 7 depicts aspects of another machine learning example structure. The example structure is configured for multi-level taxonomic validation. In implementations, the structure of Fig. 7 may be used to perform categorization in parallel and may output family category assignment 702, genus category assignment 704, and species category assignment 706. The algorithm may be applied at the Family, Genus, and Species levels and categorize samples in parallel. The structure may further include a layer 708 to adjust the output confidence level or likelihood based on the presence/absence of results consistency between levels and output final categorization 710. The multi-Level parallel characterization of the structure may improve algorithm accuracy and certainty. [00169] Fig. 8 depicts aspects of another machine learning example structure. The example structure is configured for the pre-algorithm qualifier screen. In implementations, the structure may be used to pre-screen data for "qualifier" signals (e.g., samples w/ >40% P.Gingivalis are always high-risk). In implementations, the structure of Fig. 8 may be used to apply any machine learning method/embodiment described to the remaining samples to categorize to generate a final categorization 802. The structure may be used to provide initial screening and generate a categorized subset 804 and may reduce samples requiring deeper or more costly characterization.

[00170] Fig. 9 depicts aspects of another machine learning example structure. The example structure is configured for post-algorithm adjustment. In implementations, the structure of Fig. 9 may be used to perform machine-learning categorization to produce initial categorization 904. The structure may adjust parameters applied to relevant samples and adjustments made to the raw disease state or risk classifications produced by the machine learning algorithm. In embodiments, adjustors 906 may further improve the algorithm accuracy for a final categorization 908 and act as an additional lever for optimization.

[00171] In embodiments, the step of patient assessment 108 may include utilizing machine learning methods to map samples to disease states and subsegments. The step may collect and process the patient's oral microbiome sample, apply machine learning and selection algorithms to prioritize features/signals, and produce reports specific to the patient's risk of disease onset/progression. In implementations, the step may include collecting samples 1002 and creating libraries 1004. Collecting 1002 may include collecting oral microbiome samples (saliva, plaque, subgingival plaque, supragingival plaque, buccal mucosa, any other oral microbiome sample, other specimen from the oral cavity, and/or any combination thereof) either to be collected by a clinician or in an at-home setting and preparing the sample library for 16S sequencing. Sample types may include saliva, plaque, subgingival plaque, supragingival plaque, buccal mucosa, any other oral microbiome sample, specimens from the tongue, other specimens from the oral cavity, and/or any combination thereof may be collected using saliva collection, from a swab, scraping, and the like and the sample may be collected using a disposable device. Sample library prep 1004 protocol/method may include whole genome sequencing (WGS). In implementations, step 1004 may further include performing 16S metagenomic sequencing to characterize the oral microbiome.

[00172] In some embodiments, collecting samples 1002 may include cascading sampling methods. In some implementations, sampling may be performed based on a sampling hierarchy. A sampling hierarchy may be based on one or more of criteria such as the difficulty of taking samples, time to take samples, assistance requirements for taking samples (i.e., can a patient take the sample by themselves or do they need assistance), cost of processing samples, cost of collection the sample, and the like. Sampling and analysis may be performed based on the hierarchy. For example, sample locations with easier access may be sampled and analyzed first. If the first samples indicate a disease state, additional samples and analysis may be performed in harder-to-access locations. In one example, sampling may first include a sampling of supra-gingival plaque (i.e., a swab of the tooth above the gumline) either as a first pass for hard-to-access implants or as a less invasive/more patent friend method to collect a sample. The analysis results of the first pass may be used to inform further testing or clinical interventions.

[00173] In some embodiments, alternative methods, as described herein, may be used to quantify microbial relative abundance. In some implementations, alternative sample characterization may be performed (e.g., metabolite concentration or other methods described with reference to Figs. 6-9). In some implementations, as described herein, samples may be processed at a cascading depth. In implementations, the step may further include processing sequences and aggregate signals. Raw sequences may be processed through 16S bioinformatic pipelines for taxonomic classification, relative abundance, and diversity. In implementations, the step may further include applying algorithms and generating reports 1012, such as applying the decision-tree algorithm selected during machine learning optimization.

[00174] In embodiments, patient samples may be collected by the patient in their home. At-home testing bridges the gap betw een dental visits and allows for even earlier identification of risk states. At-home testing and results may inform increased dental visits or tailor/optimize routine visits to address identified issues. In some implementations, samples may be collected both at home and during clinical visits and may be collected longitudinally. In some implementations, saliva and plaque samples may be both collected/characterized. Plaque and saliva both offer insights into the health of implants, teeth, and other oral health indicators. Plaque offers a direct assessment of the community potentially impacting the mouth, as well as the actors most likely to diffuse into the blood. Saliva provides an overall average view of the mouth and thus can provide an extremely non- invasive method to assess overall oral health status, as well as identity signatures consistent with oral and systemic disease states. The samples can offer insights individually but may also improve accuracy with a combo analysis.

[00175] Fig. 10 shows aspects of a process of patient assessment. As described herein, the process may include sample collection 1002, DNA extraction and library preparation 1004, 16S metagenomic sequencing 1006, bioinformatics taxonomic processing 1008, machine learning algorithm categorization 1010, and report generation 1012.

[00176] Fig. 11 depicts example aspects of sample collection. Sampling location embodiments may include in-office (i.e., at the location of a healthcare provider), at-home, or a combination of the two. At-home testing can be used as a non-invasive method for risk monitoring between check-ups. Inoffice monitoring can be performed during routine clinical workflows.

[00177] Fig. 12 depicts example aspects of saliva pre-screening. Pre-screening may include collecting and characterizing a saliva sample for disease-associated signatures/categorization. The saliva results may be used as a pre-screen for additional, targeted testing of implant plaque such that additional direct site-specific sampling only on patients with saliva characterized as high-risk for infection onset or progression may be performed.

[00178] Fig. 13 depicts example aspects of generating libraries with cascading sequencing depth. Sample libraries may be generated with flexible sequencing scenarios. The generation may include sequencing a subset/aliquot from the library at a low depth 1302 and applying machine learning algorithms. If the categorization output is conclusive, a report may be generated without further processing 1306. If output is inconclusive, a second aliquot may be sequenced at a deeper depth 1304, and the process may be repeated until confident categorization is achieved.

[00179] Fig. 14 depicts example aspects of processing saliva and plaque combination. The processing may include collecting both saliva and plaque samples. In one implementation, the saliva sample may be processed first, and results/signals may be used to inform whether to characterize the plaque sample and if so, inform sequencing parameters (e.g., depth of run - increase/decrease depth based on saliva characterization certainty). In another implementation, both samples may be processed and used to triangulate conclusions/results between the two.

[00180] In embodiments, the step of intervention guidance may include applying models to patient oral microbiome samples and recommending treatment based on sample specifics. The report may provide oral care preventative and reactive strategies based on the unique microbial actors affecting the patient. In implementations, the step may include report generation. A summary' report may be generated for each patient sample summarizing the disease state classification, risk profile, and interventional recommendations/insights. In some implementations, reports may include patientspecific and clinician-specific reports and any single or combinations of categories described herein may be included. The step may include report distribution. The report may be distributed electronically to the patient and their clinician for review using an online portal and/or mobile device or application. The step may include clinical review and care planning. The report may be reviewed with patient to update care and intervention plan. The review may include dental implant procedure scheduling/timing. Adaptive procedure scheduling may be generated based on the microbial riskprofile (high-risk, delay operation pending intervention/improvement). The review may include dental implant procedure precision infection control (tailoring surgical infection control strategy based on the specific microbial profile of site). The review may include analysis/recommendation of systemic vs. targeted antibiotics vs. oral/topical (and selection within each category ), post-operation oral care instruction (e g., use of anti-microbial rinse), post-operation monitoring frequency (tailoring appointment schedule frequency based on patient risk profile). The review may include longitudinal post-operative preventative and reactive (following mucositis/implantitis diagnosis) care selection, oral health care plan variants, visit frequency for mechanical cleanings (mechanical cleaning method selection), and/or antibiotic use and/or selection. The step may further include longitudinal monitoring to perform ongoing sampling with the patient to enable early identification and intervention. In implementations, samples may be collected and characterized at routine check-ups and assessed for population-derived signals as well as individual trending/shifts.

[00181] In embodiments, the methods and system described herein may include relative abundance signal prioritization and triangulation. The process may include the following steps:

[00182] 1) Conduct classical and Bayesian methods to identify, assess, and prioritize significant associations of genera relative abundance to disease or disease-risk states using the “supervised” and “as-is” datasets.

[00183] -Classic regression - ID p/q-value associations between genus relative abundance and disease/risk-states.

[00184] -Random Forest Machine Learning - ID most significant contributors to random forest categorization.

[00185] -Bayesian Network Analysis - ID and assess the implication of positive and negative genuslevel relative abundance correlations.

[00186] 2) Compare/validate significant associations between datasets & analysis methodologies. [00187] -Verify consistency between genera signal/importance between classical regression and random forest machine learning outcomes.

[00188] -Explore/ document any significant discrepancies between analyses to justify inclusion/ exclusion during panel selection.

[00189] 3) Assess feature dependencies/correlation and batch positively correlated genera to improve signal strength.

[00190] -Perform both classic regression and random forest machine learning analyses on combinations of correlated genera and single genera.

[00191] 4) Prioritize single/batch relative abundance panel candidates based on preliminary insights. [00192] - Select high-impact single genus and correlated genera combinations as candidates for selection into the assay panel. [00193] -Select a subset of microbes to quantify relative abundance (i.e., subset of 10 bacteria families). The relative abundance within the family subset may be quantified as a signal with respect to the subset and/or the bacteria overall.

[00194] - Grouping and aggregating microbes based on either their characteristics (e.g. Gram stain, oxygen relationship), or their positive/negative correlation with other features/signals (e.g. rather than using the relative abundance of specific bacterial families, aggregate relative abundance based on taxa characteristics (e.g. relative abundance of facultative anaerobes overall, versus each individual).

[00195] In embodiments, the methods and system described herein may include a process for machine learning and validation. The process may include the following steps:

[00196] 1) Assess combinations of relative abundance genera-level associations utilizing a supervised Bayesian decision tree model.

[00197] - Test multiple single genera/combo genera panels.

[00198] - Assess relative abundance qualifiers or additional criteria.

[00199] -Explore sensitivity etc.

[00200] - Determine Panel Options.

[00201] 2) Validate panel accuracy for disease classification by: Running Markov simulations and/or applying decision tree to unsupervised dataset and confirming accuracy.

[00202] 3) Develop/assess/profiles transitional profiles.

[00203] - Determine risk-scale and thresholds for scenarios where a signature may be high-risk of progressing to disease states etc.

[00204] 4) Validate panel accuracy for transitional and disease classifications.

[00205] - Rerun simulations and unsupervised testing

[00206] 5) Select/Prioritize decision panel utilization and weighting based on accuracy.

[00207] In embodiments, the methods and system described herein may include a process for application of tools in the clinical setting. The process may include the following steps:

[00208] 1) Select/Create 16S primers based on the target panel make-up determined in the prior step. [00209] - Design 16S primer panel optimized for the accurate identification of the key genus and genera combinations producing significant disease risk/state associations.

[00210] 2) Collect Sample.

[00211] - Produce an oral microbiome samples (saliva, plaque, or other specimen from the oral cavity) either to be collected by a clinician or in an at-home setting.

[00212] - Place sample in an approved/validated stabilizing buffer and submit for processing. [00213] 3) Create sample libraries and perform 16S sequencing. [00214] - Prepare sample for 16S sequencing by performing the following processes: DNA extraction/purification; library preparation and amplification; library dilution and sequencing alignment; library quality check and verification.

[00215] - Run 16S sequencing using optimized primers.

[00216] 4) Perform Bioinformatics on raw sequences and quantify the relative abundance of bacteria in the sample at the genus level.

[00217] - Process raw sequences through 16S bioinformatic pipelines for taxonomic classification and relative abundance.

[00218] 5) Apply the categorization algorithm by applying Bayesian decision tree thresholds to the sample using the relative abundance data produced.

[00219] 6) Deliver risk assessment and findings to clinicians (and patients) to support patient interventions and risk-mitigation activities.

[00220] Fig. 15 depicts example aspects of the report scope. A report may include aspects directed to disease state 1502, risk of progression 1504, pathogen profile 1506, and/or intervention recommendations 1508.

[00221] Fig. 16 depicts example aspects of a use case related to peri-implantitis. Peri-implantitis is polymicrobial in nature as no single pathogen is responsible for disease progression or manifestation. Multiple pathogens are known to be associated with the disease, but their presence/abundance alone is insufficient to understand the disease state. Plaque microbial composition evolves with disease progression from healthy to peri-implant mucositis (PM) to peri-implantitis (PI). Both PM/PI are diagnosed based on a clinical inspection of disease-associated indicators (inflammatory signs, bleeding/suppuration on probing, probing depth, presence of bone loss). Plaque microbial composition is not currently assessed to support the diagnosis. PM is reversible, and treatment options include mechanical therapies, oral hygiene instruction, and adjunctive antimicrobial agents. PI is irreversible, and treatment options include PM methods, several surgical treatments, or implant removal. PM treatments can restore the implant to a healthy state. PI treatments are to prevent further damage/loss. Currently, there are limited early identification options. Clinical diagnosis often occurs once the disease has progressed to require more challenging treatment and treatment decisions are blind to the specific microbial actors responsible for disease manifestation.

[00222] The methods and systems described herein provide for numerous benefits in the diagnosis, prevention, and treatment of PM/PI. In one aspect, they enable ID/classification of healthy implants at high risk of developing mucositis. In another aspect, the methods and sy stems described herein enable ID/classification of mucositis patients at high risk of progressing to peri-implantitis. In another aspect, the methods and systems enable at-home non-invasive monitoring of disease state risk/progression via saliva. Dentists can proactively adjust care protocols, interventions, visits/cleaning frequency, etc., before physical symptoms present and thus avoid more complex or costly interventional needs. In another aspect, the methods and systems enable targeted/personalized treatment optimization as they enable visibility into the specific microbes responsible for disease progression and thus enable crafting targeted/specific interventional strategies with a higher likelihood of success. Ongoing use following procedure allows for an intra-patient assessment tailored/optimized based on observed changes to the microbial dynamics, allowing for longitudinal visibility.

[00223] Fig. 17 depicts example aspects of a use case of the systems and methods described herein. The figure shows the stages of implant procedure timing. In pre-implant surface surgeries (stages 1, 2, and 3) methods and systems may use at-home saliva sampling. Sampling may be performed during post-tooth removal healing and during post-implant insertion healing. Samples may be analyzed to determine general infection risk per known pathogens and periodontitis/peri-implantitis known pathogens. The microbiome testing may provide a dentist with currently absent visibility into the infection risk profile of patients driven by oral microbes. The data may allow care providers to perform risk assessment and interventional guidance regarding modifications to at-home oral health guidance, use and targeted selection of antibiotic treatments, and/or inform following procedure scheduling/readiness.

[00224] In post-implant surface surgeries (stages 4, 5. and 6), methods and systems may be used in office plaque sampling. Sampling may be performed prior to temporary crown placement, prior to permanent crown placement, and/or post-procedure check-up. In these stages, there is a targeted risk of patient progression from healthy to peri-implant mucositis and/or peri-implantitis. Sample analysis may provide dentists with currently absent visibility into the risk of implant disease progression and the specific microbes responsible. The data may allow care providers to perform a risk assessment and interventional guidance regarding modifications to at-home oral health guidance, use and targeted selection of antibiotic treatments, and optimize check-up visit frequency for risk-based plaque mgmt.

[00225] In longitudinal care (stage 7), methods and systems may use both plaque and saliva testing in the office and at-home locations. Saliva sampling may be performed between dentist visits. Plaque sampling may be performed at dentist visits. The sampling and testing of the sample may provide data regarding the risk of patient progression from healthy to peri-implant mucositis and peri- implantitis. The data may allow care providers and patients to optimize the combination of remote and in-office longitudinal monitoring and risk assessment. The data may allow interventional guidance, at-home oral health protocol changes, antibiotic use/selection, and/or dentist visit frequency optimization.

[00226] Fig. 18 depicts example aspects of another use case of the systems and methods described herein. In another use case example, methods and systems described herein may be used to detect, prevent, and/or treat periodontitis and caries. The primary cause of periodontitis is poor oral hygiene which leads to the buildup of bacterial plaque. This plaque hardens into tartar, leading to gum inflammation (gingivitis) which, if left untreated, can progress to periodontitis. Symptoms include swollen or puffy gums, bright red, dusky red or purplish gums, gums that feel tender when touched, gums that bleed easily. Certain factors can increase your risk of developing periodontitis, including smoking, diabetes, poor nutrition, stress, certain medications, genetic susceptibility, certain infections and diseases, and hormonal changes in females. Increasing research suggests that periodontitis may be associated with other health conditions such as heart disease, diabetes, and Alzheimer's disease. The theory is that inflammation in the mouth can trigger inflammation in other parts of the body.

[00227] Saliva and/or plaque samples may be collected and characterized during and in between check-ups. The results of analysis may be used to assess patient periodontitis status and risk of progression based on their unique microbial signatures. Characterization can inform intervention selection, check-up frequency, and oral care recommendations to resolve or prevent periodontitis. Benefits of microbiome screening may include earlier at-home identification (routine test patient performs at-home between dental visits, say halfway between, and provides a risk assessment and either directs oral homecare improvements or an extra dental visit); early identification and risk assessment which provides insight into disease onset and progression risk prior to visually observable. Benefits further include improved treatment guidance. Based on the community profile of saliva/plaque/combo, treatment options are made available to support dentist clinical decisionmaking. Further benefits include the ability to perform easier post-diagnosis longitudinal progress monitoring since at-home sampling/testing between dentist visits to monitor the progression of the disease and ensure treatment is working.

[00228] Although the methods and system described herein used examples related to diagnosing and detecting peri-implantitis, periodontitis, and caries, the methods and systems described herein maybe used in other applications. In one example, the methods and systems may be used to determine implant placement candidacy. In some aspects, the analysis of the oral microbiome composition and its changes over time may provide valuable insights into the suitability of a patient for dental implant procedures. The Al models described herein may be trained to identify microbial profiles that are associated with successful implant outcomes, as well as those that may indicate a higher risk of complications. In some cases, the system may analyze the relative abundance of specific bacterial species known to impact implants or contribute to peri-implantitis. This analysis may help clinicians assess the potential risks and benefits of implant placement for individual patients. The system may generate a candidacy score based on the microbial composition, taking into account factors such as the presence of pathogenic bacteria, overall microbial diversity, and the balance between beneficial and harmful microorganisms. In embodiments, systems and methods described herein may be configured to provide treatment plans for improving the oral microbiome prior to implant placement. In some implementations, the system may incorporate longitudinal data analysis to track changes in the oral microbiome over time, allowing for the assessment of a patient's response to preparatory' treatments and their readiness for implant placement.

[00229] In another example, the methods and systems may be used to generate or recommend specific antibiotic regimens based on outcome, recommend local antibiotic application with biofilm disruptor, diagnose or detect Alzheimer’s, rheumatoid arthritis, IBD, cancer (several ty pes), metabolic indicators, cognitive decline, pre-term birth and inflammatory markers. In embodiments, methods and systems may be used to test histamine response, detect allergies, and the like. In embodiments, the methods and systems described herein may be used to analyze microbes in breast milk to make recommendations for mother/baby nutrition, screen for post-intubation pneumonia infection, adverse pregnancy outcomes, and/or analyze inform supplement recommendations. In some embodiments, methods and systems described herein may be used for nutraceuticals to inform supplement recommendations based on microbiome testing.

[00230] In embodiments, the methods and system may be used in veterinary applications. In embodiments, veterinarians and other animal caregivers use microbiome testing for applications such as detecting periodontitis, inflammatory markers, etc. In embodiments, microbiome testing may be used for the diagnosis and treatment of any animal such as dogs, cats, horses, and other organisms. [00231] In embodiments, the methods and system may be used with mental health treatment and diagnosis, such as depression, stress, and anxiety.

[00232] The term healthcare provider and similar terms as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, a healthcare provider may include various entities depending on the condition, illness, and the like. In some situations, a healthcare provider may be a dentist, clinician, doctor, veterinarian, psychologist, or even the patient themselves. Any of the terms dentist, clinician, doctor, veterinarian, psychologist, or even the patient themselves, include any healthcare worker.

[00233] The term biome dataset and similar terms as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, biome dataset may refer to a collection of data derived from biological samples collected from the oral cavity of a subject. This dataset may include information on the relative abundance of bacteria present in the oral microbiome. In some embodiments, the systems and methods described herein may be adapted to analyze not only bacteria but also viruses or a combination of both bacteria and viruses within the oral microbiome. The analysis module may be configured to detect changes in the relative abundance of viruses, or a combination of bacteria and viruses, between different biome datasets. For cases where viral pathogens are identified, or where a combination of bacterial and viral imbalances is detected, the treatment module may be configured to generate a treatment plan that includes antiviral medications, antibiotics, or a combination thereof. In embodiments, the systems and methods described herein may be designed to handle a broader category of pathogens, including bacteria, viruses, fungi, and other entities. The term "pathogen" as used herein may refer to any agent that can cause disease.

[00234] The term relative abundance and similar terms as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, relative abundance may include the proportion or percentage of a particular bacterial taxon in relation to the total bacterial population in a sample. This may be measured at various taxonomic levels, including family level and genus level.

[00235] The term Al model and similar terms as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, Al model may include an artificial intelligence algorithm, such as a random forest model or decision tree model, neural network, Bayesian network, and the like, that analyzes the relative abundance data of bacteria in the biome dataset to generate a risk score for oral disease. Al models may include models that are trained using training data.

[00236] The term weighted tree data structure and similar terms as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, a weighted tree data structure may include a hierarchical data organization used by the Al model to analyze the relative abundance of bacteria and calculate risk scores and/or treatment regimen recommendations .

[00237] The term risk score and similar terms as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, risk score may include a numerical value generated by the Al model that indicates the likelihood of a subject developing or having an oral disease based on the analysis of their oral microbiome composition. In embodiments, scores may further include scores for treatment regimens that can be predicted or recommended to a subject. [00238] The term treatment regimen and similar terms (i.e., treatment plan) as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, the treatment regimen may include a wide range of approaches, which may be used individually or in combination, depending on the patient's specific needs. The treatment regimen may be tailored to the individual patient's needs and may be adjusted based on the patient's response to treatment and changes in their oral health status over time. These approaches may include, but are not limited to:

[00239] 1. Antibiotic therapies, which may involve local or systemic administration of antibiotics. In embodiments, planned course of treatment may include multiple antibiotics, selected based on the relative abundance of bacteria in the subject's oral microbiome and the antibiotic resistance profiles of those bacteria.

[00240] 2. Probiotic treatments, which may introduce beneficial bacteria to promote a healthy oral microbiome.

[00241] 3. Mechanical interventions, such as scaling and root planing, air cleaning, or mechanical debridement, which may physically remove plaque and tartar.

[00242] 4. Advanced therapeutic techniques, including laser treatments (e.g.. LANAP/LAPIP), photo therapy, or the use of custom trays for delivering antimicrobial or peroxide medicaments.

[00243] 5. Surgical procedures, which may include gum grafting, implant removal (partial or complete), soft tissue grafting, or surgical soft tissue resection.

[00244] 6. Chemical treatments, such as chemical debridement, acid etchant cleaning, or gingival pocket irrigation and disinfection.

[00245] 7. Lifestyle modifications, which may involve stress reduction techniques, smoking cessation programs, or dietary adjustments.

[00246] 8. Oral hygiene enhancements, including the use of prescription toothpaste or floss, increased cleaning frequency, or the use of saliva-increasing or xylitol chewing gums.

[00247] 9. Diagnostic procedures, such as CT scans, which may be used to visualize potential irritants or guide treatment planning.

[00248] 10. Specialized treatments like Perio-protect or Hibba cleanse, which may offer targeted approaches to oral health management.

[00249] The term anaerobic score and similar terms as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, anaerobic score may include a measure derived from the relative abundance of anaerobic bacteria in the oral microbiome, which is used as an input for the Al model in assessing oral disease risk and/or disease treatment recommendations and plans. [00250] The term gram stain profile and similar terms as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, gram stain profile may include the distribution of gram-positive and gram-negative bacteria in the oral microbiome, determined from the relative abundance data and used as an input for the Al model. [00251] The term alpha diversity and similar terms as utilized herein should be understood broadly. Without limitation to any other aspect or description of the present disclosure, alpha diversity may include a measure of the microbial diversity within a single sample from the oral cavity, calculated from the relative abundance data and used as an input for the Al model in assessing oral disease risk and/or disease treatment recommendations and plans.

[00252] The biome analysis process described herein may include a lab process and a data analysis process or a bioinformatics analysis process. The lab process may involve sample collection, preparation, and sequencing, while the data analysis process may include bioinformatics, machine learning, and risk assessment. In some aspects, the lab process may begin with sample collection from the oral cavity', followed by DNA extraction and purification. The lab process may then proceed to library preparation and sequencing. In some cases, the data analysis process may start with processing raw sequences through bioinformatic pipelines for taxonomic classification and relative abundance quantification. The analysis may then apply machine learning algorithms, such as decision trees, to categorize samples and generate risk assessments. These processes may work in tandem to provide comprehensive insights into oral microbiome composition and associated health risks.

[00253] The lab process for microbiome analysis may involve several steps, including sampling, DNA extraction, library preparation, and sequencing. These lab process steps generate high-quality data that can be used for downstream bioinformatics analysis and interpretation. Following sample collection, DNA extraction may be performed to isolate microbial DNA from the collected specimens. This process may involve lysing the microbial cells to release their genetic material, followed by a series of purification steps to remove cellular debris and other contaminants. Various DNA extraction kits and protocols may be employed, depending on the specific requirements of the downstream applications. Library preparation may be the next step in the process, where the extracted DNA is prepared for sequencing. This may involve fragmenting the DNA, adding sequencing adapters, and amplifying specific regions of interest, such as the 16S rRNA gene for bacterial identification. The library' preparation process may be optimized to ensure adequate coverage and representation of the microbial community present in the sample. Sequencing may be the final step in the lab process, where the prepared DNA libraries are sequenced using high- throughput sequencing platforms. The sequencing process generates raw data in the form of DNA sequences, which can then be processed and analyzed using bioinformatics tools to identify and quantify the microbial species present in the sample. In some cases, qualify control steps may be implemented throughout the wet lab process to ensure the reliability and reproducibility of the results.

[00254] Fig. 19 illustrates a flowchart for a sample DNA extraction and purification lab process flow. The process begins with sample collection and/or inspection 1902. Sampling may include sampling of areas in the oral cavity of a subject. Sampling may include isolation of the sample site (i.e., isolation of implant site, using cotton rolls). In some cases, isolation may be important to prevent contamination from other areas of the oral cavity and to ensure the sample accurately represents the microbiome of the specific site being sampled. In some cases, the site may be dried (i.e., using compressed air) which helps to remove any loose debris or saliva that could interfere with the sample collection. A sterilized sampling device (i.e., a Gracey Curette or Periodontal Explorer) may be used for sampling. The sampling protocol may be applied to various locations within the oral cavity, including anterior (front) and posterior (back) sites. Samples can be collected from subgingival (below the gum line), supragingival (above the gum line), or surface locations, depending on the specific sampling requirements. In one example, a sampling device may be used to scrape plaque from the subgingival/submucosal region. Collected samples may be transferred to a vial containing DNA/RNA Shield solution for preserving the genetic material of the microorganisms in the sample and stabilizing DNA and RNA, preventing degradation and maintaining the sample's integrity for subsequent analysis. The samples may be stored at room temperature for short-term storage. To maintain sterility’ and prevent cross-contamination, the sampling device is not reintroduced into the patient's mouth after exposure to the DNA/RNA Shield solution until it has been properly cleaned and sterilized so as to not to affect the accuracy of the microbiome analysis. [00255] Upon collection, each sample may undergo a quality control process. Visual inspection, either by trained laboratory personnel or automated imaging systems, is performed to detect any signs of blood or foreign material that could interfere with downstream analysis. Samples containing such contaminants may be rejected or selected for different processing. Once a sample passes the initial inspection, it moves to the DNA extraction and purification stage. This process involves lysing the microbial cells to release their genetic material, followed by a series of purification steps to isolate the DNA from other cellular components. Various methods may be employed, such as magnetic bead-based extraction or silica membrane spin columns, depending on the specific protocol optimized for oral microbiome samples.

[00256] The purified DNA then undergoes preparation for sequencing. In one example, purification may focus on the 16S rRNA genes, specifically the v3/v4 hypervariable regions. In one example, the DNA extraction and purification process may extract and purify DNA from oral plaque samples using an automated magnetic bead-based system. The method may include the use of the Zymo Extraction Kit, the IsoPure System, and the Qubit4 Fluorometer, with a focus on achieving high dsDNA concentration and observational purity. The process may include the dispensing of reagents (i.e., ZymoBIOMICS DNA extraction kit reagents) into the appropriate wells of a plate 1904. Next, samples may be processed with magnetic particles coated with a DNA-binding surface (i.e., MagBeads). Samples may be thoroughly vortexed to ensure even distribution. A specified volume of these beads is then transferred to the plate containing the extraction reagents. Similarly, the oral plaque samples are vortexed to homogenize the microbial content, and a predetermined volume is transferred to the plate. An automated extraction system (i.e., IsoPure Automated Extraction System) is then prepared and initiated to perform steps that may include cell lysis, DNA binding to the magnetic beads, washing of contaminants, and elution of purified DNA 1906. Following the extraction process, the purified DNA is transferred to a PCR plate 1910. This transfer step prepares the DNA for potential downstream applications such as PCR amplification or library preparation for sequencing. To assess the success of the extraction and quantify’ the yield, an aliquot of the extracted DNA is tested for double-stranded DNA (dsDNA) concentration using a fluorometer 1912.

[00257] Figure 20 figure illustrates a detailed workflow for DNA library preparation and sequencing. The process begins with a well plate containing DNA samples 2002. The first step is PCR Prep 2004, performed using a red automated liquid handling system. This system dispenses reagents and samples for the initial PCR reaction. Following PCR Prep, the samples undergo Pre-PCR Mix 2006. ensuring thorough combination of all components. The next stage is PCR Cycling 2008, carried out in a thermal cycler, which amplifies the DNA fragments. After amplification, the samples return to the liquid handling system for Clean-Up and Pooling 2010. This step purifies the amplified DNA and combines samples as needed. The final stage before sequencing is Pre-Sequence Storage, where the prepared libraries are stored in a temperature-controlled environment 2012. Additionally, a small portion of the pooled library undergoes Pooled Library Quantification 1914 to determine the concentration and qualify of the DNA library’ before sequencing.

[00258] The 16S V3/V4 amplicon library preparation may include a semi-automated process. In one example, preparation may include the use of the Myra liquid handling system and TurboCycler2 PCR Cycler to produce high-quality 16S rRNA V3/V4 amplicon libraries. The preparation process may utilize the Zymo 16S V3/V4 Library Prep Kit and includes key steps such as DNA concentration measurement. Using the extracted DNA from previous steps and additional DNA standards, three replicate library plates may be produced to ensure consistency and accuracy in sequencing. [00259] The 16S rRNA V3/V4 library preparation process begins with the setup and loading of the liquid handling system (i.e., Myra liquid handling system), which automates many of the precise liquid transfer steps required for library preparation. The system may perform pre-PCR sample, reagent, and index transfers. Following the transfers, a plate shaker may be used for pre-PCR mixing. This step ensures thorough mixing of the samples with reagents and indexes, promoting uniform amplification in the subsequent PCR step. PCR amplification may then be carried out using a thermal cycler (i.e.. TurboCycler2). This step selectively amplifies the V3/V4 region of the 16S rRNA genes, creating millions of copies of these specific DNA fragments. After PCR, the system undergoes a decontamination process to prevent any cross-contamination betw een the pre- and post-PCR steps. The system may be used again for post-PCR pooling and cleanup. This step involves combining the amplified products and removing unused reagents, primers, and any potential contaminants that could interfere with sequencing. The concentration of the prepared libraries may be measured using a fluorometer (i.e., a Qubit 4 Fluorometer).

[00260] Fig. 21 illustrates a flow chart for a DNA sequencing process. The process begins with pooled libraries 2102. The next step involves adding PhiX control to the pooled libraries 2104. Following this, the mixture is dispensed into a NextSeq Reagent Cartridge 2106. The process then moves to the Sequencing stage, depicted by a sequencing machine 2108. The final step in the process is File Generation 2110.

[00261] The pooling and sequencing process involves 16S amplicon multiplex sequencing and may use, for example, the Illumina NextS eq 1000 system with the 2x300 Pl Reagent Kit. The process produces raw? sequencing output files that meet quality specifications and are suitable for downstream bioinformatic analysis. This process includes pooling the three library plates into a single pool, adding a PhiX control, and performing multiplex sequencing (for example, using the NextSeq 1000 platform). The sequencing process may begin with pooling libraries from separate preparation plates, including different primer and index combinations. The pooled libraries are then quantified using a fluorimeter to determine the DNA concentration. Based on the quantification results, the pooled libraries may be diluted using RSB (Resuspension Buffer) with Tween. This step ensures that the DNA concentration is within the optimal range for sequencing. Next, a control solution (i.e.. PhiX solution) is prepared and spiked into the diluted library pool at a predefined concentration of (i.e., 30%). The reagent kit seal is then broken, and the pooled libraries with PhiX are dispensed into the reagent cartridge. The flow' cell and reagent cartridge are installed into the sequencer (i.e.. NextSeqlOO sequencer). The sequencing run is then initiated, during which the instrument will perform cluster generation, sequencing by synthesis, and base calling. Upon completion of the sequencing run, the files containing the raw sequencing data are transferred from the sequencer for downstream bioinformatic processing. These files contain the sequencing reads and associated quality scores which can be used for subsequent analysis.

[00262] Following sequencing, the raw data may undergo bioinformatics analysis to transform the sequences into meaningful biological information. This process may involve one or more steps such as: quality⁷ filtering (e.g., trimming of raw reads to remove low-quality⁷ bases and sequencing artifacts), denoising and clustering of sequences into Amplicon Sequence Variants (ASVs) or Operational Taxonomic Units (OTUs), taxonomic assignment of the ASVs/OTUs using reference databases, processing with the BLAST algorithm to identify the bacterial species present in the sample, calculation of relative abundances for each identified taxon, determination of metrics (i.e. alpha diversity metrics). In one example, the QIIME2 platform may be used as a framework for these analyses.

[00263] The lab process may include a number of validation procedures to ensure that the entire lab process produces accurate and reliable bacterial relative abundance profiles. Appropriate validation processes may be used to assess the accuracy, precision, sensitivity, specificity, reportable range, and robustness of the complete workflow, as well as demonstrate consistency in results across different sample types, reagent lots, operators, and processing dates.

[00264] In embodiments, various processes may be used to calculate the relative abundance of microbial species in microbiome samples. After quality⁷ filtering of the raw sequencing data, the sequences may be clustered into Amplicon Sequence Variants (ASVs) or Operational Taxonomic Units (OTUs) using algorithms that group similar sequences together. These ASVs or OTUs may then be assigned taxonomic classifications by comparing them to reference databases of known bacterial 16S rRNA gene sequences.

[00265] To calculate relative abundance, the number of sequences assigned to each taxonomic group (e.g., species, genus, or family) may be counted and divided by the total number of sequences in the sample. This provides a measure of the proportion of each microbial taxon in the community. In some cases, the relative abundance calculations may be normalized to account for factors like 16S rRNA gene copy number variations between species. Additional statistical analyses may be performed to assess the significance of abundance differences between samples or groups. Visualization tools may also be used to generate graphs and charts depicting the relative abundances across samples.

[00266] The bioinformatics pipeline may be optimized for the specific sequencing platform and sample ty pe used. Quality control steps may be incorporated throughout to ensure the reliability of the relative abundance measurements. The resulting relative abundance data may then be used for downstream analyses to gain insights into the oral microbiome composition and its potential associations with oral health or disease.

[00267] In embodiments, relative biome measures may include various quantitative and qualitative assessments of the microbial community composition within a given sample. These measures may provide insights into the diversity⁷, abundance, and interactions of microorganisms present in the oral cavity. Some relative biome measures that may be included are the relative abundance of specific taxa at different taxonomic levels, such as phylum, class, order, family, genus, and species. Alpha diversity metrics, which measure the diversity⁷ within a single sample, may include the Shannon diversity index, Simpson diversity index, Chaol richness estimator, and observed species count. Beta diversity metrics, which compare the diversity between samples, may include Bray-Curtis dissimilarity. UniFrac distance (weighted and unweighted), and Jaccard index. Other measures may include the ratio of gram-positive to gram-negative bacteria, proportion of aerobic, anaerobic, and facultative anaerobic bacteria, and relative abundance of potential pathogens or beneficial microorganisms. Functional gene abundance related to specific metabolic pathways or virulence factors, microbial community network analysis including co-occurrence patterns and keystone species identification, and the ratio of early colonizers to late colonizers in dental plaque formation may also be considered.

[00268] In embodiments, relative biome measures may include the percentage of biofilm-forming bacteria, which can indicate the propensity for plaque formation and adherence to dental surfaces. The proportion of gram-negative bacteria may also be calculated, as these microorganisms often play a significant role in periodontal diseases and other oral health issues. Additionally, the percentage of anaerobic and facultative anaerobic bacteria may be determined, as these organisms can thrive in oxygen-depleted environments and are often associated with various oral infections.

[00269] In embodiments, analysis of the relative abundance of biome bacteria may include analysis of the progression of the relative abundance of different colonizers and/or different types of colonizers. In one example, analysis may include tracking changes in the relative abundance of early, middle, and late colonizers over time.

[00270] Early colonizers primarily are the initial inhabitants of the oral microbiome. These bacteria adhere to the tooth surface, forming a foundation for subsequent colonization. Their abil i ty to attach to dental surfaces and produce extracellular polymeric substances creates a favorable environment for other microorganisms to colonize. As the biofilm develops, middle colonizers (such as Fusobacterium and Veillonella species) increase in relative abundance. These bacteria play a role in bridging the gap between early and late colonizers. Fusobacterium species, for example, possess multiple adhesins that allow them to coaggregate with both early and late colonizers, facilitating the attachment of more pathogenic species. Late colonizers, which may include potentially pathogenic species (such as Porphyromonas gingivalis and Tannerella forsythia), typically increase in relative abundance as the biofilm reaches maturity. These bacteria are often associated with periodontal diseases and can thrive in the anaerobic environment created by the mature biofilm.

[00271] The analysis of colonizer progression may include temporal analysis. Temporal analysis may include tracking changes in relative abundance over time, allowing for the identification of patterns or shifts in microbiome composition. For example, a sudden increase in late colonizers might indicate a transition towards a disease state. The analysis of colonizer progression may include ratio analysis. Analysis may include calculating and monitoring ratios between early, middle, and late colonizers. The analysis of colonizer progression may include network analysis. Analysis may include examining co-occurrence patterns and interactions between different colonizer groups. The analysis of colonizer progression may include correlation with clinical parameters. Analysis may include associating changes in colonizer progression with clinical indicators of oral health or disease. The analysis of colonizer progression may include predictive modeling. Analysis may include using the progression patterns of different colonizers to train models to predict the onset or progression of oral diseases. These models may incorporate machine learning algorithms to improve accuracy over time. The analysis of colonizer progression may include treatment response assessment. Analysis of the progression of colonizers can be evaluated in response to various treatments or interventions.

[00272] In embodiments, a trained Al model may be used to determine a disease risk score by analyzing relative abundance data from oral microbiome samples.

[00273] In one example, an Al model may be a tree structure such as a weighted tree data structure. A tree structure is a structure that represents a series of decision points based on the relative abundance of specific bacterial taxa or groups of taxa. The model may start with a root node that represents a possible dataset. From this root, the tree may branch into child nodes based on thresholds of relative abundance for certain key bacterial species or genera. For example, one branch may split based on the relative abundance of Porphyromonas gingivalis, with samples above a certain threshold following one path and those below following another. As the tree branches further, it may incorporate additional taxa or combinations of taxa. Each branching point may be determined by analyzing which split in the data provides the most information gain or reduction in entropy with respect to disease risk. This process may continue until leaf nodes are reached, where each leaf represents a specific risk score or category'.

[00274] The path from the root to a leaf node may represent a series of decisions based on the relative abundance of multiple bacterial species. For instance, a high-risk path might involve high levels of late colonizers, combined with low levels of beneficial early colonizers. In some implementations, the tree model may incorporate temporal data, allowing it to assess changes in relative abundance over time. This may enable the model to capture dynamic shifts in the microbiome that may be indicative of increasing disease risk.

[00275] The tree model may be trained on a dataset of samples with known disease outcomes, allowing it to leam the patterns of relative abundance associated with different levels of disease risk. Once trained, the model may be applied to new samples, traversing the tree based on the sample's relative abundance data to arrive at a risk score.

[00276] In some embodiments, multiple trees may be combined in an ensemble method, such as a random forest, where the final risk score is determined by aggregating the outputs of many individual trees.

[00277] The tree model may also incorporate other relevant factors beyond relative abundance, such as alpha diversity measures or the presence of specific functional genes. These additional features may be integrated into the decision-making process at various levels of the tree.

[00278] In embodiments, Bay esian methods may be used for the analy sis of relative abundance data in oral microbiome studies to provide a probabilistic framework for understanding microbial community composition and its relationship to oral health outcomes. These methods can incorporate prior knowledge and uncertainty into the analysis. In embodiments, Bayesian statistical methods may be used to estimate the relative abundance of bacterial taxa while accounting for uncertainty in the measurements. This may involve using hierarchical models that can handle the compositional nature of microbiome data. In embodiments. Bayesian state-space models or hidden Markov models may be used to analyze the progression of relative abundance over time, capturing the dynamics of colonizer succession and biofilm maturation. In embodiments, Bayesian predictive models may be used to assess the risk of oral diseases based on relative abundance profiles. These models can provide probabilistic predictions and quantify uncertainty in the estimates.

[00279] In embodiments, the Al model may include a neural network model. A neural network model may be used to determine risk scores based on relative abundance data from oral microbiome samples. The model may consist of multiple layers of interconnected nodes, or neurons, that process and transform the input data to generate a risk score output. The input layer of the neural network may receive the relative abundance data for various bacterial taxa. Each input neuron may correspond to a specific taxon, with its activation level representing the relative abundance of that taxon in the sample. Hidden layers in the network may then process this input data through a series of weighted connections and activation functions. These layers may capture complex patterns and relationships within the relative abundance data that are indicative of disease risk. The output layer of the network may produce a risk score, which could be a continuous value representing the probability of disease or a categorical classification of risk level (e.g., low, medium, high). During training, the neural network may be presented with labeled datasets containing relative abundance profiles and known disease outcomes. The network may adjust its internal weights through backpropagation to minimize the difference between its predicted risk scores and the actual outcomes.

[00280] The neural network may incorporate various architectural features and may include convolutional layers (may be used to capture local patterns in the relative abundance data, such as co-occurrence of certain bacterial groups), recurrent layers (may be employed to analyze temporal sequences of relative abundance data, allowing the model to capture dynamic changes in the microbiome over time), attention mechanisms (may be implemented to focus on the most relevant taxa or combinations of taxa for risk prediction).

[00281] In some implementations, ensemble methods may be used, where multiple neural networks are trained, and their outputs are aggregated to produce a final risk score. This approach may improve the robustness and accuracy of the predictions. The neural network model may be periodically retrained or fine-tuned as new data becomes available, allowing it to adapt to changes in microbial populations or emerging patterns of disease risk. This continuous learning process may help maintain the model's accuracy and relevance over time.

[00282] The risk score generated from the analysis of relative abundance data may be utilized to determine an appropriate treatment regimen. In some implementations, the treatment determinations may include a multi-step process that may involve additional modeling and decision-making algorithms. This approach may allow for personalized treatment strategies based on the specific microbial composition of an individual's oral microbiome.

[00283] In some implementations, a decision support system may be used that takes the risk score as input and outputs treatment recommendations. This system may be based on a series of if-then rules, decision trees, or more complex machine learning models such as random forests or neural networks. [00284] For example, a decision tree model may be used where the risk score serves as the primary input. The tree may branch based on different thresholds of the risk score, with each branch leading to different treatment options. Low risk scores may lead to recommendations for preventive measures such as improved oral hygiene practices or more frequent dental check-ups. Moderate-risk scores may suggest more aggressive preventive measures or early interventions, while high-risk scores may indicate the need for immediate therapeutic interventions.

[00285] The model may also incorporate other factors alongside the risk score, such as: specific bacterial compositions (the presence or absence of certain key bacterial species may influence treatment decisions, such as specific antibiotics), patient history, and/or treatment efficacy data. [00286] The treatment regimen determined by the model may include various components and may include one or more of the treatment options described herein.

[00287] In some implementations, an output of analysis may include a treatment report or a treatment plan. A treatment report may include a comprehensive overview of the patient's oral health status and recommended interventions based on the analysis of their oral microbiome. The report may be structured to provide information for both the healthcare provider and/or the patient. A report may include sample collection details (e.g., information about when and how the oral microbiome samples were collected), microbiome analysis summary (e.g., an overview of the key findings from the microbiome analysis, including the relative abundance of different bacterial), risk assessment (e.g., a risk score or category, such as low, moderate, high, for various oral health conditions based on the microbiome analysis and other factors), and treatment recommendations.

[00288] Fig. 22 is a flowchart of an example method 2200 for generating a post-dental implant treatment. At step 2202, method 2200 may include collecting biological samples from an oral cavity of a subject. This step may include obtaining samples from various sites within the subject's mouth, such as saliva, dental plaque, or subgingival plaque. Depending on the specific analysis requirements and research objectives, these samples may be handled in different ways. In some cases, samples from different sites may be combined into a single composite sample to provide an overall representation of the oral microbiome. This approach can be useful for general assessments or when resources are limited. How ever, combining samples may result in the loss of site-specific information and could potentially mask important localized variations in microbial communities. Alternatively, samples from different sites may be kept separate to maintain the integrity' of sitespecific microbial populations. This approach allows for a more detailed analysis of the distinct microbiomes associated with various oral niches. This method preserves the unique microbial compositions of each sampled area, enabling more precise characterization of site-specific microbial communities and their potential associations with oral health or disease states. Maintaining separate samples also allows for comparative analyses between different oral sites within the same individual, which can provide insights into the spatial distribution and diversity' of oral microbiota. The collection process may utilize tools like swabs, curettes, or other dental instruments. The collected samples may be processed using one or more techniques described herein to obtain relative abundance data of bacteria.

[00289] At step 2204, method 2200 may include identifying the subject as at a first risk of oral disease based on a risk score. In embodiments, different ty pes of risk scores may be generated to assess the likelihood of various oral health conditions based on the analysis of the oral microbiome. These risk scores may include: periodontal disease risk score, peri-implantitis risk score, halitosis risk score, oral cancer risk score, systemic health risk score (score of systemic health conditions such as cardiovascular disease or diabetes), treatment response risk score (the likelihood of a positive response to specific treatments based on the current microbial composition), recurrence risk score (the likelihood of condition recurrence based on current microbial profiles), and/or oral dysbiosis score. These risk scores may be presented individually or combined into an overall oral health risk assessment.

[00290] The values of the risk scores may represent different levels of risk for health conditions. In some implementations, the risk scores may be represented as continuous values, providing a more granular assessment of risk. For example, a periodontal disease risk score may range from 0 to 100, where higher values indicate a greater likelihood of developing or progressing periodontal disease. In other implementations, the risk scores may be categorized into discrete levels, such as low, medium, and high. This categorization may simplify interpretation for both healthcare providers and patients. For instance, “low risk” score may indicate minimal likelihood of developing the condition or a stable oral health status, “medium risk” score may suggest an increased chance of developing the condition or a need for preventive measures, and a “high risk” score may signify a significant likelihood of developing the condition or an urgent need for intervention. In one example, the continuous values may be mapped to these categories using predefined thresholds. For example, a periodontal disease risk score of 0-30 may be categorized as low risk, 31-70 as medium risk, and 71- 100 as high risk. In some cases, the risk assessment may incorporate both continuous and categorized values. The continuous value may provide a precise measure for tracking changes over time, while the categorized value may offer an easily understandable overview for quick decision-making. The interpretation of these risk scores may vary based on factors such as the specific oral health condition, patient demographics, and clinical context.

[00291] The risk score may be generated by an Al model. In one example, the Al model may be a model that traverses a weighted tree data structure to analyze the relative abundance of bacteria in the biome dataset. The Al model referred to here may be a machine learning algorithm specifically designed to analyze and interpret data related to the relative abundance of bacteria in a biome dataset. The Al model utilizes a weighted tree data structure, which includes a hierarchical arrangement of nodes where each node represents a decision point or a feature of the data. The process of traversing this tree structure may include the Al model moving through the nodes, making decisions at each point based on the input data. In this case, the input data is the relative abundance of various bacterial species or genera in the oral microbiome sample. As the model traverses the tree, it evaluates the abundance of specific bacteria or groups of bacteria at each node, using predetermined thresholds or rules to navigate to the next appropriate node. The "weighted" aspect of the tree structure indicates that certain features or decision points may have more influence on the final output than others. This weighting is determined during the training phase of the model, where it leams from a large dataset of known samples and outcomes. By traversing this weighted tree structure, the Al model can efficiently process complex microbiome data and classify it into meaningful categories or risk scores as described herein.

[00292] In some implementations, the input to the Al model may include various metrics and may include relative abundance values of specific bacterial species, family, or genera, alpha diversity scores, gram stain profiles, anaerobe scores, and/or the like.

[00293] At step 2206, method 2200 may include generating a plan for an antibiotic regimen for the subject at the first risk of oral disease after dental implant surgery . In embodiments, a plan may be generated when the first risk is above a threshold value (e.g., score of 50 or higher, or a score of “medium risk” or “high risk”).

[00294] The antibiotic regimen may include a plurality of antibiotics selected based on the relative abundance of the bacteria and antibiotic resistance of the bacteria. The antibiotic regimen may be determined by incorporating data from an antibiotic resistance database that may contain comprehensive information on the resistance profiles of various bacterial species to different antibiotics. By cross-referencing the relative abundance data of bacteria in the subject's oral microbiome with the antibiotic resistance database, the system may generate a more targeted and effective antibiotic regimen. The database may include aspects of known resistance mechanisms for different bacterial species, geographical variations in antibiotic resistance patterns, temporal trends in the development of resistance and the like. Antibiotics for the regimen, and the details of the regimen may consider the effectiveness of each antibiotic against the most abundant bacterial species in the subject's oral microbiome, the likelihood of resistance based on the subject's bacterial profile and the resistance data in the database, potential synergistic effects between different antibiotics, and/or the subject's medical history, including any previous antibiotic treatments and known allergies. The system may use machine learning algorithms to analyze the interactions between the subject's microbiome data and the antibiotic resistance database. In some cases, the system may recommend alternative treatment strategies if the antibiotic resistance database indicates a high likelihood of resistance to commonly used antibiotics.

[00295] In embodiments, the treatment plan may be administered to a patient. Administering may include a comprehensive approach that includes patient education, prescription and dispensing of antibiotics, regular monitoring and follow-up appointments, adjustments to the regimen as needed, complementary treatments, probiotic supplementation, long-term management strategies, integration with dental procedures, use of digital health tools, and the like. [00296] In some cases, a second biological sample may be collected from the oral cavity of the subject at a later time point. The later time point for collecting a second biological sample may vary depending on the specific clinical context and treatment goals. In some cases, the second sample may be collected after a couple of months, allowing sufficient time for the oral microbiome to potentially stabilize following initial interventions. In other instances, the later time point may coincide with the completion of the antibiotic regimen, providing insights into the immediate effects of the treatment on the oral microbiome composition. Additionally, the healthcare provider may choose to collect samples at various intervals, such as at 3 months, 6 months, or 1-year post-treatment, to monitor long-term changes in the oral microbiome. The timing may also be influenced by factors such as the patient's recovery' progress, the presence of any complications, or the need for additional dental procedures. This second sample may be processed and analyzed to generate a new biome dataset. The Al model may then analyze the relative abundance data from this second dataset to generate an updated risk score.

[00297] If the updated risk score remains high or has increased, the healthcare provider may consider modifying the treatment plan. This modification may involve adjusting the antibiotic regimen, changing the types of antibiotics used, altering the dosage, or extending the duration of treatment. In some instances, the treatment plan modification may include additional interventions beyond antibiotics, such as any of the other treatment options described herein.

[00298] If the updated risk score decreases or is low, the healthcare provider may consider modifying the treatment by pausing the treatment or adjusting the antibiotic regimen (reducing dosage or changing to less aggressive antibiotics). In some instances, the treatment plan modification may include stopping antibiotics and the usage of other interventions.

[00299] In some implementations, the Al model may be designed to compare the first and second datasets, and analysis may include not only a comparison of a risk score but also the changes in relative abundance between the two time points. This comparative analysis may provide insights into the effectiveness of the current treatment and guide any' necessary modifications to the treatment plan.

[00300] Fig. 23 is a flowchart of an example method 2300 for generating a treatment. At step 2302, method 2300 may include collecting biological samples from an oral cavity of a subject to provide a biome dataset. At step 2304, method 2300 may include identify ing the subject as having a first risk score for a disease based on a risk score that is generated from relative abundance data in the biome dataset by a trained Al model that predicts scores based on training examples.

[00301] In some implementations, a random forest model may be used for the Al model. This ensemble learning method may be particularly effective for handling the complex, high-dimensional data associated with microbiome analysis. The random forest model may include multiple decision trees, each trained on a random subset of the input features and samples. For microbiome data, these features may include the relative abundances of different bacterial taxa at various taxonomic levels (e.g., phylum, class, order, family, genus, or species). When analyzing new samples, each decision tree in the forest may independently predict a risk score or classification based on the relative abundance data. The final prediction may then be determined by aggregating the results from all trees. Aggregation may include majority voting or averaging. A random forest model may be trained on a large dataset of oral microbiome samples with known health outcomes or risk levels. The model may then be used to analyze new samples, predict risk scores, or classify samples into different health categories based on their relative abundance profiles. In embodiments, the random forest model may also be integrated with other data sources, such as clinical measurements, patient history, or genetic information, to provide a more comprehensive risk assessment.

[00302] In some implementations, the Al model may be designed to compute risk scores for various diseases beyond oral health conditions. The model may analyze the biome dataset to identify patterns and correlations that may be indicative of systemic health issues or risks. For example, the Al model may generate risk scores for conditions such as cancer, cognitive decline, rheumatoid arthritis, or Alzheimer's disease. In one example, for cancer risk assessment, the Al model may analyze the relative abundance of certain bacterial species that have been linked to increased cancer risk. Some oral bacteria may produce metabolites or trigger inflammatory responses that could potentially contribute to cancer development or progression.

[00303] At step 2306, method 2300 may include generating a treatment plan for the subject having the first risk score for disease. The treatment plan may be based on the first risk score and the relative abundance data. A treatment plan may include any of the treatments outlined herein. The treatment plans may be customized to the patient, the absolute or relative abundance data, and/or risk scores. The treatment plans may be administered to the patient according to the plan.

[00304] Fig. 24 is a flowchart of another example method 2400. At step 2402, method 2400 may include receiving a sample comprising bacteria from an oral cavity of a subject. At step 2404, method 2400 may include analyzing the sample to generate an analysis comprising a relative abundance of bacteria in the sample. The samples may be analyzed using, for example, the methods described with respect to figures 19-21. At step 2406, method 2400 may include determining, using a trained machine model, a classification of the relative abundance indicative of an oral health issue. Any trained Al model described herein may be used. At step 2408, method 2400 may include generating a report based on the classification. The report may include personalized intervention recommendations. In some implementations, the report may be accessed electronically via the Internet. The report may be made available through a secure online portal, allowing authorized users such as healthcare providers or patients to view and interact with the information. This portal may provide a user-friendly interface for accessing detailed analysis results, visualizations of microbial compositions, and tailored treatment suggestions. In some cases, the portal may offer features such as historical data comparisons, enabling users to track changes in oral health over time.

[00305] Fig. 25 is a schematic of an example system 2500 for determining a post-dental implant treatment. A system for determining a post-dental implant treatment method may include various components configured to collect 2502, analyze 2504, interpret 2506, and report 2508. System 2500 may include one or more collection devices 2510 that may be introduced into the oral cavity' to gather biological samples from various surfaces. This device may be designed to collect samples from specific areas of interest, such as around dental implants, saliva, gums, and the like. System 2500 may include a storage unit 2512 containing a storage medium may be included to preserve the collected samples. This unit 2512 may utilize stabilizing agents or temperature-controlled environments to maintain the integrity of the biological material for subsequent analysis. System 2500 may also include a processing module 2514 configured to extract biological material from the collected samples. This module may employ various techniques such as centrifugation, filtration, or chemical lysis to isolate bacteria from the samples as described herein.

[00306] System 2500 may further include an analysis unit 2516 configured to identify and quantify microbial species present in the isolated bacterial material. This unit may utilize techniques such as DNA sequencing, PCR, or mass spectrometry to characterize the microbial community’ and may include devices and platforms described herein. A data recording module 2518 may be included to record the relative abundance of the identified microbial species.

[00307] System 2500 may further include a models module. The models module may stage, execute, and monitor Al models. The Al models may be configured to correlate the recorded relative abundance data with a health score, such as an oral health score. A treatment identification unit 2522 may be included to generate one or more treatment regimens based on the health score and/or the relative abundance of bacteria. This unit may draw upon a database of treatment options and clinical guidelines to provide tailored recommendations. System 2500 may further incorporate a user interface 2524 that provides a platform for users to access the recorded relative abundance data and the generated treatment regimens.

[00308] In embodiments, system 2500 may include a training unit 2526 configured to train Al models 2520 for determining risk scores and/or treatment options. In some implementations, the training unit 2526 may be part of the bioinformatics interpretations 2506 part of the system 2500 or may be a separate module. [00309] In embodiments, the training unit 2526 may use diverse and comprehensive training data 2528 to effectively train Al models 2520 for determining risk scores and/or treatment options. The training data 2528 may include microbiome composition data. The microbiome composition data may include the relative abundance profiles of microbial species at various taxonomic levels (e.g., phylum, class, order, family, genus, species) from a number of oral samples. The data may include samples from healthy individuals as well as those with various health conditions, such as oral health conditions. The training data 2528 may include clinical outcome data. Clinical outcome data may include information on diagnosed health conditions, disease progression, and treatment outcomes associated with each microbiome sample. The data may cover a range of oral health issues such as gingivitis, periodontitis, peri-implant mucositis, and peri-implantitis. The training data 2528 may further include patient metadata. Patient metadata may include demographic information (e.g.. age, sex, ethnicity), medical history, lifesty le factors (e.g., smoking status, diet), oral hygiene habits, and genetic information when available.

[00310] In some implementations, training data 2528 may include temporal data. Temporal data may include longitudinal samples from the same individuals over time, allowing the model to learn patterns of microbiome changes associated with disease onset or progression. Training data 2528 may include treatment data. Treatment data may include details of treatment regimens applied to patients and their outcomes. Training data 2528 may include antibiotic resistance data. Antibiotic resistance data may include information on the antibiotic resistance profiles of various bacterial species to inform appropriate antibiotic recommendations. Training data 2528 may include systemic health data. Systemic health data may include information on the patient's overall health status and any diagnosed systemic diseases. Training data 2528 may include environmental factors. Environmental factors may include factors that may influence oral microbiome composition, such as geographical location or local water fluoridation levels. Training data 2528 may include control group data. Control group data may include samples from a control group of healthy individuals without health issues.

[00311] Training unit 2526 may process and integrate the training data. In some implementations, the training unit 2526 may perform data normalization, feature selection, and dimensionality reduction to prepare the data for model training.

[00312] Data normalization may include operations such as scaling the input features to a standard range, typically between 0 and 1 or -1 and 1. This process may help prevent features with larger numerical ranges from dominating those with smaller ranges, potentially improving the model's performance and convergence speed. Normalization techniques may include min-max scaling, z- score normalization, or decimal scaling. Feature selection may be employed to identify the most relevant features for the prediction task. Feature selection may include correlation analysis, mutual information, or recursive feature elimination. Dimensionality reduction techniques may include principal component analysis (PCA). In some cases, training unit 2526 may implement feature engineering to create new, more informative features from the existing data. This may involve combining multiple features, applying mathematical transformations, or leveraging domain knowledge to create more predictive inputs for the model.

[00313] The training unit 2526 may use the training data 2528 to train weighted tree structures. In one example, training unit 2526 may be used to train a random forest model through a process that includes bagging and decision tree construction. The random forest algorithm may create multiple decision trees and combine their outputs to make predictions. In embodiments, the preprocessed and normalized training data may be used as input for training the random forest model.

[00314] Training may include a bagging process. Using the bagging process, multiple subsets of the training data are created through random sampling with replacement. Each subset may contain a random selection of samples and features from the original dataset. For each subset created through bagging, a decision tree may be constructed. The tree may be built by recursively splitting the data based on the most informative features at each node. At each node of the decision tree, a random subset of features may be considered for splitting. Each decision tree may be grown to its full depth or until a stopping criterion is met, such as a minimum number of samples in a leaf node. Multiple decision trees may be created, forming an ensemble of trees that constitute the random forest. For making predictions on new data, each tree in the forest may provide its own prediction. These individual predictions may then be aggregated, typically through majority voting or averaging. [00315] The performance of the random forest model may be assessed, and the model's performance may be optimized by adjusting hyperparameters such as the number of trees, maximum tree depth, or minimum samples per leaf. Hyperparameter tuning may involve systematically searching for the best combination of parameters that yield the highest model performance. In one example, the number of trees in the forest may be adjusted to balance between model complexity and computational efficiency. Increasing the number of trees may improve the model's stability⁷ and predictive power but may also increase prediction time. The training unit 2526 may experiment with different numbers of trees to find an optimal balance. In another example, maximum tree depth may be tuned to control the complexity of individual trees in the forest. Deeper trees may capture more complex relationships in the data but may also be prone to overfitting. The training unit 2526 may adjust this parameter to find a depth that provides good predictive performance without overfitting to the training data. Other hyperparameters that may be adjusted include the minimum number of samples required to split an internal node, the maximum number of features to consider when looking for the best split, and the method used to measure the quality of a split. Training unit 2526 may employ various hyperparameter tuning techniques such as grid search, random search, or Bayesian optimization to tune the hyperparameters.

[00316] To assess the model's performance for each hyperparameter combination, a loss function may be incorporated to quantify the difference between the model's predictions and the actual values in the training data. The loss function is used as a measure of the model's performance, with lower values indicating better performance. For random forest models, loss functions may include mean squared error (MSE) or cross-entropy loss functions. The loss function may be calculated for each fold of the cross-validation process. The training unit 2526 may use the average loss across all folds as a metric to compare different hyperparameter configurations. By minimizing this loss, the training unit may identify the optimal hyperparameter settings that yield the best performance on the validation data.

[00317] Training unit 2526 may use the training data 2528 to train a neural network-based model. Training may include data preprocessing to ensure compatibility with the neural network architecture. This may include normalization, encoding categorical variables, and handling missing values. In one example, categorical variables may be converted into numerical values, often using techniques such as one-hot encoding or label encoding. One-hot encoding creates binary columns for each category⁷, while label encoding assigns a unique integer to each category⁷. The structure of the neural network may be designed based on the number of variables in the data. This may include determining the number of layers, neurons per layer, and activation functions.

[00318] The initial weights of the neural netw ork may be set using various techniques such as random initialization. During training, the input data may be fed through the netw ork, with each neuron computing its output based on its inputs and w eights. The network's predictions may be compared to the actual values in the training data using a loss function. The gradient of the loss function with respect to each weight in the netw ork may be computed, allowing for the calculation of how each weight contributes to the overall error. The weights may be adjusted using an optimization algorithm such as Stochastic Gradient Descent (SGD) to minimize the loss function. Training may be repeated for multiple epochs, with each epoch representing a pass through the training dataset. A separate validation dataset may be used to monitor the model's performance on unseen data and prevent overfitting. Various hyperparameters such as learning rate, batch size, and network architecture may be adjusted to optimize performance.

[00319] During the training process, the loss function may be calculated after each forward pass through the network. This value may then be used in the backpropagation step to compute gradients and update the network's weights. The goal of training may be to minimize this loss function, thereby improving the model's predictions. In some implementations, the loss function may be modified to incorporate domain-specific knowledge about oral microbiome data. For example, it may be weighted to give more importance to certain bacterial species known to be particularly indicative of oral health issues. Additionally, regularization terms may be added to the loss function to prevent overfitting, which may be especially important when dealing with high-dimensional microbiome data. In embodiments, loss functions may include cosine similarity loss, mean squared error, binary cross-entropy, and the like.

[00320] In some implementations, Large Language Models (LLMs) may be utilized to enhance the analysis and interpretation of oral microbiome data. The training process for LLMs for microbiome analysis may include fine-tuning a pre-trained model. A pre-trained LLM model may be fine-tuned on more specific tasks related to microbiome analysis. Fine tunning may include training on datasets that pair microbiome profiles with health outcomes or treatment recommendations. Fine-tuning an LLM may include providing natural language descriptions of biome abundance and outcomes. For instance, a training example might include a detailed breakdown of bacterial species abundance alongside a textual description such as "High abundance of Streptococcus and low diversity of commensal bacteria, indicating an increased risk of dental caries." The LLM model may learn to recognize patterns in the abundance data and associate them with specific health implications, enabling it to generate informative and context-rich natural language outputs.

[00321] In embodiments, an LLM may be adapted for specific tasks such as generating treatment recommendations, interpreting microbiome profiles, or answering queries about oral health based on microbiome data. For example, given a microbiome profile, an LLM may generate human -readable interpretations, explaining the significance of the microbial composition and potential health implications. In another example, LLMs may assist in generating personalized treatment recommendations based on microbiome profiles, patient history’, and current scientific knowledge. [00322] Fig. 26 is a flow chart of an example method 2600 for using a trained machine learning model to identify a treatment based on oral biome composition. At step 2602, method 2600 may’ include training a machine model. A computer may train a machine model using input data and a selected training algorithm. The input data may include oral microbiome samples, patient health records, and known associations between microbial compositions and oral diseases. The training algorithm may incorporate a loss function to measure the model's performance and guide the learning process. The loss function may quantify the difference between the model's predictions and the actual outcomes in the training data. The loss function may be selected based on the type of model being trained. Models may include random forest models, neural network-based models, and/or LLMs. [00323] At step 2604, method 2600 may include detecting health-adverse biome compositions. Once trained, the model may analyze new oral microbiome samples to detect compositions associated with poor oral health. This may involve identifying specific bacterial species, ratios between different types of bacteria, or overall diversify metrics that are indicative of an unhealthy oral environment. At step 2606, method 2600 may include associating biome compositions with oral diseases. The model may then determine which specific oral diseases are likely associated with the detected health- adverse biome compositions. This step may involve mapping the detected microbial patterns to known disease states based on the relationships learned during training.

[00324] At step 2608, method 2600 may include identifying treatment types. Based on the detected health-adverse biome compositions and their associated oral diseases, the model may identify appropriate types of treatments. These treatments may include antibiotics, probiotics, dietary changes, specific oral hygiene practices and/or any other treatments. At step 2610, method 2600 may include generating a treatment regimen. A detailed treatment regimen based on the identified treatment type and the specific biome composition of the patient may be generated. This regimen may include factors such as the type and dosage of medications, duration of treatment, and any complementary therapies or lifestyle modifications. The generated treatment may be administered to a patient.

[00325] Fig. 27 is a flowchart of an example method 2700 for training a machine learning model for biome analysis. At step 2702, method 2700 may include collecting a set of biome data for a group of patients from a database. This biome data may include information about the microbial composition of the oral cavity, such as relative abundances of different bacterial species or genera. At step 2704, one or more transformations may be applied to each biome dataset to create a modified set of biome data. These transformations may include normalization techniques, feature scaling, dimensionality reduction methods, and the like. At step 2706. a first training set may be created by combining the modified set of biome data with a set of patient data. This patient data may include demographic information, medical history, or other relevant factors that could influence oral health. At step 2708, the machine learning model may undergo a first stage of training using the first training set. This initial training may allow the model to learn basic patterns and relationships between the biome data and patient characteristics. At step 2710, a second training set may be assembled for a subsequent stage of training. This set may include the first training set, the modified biome data, and biome classification data generated by themachine learning model after the first stage of training. The inclusion of the model's initial classifications may allow for refinement and improvement of its predictive capabilities. At step 2712, the machine learning model may undergo a second stage of training using the expanded second training set. This stage may enable the model to fine-tune its predictions and potentially capture more complex relationships within the data. This two-stage training approach may allow for iterative improvement of the model's performance. By incorporating the model's own classifications from the first stage into the second stage of training, the method improves the model's ability to recognize and interpret patterns in biome data.

[00326] In one example, the machine learning model may be a random forest machine learning model. In the first stage, training may include constructing a multitude of decision trees at training time and outputting the class that is the mode of the classes (classification) or mean prediction (regression) of the individual trees. In the second stage, training may include categorization methods. This stage focuses on refining the model's ability to categorize data based on learned patterns from the first training stage.

[00327] In another example, the machine learning model may be a neural network model. The first stage of training may involve training the model and adjusting weights and biases. In the second stage, the pre-trained neural network model may be fine-tuned. Fine-tuning involves continuing the training process to refine the w eights of the netw ork for the specific task by training on a smaller dataset that is more focused on the particular features of interest.

[00328] Fig. 28 is a flowchart of an example method 2800 for multi-site biome sampling and analysis. At step 2802, method 2800 may include introducing one or more collection devices into an oral cavity⁷ and collecting biological samples from a first site and a second site within the oral cavity of a subject. These sites may be selected based on their relevance to specific oral health conditions and/or to provide a comprehensive overview of the oral microbiome. At step 2804. following sample collection, biological material may be extracted from the collected samples to isolate bacteria. This step may involve various techniques such as centrifugation, filtration, or chemical lysis to separate bacterial cells from other components in the sample and may include one or more of the processes described herein. At step 2806, the isolated bacteria may then be analyzed to identity- and quantify microbial species at both the first and second sites. This analysis may utilize methods such as 16S rRNA sequencing, quantitative PCR, or metagenomic sequencing as described herein.

[00329] At step 2808, method 2800 may include recording the relative abundance of the microbial species at both the first and second sites. At step 2810, the recorded relative abundance at the first site may be correlated with a first oral health score using a first model. Similarly, the relative abundance at the second site may be correlated with a second oral health score using a second model. These models may be tailored to interpret site-specific microbial profiles and their potential implications for oral health. At step 2812, the method may further involve analyzing the first and second oral health scores using a third model to determine a third, comprehensive oral health score. This step may allow for the integration of site-specific information to provide a more holistic assessment of the subject's oral health status.

[00330] At step 2814, based on the third oral health score, one or more treatment regimens may be identified. These regimens may be tailored to address the specific microbial imbalances or oral health issues indicated by the comprehensive analysis. Step 2814 may include providing a platform through a user interface for users to access the identified treatment regimens.

[00331] The three models used in the multi-site biome sampling and analysis method may differ from one another and can be trained on different data to optimize their performance for specific tasks. The first and second models, which correlate the relative abundance of microbial species at the first and second sites with oral health scores, may be tailored to the unique characteristics of each sampling site. These models may be trained on datasets specific to their respective sites, taking into account the typical microbial compositions and health implications associated with each location. For example, the first model may be trained on data collected from subgingival plaque samples, focusing on microbial profiles associated with periodontal health. This model may be optimized to detect patterns indicative of gingivitis or periodontitis. The training data for this model may include historical samples from subgingival sites, along with corresponding clinical assessments of periodontal health. The second model may be trained on data from supragingival plaque samples or saliva, which may be more relevant for assessing overall oral hygiene or the risk of dental caries. The training dataset for this model may include samples from these sites, paired with clinical evaluations of dental health and caries risk.

[00332] The third model, which analyzes the outputs of the first two models to determine a comprehensive oral health score, may be trained on a more diverse dataset that includes combined information from multiple oral sites. This model may leam to integrate and w eigh the importance of different site-specific indicators to provide an assessment of oral health. The training data for the third model may include historical cases where both site-specific and overall oral health assessments w ere performed, allowing the model to leam the complex relationships between localized and systemic oral health indicators.

[00333] In some implementations, the models may use different machine-learning algorithms suited to their specific tasks. For instance, the site-specific models may employ random forest algorithms to handle the high-dimensional nature of microbiome data, while the third model may use a neural netw ork to integrate the complex inputs from the first tw o models. In some implementations, all three models may be tree-based models or neural network models.

[00334] Fig. 29 is a schematic of an example system 2900 for multi-site biome sampling and analysis. A system for multi-site biome sampling and analysis may include various components configured to collect 2902, analyze 2904, interpret 2906, and report 2908. System 2900 may include one or more collection devices 2910. In one example, first and second collection devices may be configured for gathering biological samples from different sites. These devices may be designed to collect samples from specific areas of interest, such as subgingival and supragingival regions, from different tooth surfaces, and the like.

[00335] System 2900 may include a processing module 2914 to extract biological material from the collected samples. This module may employ various techniques such as centrifugation, filtration, or chemical lysis to isolate bacteria from the samples. System 2900 may include an analysis unit 291 configured to identify and quantify microbial species present in the isolated bacterial material from both the first and second sites. This unit may utilize techniques such as DNA sequencing, PCR, or mass spectrometry to characterize the microbial communities. System 2900 may include a data recording module 2918 configured to record the relative abundance of the identified microbial species at both the first and second sites. This module may generate detailed profiles of the microbial composition for each sample site.

[00336] System 2900 may include a model module 2920 configured to correlate the recorded relative abundance data with oral health scores. This module may employ a plurality of models. Models may be configured to generate one or more correlations 2930 and/or scores 2932. In one example, three models may be used. A first model may be configured to correlate the relative abundance at the first site with a first oral health score. A second model may be configured to correlate the relative abundance at the second site with a second oral health score. A third model may be configured to analyze the first and second oral health scores and determine a third oral health score.

[00337] System 2900 may include a treatment identification unit 2922 and may be configured to suggest one or more treatment regimens based on the third oral health score. This unit may draw upon a database of treatment options and clinical guidelines to provide tailored recommendations. System 2900 may include a user interface 2924 that provides a platform for users to access the identified treatment regimens.

[00338] System 2900 may include a training unit 2926 configured to train one or more models. In one example, the training unit 2926 may be used to train the three distinct models used in the model module 2920. In embodiments, the training unit 2926 may utilize different training data and methodologies for each model to optimize their performance for specific tasks. For the first model, which correlates relative abundance at the first site with a first oral health score, the training unit may use data specifically collected from that site. This training data may include historical samples from the first site, such as subgingival plaque samples, along with corresponding clinical assessments of oral health. The second model, correlating relative abundance at the second site with a second oral health score, may be trained using a separate dataset. This training data may comprise samples from the second site, which could be supragingival plaque or saliva samples, paired with relevant clinical evaluations. The training unit may use algorithms similar to the first model but optimized for the specific characteristics of the second site's microbiome data. For the third model, the training data may include cases where both site-specific and overall oral health assessments were performed, allowing the model to leam complex relationships between localized and systemic oral health indicators. The training unit may implement different machine learning techniques for this model, such as neural networks, to effectively integrate and weigh the importance of various health indicators. The training unit may continuously refine these models using new' data inputs and feedback from clinical outcomes.

[00339] In some implementations, multi-site biome sampling and analysis may extend to collecting biological samples from more than two sites within the oral cavity of a subject. Each additional site may be selected based on specific diagnostic interests or to provide a more comprehensive analysis of the oral microbiome across different oral environments. For each additional site sampled, a corresponding model may be employed to analyze the relative abundance of microbial species at that site. These models, like the first and second models, may be tailored to the unique microbial characteristics of each additional sampling site. The models may be trained on datasets specific to their respective sites, incorporating data that reflect the ty pical microbial compositions and health implications associated with each location. A final model may be utilized to selectively integrate the results from the models associated with each site. This final model may be designed to weigh the outputs of the individual site-specific models according to their relevance to the overall oral health assessment or specific oral health conditions being investigated.

[00340] Fig. 30 is a flowchart of an example method 3000 for diagnosis using characteristics of biome progression. In some implementations, a post-dental implant treatment method may involve a multi-stage approach to assess and manage oral health. This method may utilize sequential sampling and Al analysis to monitor changes in the oral microbiome over time and determine appropriate treatment strategies. At step 3002, method 3000 may include collecting a first set of biological samples from the oral cavity of a subject at an initial time point. These samples may be processed to generate a first biome dataset. In some cases, the first samples may serve as a baseline representation of the subject's oral microbiome composition. In some cases, the first samples may' be collected when a health issue is suspected, and the sample may already represent a disease state.

[00341] At step 3004, at a subsequent time point, a second set of biological samples may be collected from the same subject's oral cavity. This second sampling may occur after a predetermined interval or in response to specific clinical indicators. The samples from this second collection may be processed to create a second biome dataset. At step 3006, method 3000 then employs an Al model to analyze the changes in relative abundance of bacteria between the first and second biome datasets. In one example, this Al model may utilize a weighted tree data structure to process the microbiome data. By traversing this tree structure, the model may evaluate various features and patterns in the data to generate a risk score. Based on this risk score, the method may determine the subject's risk level for oral diseases. This risk assessment may take into account not only the magnitude of changes in bacterial abundance but also the specific types of bacteria involved and their known associations with oral health issues. At step 3008, method 3000 may include determining a first treatment plan for the subject. This treatment plan may be tailored based on multiple factors, including the observed changes in relative abundance of bacteria between the two datasets, the calculated risk level for oral disease, and/or the specific relative abundance of bacteria in the second biome dataset.

[00342] In some implementations, the processing of the first and second biological samples may include the analysis of different bacteria. This differentiation in bacterial analysis may be strategically employed to capture a broader spectrum of microbial diversity and dynamics within the oral microbiome. For instance, the first sample may be analyzed primarily for common oral bacteria known to influence general oral health, while the second sample may focus on specific pathogens that are particularly relevant to post-dental implant complications such as peri-implantitis. By employing distinct bacterial analyses for different samples, the method may enhance the sensitivity⁷ and specificity of the microbiome assessment, thereby improving the accuracy of disease risk evaluation and the effectiveness of the subsequent treatment plans. This approach allows for a more tailored analysis that can adapt to the evolving microbial landscape in the oral cavity, especially in response to changes induced by dental implants or other dental treatments.

[00343] In embodiments, the treatment plan may be modified based on differences and/or similarities of the biome composition of the two samples. In one example, the method may- incorporate a feedback loop that continuously updates and refines the treatment recommendations based on the ongoing analysis of microbial samples. In some cases, the method may detect changes in the relative abundance of specific bacterial species that are associated with an increased risk of oral diseases. When such changes are observed, the treatment plan may be modified to include more aggressive interventions or preventive measures. For example, if an increase in pathogenic bacteria is detected, the system may recommend additional antibiotic treatments or more frequent professional cleanings. The method may also analyze the effectiveness of current treatments by comparing microbial profiles before and after interventions. If the analysis indicates that a particular treatment is not producing the desired changes in the oral microbiome, the system may suggest alternative therapies or modifications to the existing regimen. This may include adjusting the dosage of medications, changing the frequency of treatments, or recommending different types of interventions altogether. In some implementations, the method may consider the rate of change in microbial compositions over time. If rapid shifts in the microbiome are detected, the treatment plan may be modified to include more frequent monitoring or more intensive interventions. Conversely, if the microbiome remains stable or shows gradual improvement, the method may recommend a maintenance approach with less frequent interventions.

[00344] In some implementations, the method may also take into account the specific microbial signatures associated with different oral health conditions. For instance, if the analysis reveals a microbial profile indicative of early-stage periodontitis, the treatment plan may be modified to include targeted therapies for preventing disease progression.

[00345] In some aspects, the method may incorporate machine learning algorithms that can predict future changes in the oral microbiome based on current trends and historical data. These predictions may be used to proactively modify the treatment plan, potentially preventing the onset of oral diseases before they occur.

[00346] In some implementations, the method may analyze the interactions between different bacterial species within the oral microbiome. If certain beneficial bacteria are found to be declining, the treatment plan may be modified to include probiotics or other interventions aimed at restoring a healthy microbial balance. The method may also consider the presence of antibiotic-resistant bacteria when modifying treatment plans. If resistant strains are detected, the system may recommend alternative treatment strategies that do not rely solely on antibiotics, such as bacteriophage therapy or immunomodulatory approaches.

[00347] Method 3000 for diagnosis using characteristics of biome progression may be implemented in a system such as the system shown and described with respect to Fig. 29. A first module (such as 2910) of the system may be configured to collect an initial set of biological samples from the oral cavity of a subject at a specific time point. After collection, the module may process these samples to generate a first biome dataset. A second module may be designed to perform a similar function at a later time point. This module may collect a second set of biological samples from the same subject's oral cavity and generate a second biome dataset. The timing of this second sample collection may be predetermined or triggered by specific clinical indicators. An analysis module (such as 2916) may be configured to analyze the changes in the relative abundance of bacteria between the first and second biome datasets. By traversing the weighted tree structure, the Al model may process microbiome data, evaluating various features and patterns to generate a risk score. This score may be used to identify the subject's risk level for oral diseases. A treatment module (such as 2922) may be configured to determine a treatment plan based on the analysis results. This module may take into account multiple factors, including the changes in bacterial abundance between the two datasets, the calculated risk level for oral disease, and the specific relative abundance of bacteria in the second biome dataset.

[00348] Fig. 31 is a flowchart of an example method 3100 for processing biome data and generating personalized user reports. In embodiments, systems and methods for secure and confidential handling of data may be desirable. In some aspects, the method for secure data handling may involve creating secure data pipelines that isolate sensitive information at various stages of processing. This approach may allow for compartmentalized access.

[00349] Method 3100 may begin with step 3102 of receiving raw biome data collected from biological samples. This raw data may contain information about the microbial composition of the subject's oral cavity. In some cases, the raw data may be anonymized with random identifiers. At step 3104, the raw biome data may undergo cleaning and normalization procedures. This step may involve removing artifacts, such as sequencing errors or contamination, and standardizing the data format to ensure consistency across all samples. The normalized data may then be stored in a centralized database, creating a repository of preprocessed biome information. At step 3106. feature extraction may be performed on the preprocessed biome data to identify key characteristics or patterns within the microbial community. These features may include relative abundances of specific bacterial species, diversify indices, or functional gene profiles. At step 3108 the extracted features may then be analyzed using one or more machine learning models. These models may be trained to recognize patterns associated with various oral health conditions or microbial imbalances. The output of this analysis may be a set of biome classifications, which may categorize the sample based on its microbial composition and potential health implications (step 3110). At step 3112, when a request for biome analysis is received for a specific subject, the system may retrieve relevant biome data and classifications from the centralized database. This may include historical data, if available, allowing for longitudinal analysis of the subject's oral microbiome. At step 3114, in addition to biome data, the system may also retrieve the personal data of the subject. This personal data may encompass demographic information, medical history, lifestyle factors, or dental care habits, providing context for the biome analysis. At step 3106, using the retrieved biome data, classifications, and personal data, the system may generate a personalized biome report for the subject. This report may include insights into the subject's oral microbiome composition, potential health implications, and personalized recommendations for maintaining or improving oral health. The personalized biome report may be transmitted to a user device, allowing the subject or healthcare provider to access the information. [00350] In embodiments, a report may be generated for a subject in real time and/or in response to a report request. The report may be regenerated each time it is requested allowing the use of the latest version of the Al model. This capability ensures that the most current and refined algorithms are applied to the analysis of the biome data, providing up-to-date assessments and recommendations. Each report generated may reflect the latest insights and data collection.

[00351]

[00352] In embodiments, the report may include a graphical summary that presents data in a clear, visually appealing manner. The graphical summary may incorporate various visual elements such as graphs, scales, and scores to convey information. The graphical summary' may feature charts and graphs that illustrate changes in microbial composition over time. These may include line graphs showing trends in relative abundance of bacterial species, bar charts comparing different bacterial groups, or pie charts representing the overall microbial community structure. Scales may be used to depict risk levels or health scores. For example, a color-coded slider scale may represent the subject's current oral health status, ranging from "Healthy" to "At Risk" to "Disease State." This visual representation may allow users to quickly grasp their current health situation. Numerical scores may be prominently displayed, such as an overall oral health score or specific risk scores for different conditions. These scores may be accompanied by brief explanations of their significance. The graphical summary' may also include visual comparisons between the subject's microbial profile and reference profiles for healthy individuals or those with specific oral health conditions. This comparison may be presented using overlaid charts or side-by-side visualizations.

[00353] In embodiments, interactive elements may be incorporated into the graphical summary, allowing users to hover over or click on different parts of the visualization to reveal more detailed information. This interactivity may provide a layered approach to information presentation, accommodating both quick overview and in-depth exploration. In some implementations, the graphical summary' may include a timeline view, illustrating how the subject's oral micro biome and associated health metrics have changed over multiple sampling points. This longitudinal perspective may help in visualizing the effectiveness of treatments or the progression of oral health conditions. The layout of the graphical summary may be designed to guide the user through the results and highlight key findings and recommendations.

[00354] Fig. 32 illustrates one example of a part of an output of a generated report that may be provided to a user. The example includes a graphical summary' that includes a graphical scale showing a disease state classification component, a disease progression risk score component, and relevant explanations for the elements. The disease state classification component includes a visual scale representing different states. In one example, the states include a scale from Healthy to Mucositis to Implantitis, with a marker indicating the current state as Mucositis. The disease progression risk score component displays a numerical score that reflects the risk score (for example, a score of 67/100 is shown in Fig. 32, categorized as moderate risk). The system provides an interpretation of the bacterial composition of the sample, indicating an infection and provides guidance on the risk of disease progression and recommends discussing intervention options with a clinician.

[00355] In some implementations, the risk score shown in Fig. 32 may be computed using the techniques described herein such as using a trained Al model. The risk score may be derived from a combination of factors that contribute to the overall assessment of a patient's oral health status and the likelihood of disease progression.

[00356] In one example, the risk score may be computed using a weighted algorithm that aggregates multiple factors into a patient risk score. The algorithm may consider one or more components for generating risk score. A component of the risk score may be based on a microbial profile score. The Microbial Profile Score may be a composite of factors or scores and may include scores such as early colonizer relative abundance, bridgers relative abundance, late pathogens relative abundance, community diversity, and/or machine-learning derived progression risk score. A component of the risk score may be based on a patient medical history score. A patient medical history score may be a composite of factors or scores and may include scores such as diabetes status, smoking history', and/or periodontal history. A component of the risk score may be based on clinical measurements. Clinical measurements may be a composite of factors or scores and may include scores such as bleeding on probing, probing depth, oral hygiene status, local gingival inflammation, and/or implant age. A component of the risk score may be based on an implant characteristics score. An implant characteristics score may be a composite of factors or scores and may include scores such as implant location (e.g., higher risk for posterior locations), and/or implant material. The system may assign specific weights to each of these components based on their relative importance in determining the overall risk. Within each component, individual factors may be assigned sub-weights to reflect their contribution to that particular aspect of the risk assessment.

[00357] In some implementations, the risk score may also take into account additional factors that may influence the likelihood of peri-implant disease development or progression. These factors may include a history of periodontitis, diabetes history, smoking or tobacco use. The risk assessment may also consider immunocompromised status, which may result from medication use or systemic conditions, potentially affecting the body's ability to fight infections. Osteoporosis may be considered, as it can impact bone density and healing around implants. Poor oral hygiene practices and inadequate oral hygiene maintenance may be evaluated, as they can contribute to bacterial accumulation and inflammation. Other factors that may be incorporated into the risk assessment include a lack of keratinized mucosa around the implant, implant-specific factors such as surface roughness, malpositioning, poor design or fit of prosthetics, microgap at the implant-abutment interface, and cement remnants may also be considered. The risk score calculation may take into account factors related to the implant surgery⁷ itself, such as traumatic surgical technique, insufficient primary⁷ stability, contamination of the implant surface during surgery⁷, and inadequate bone volume or quality at the implant site. Additionally, the risk score may consider the patient's adherence to regular follow-up appointments and professional maintenance, as well as the presence of soft tissue defects or inadequate soft tissue coverage around the implant.

[00358] In implementations, each of the factors contributing to the risk score may be scored individually. These individual scores may then be normalized to ensure comparability across different types of factors. The normalized scores may be combined into a comprehensive risk score using weighting methods. The system may adjust the weights assigned to different factors based on their relative importance, which may be determined through statistical analysis of historical data or expert clinical knowledge.

[00359] Fig. 33 illustrates elements of an example user interface for a risk assessment which may be part of a report. The interface comprises several sections for user input and data visualization. In some implementations, the graphs may show how a user's risk of disease may change under new scenarios, such as when new risk factors are introduced. The system may allow users to interactively add or remove risk factors, visualizing in real-time how these changes affect their disease risk profile over time. For example, a user may add a risk factor like "smoking" or "diabetes" and observe how the risk curve shifts in response. This feature may help users understand the potential impact of lifesty le changes or medical conditions on their oral health. The interface may also enable the simultaneous comparison of multiple scenarios. Users may create and save different risk profiles, each representing a unique combination of risk factors. These profiles may be overlaid on the same graph, allowing for easy' comparison of how various factors or combinations of factors influence disease risk trajectories. This comparative view may be particularly useful for healthcare providers in explaining potential outcomes to patients and in developing personalized prevention or treatment strategies. In addition, the graphs may incorporate predictive modeling to project future risk based on current trends and potential interventions. For instance, the system may illustrate how adopting certain preventive measures or undergoing specific treatments could potentially alter the user's risk curve over time. This predictive capability⁷ may help in motivating patients to adhere to recommended oral health practices by visually demonstrating their long-term benefits. [00360] Fig. 34 illustrates elements of an example user interface intervention guidelines that may be part of a report. The figure illustrates an implant site-specific interventional guidance. The interface provides a recommended interventional strategy based on microbial relative abundance and disease risk profile. The diagram may include two main sections and may include a selection considerations section and an implant site specific care plan section.

[00361] The selection considerations section may include an intervention objective component and an intervention complexity component. The interv ention objective component may include a horizontal scale ranging from prevention to reduction that indicates the objective of a recommended intervention. In the example, the recommended intervention is marked with a triangle on the scale and is indicated as "Inflammation Reduction". The intervention complexity component includes a horizontal scale that ranges from Low to High and indicates the complexity of the recommended intervention. In the example, the complexity of the recommended intervention is marked with a triangle on the scale and is marked as "Moderate Complexity."

[00362] The implant site specific care plan section outlines the treatment approach. In this embodiment, it lists the Primary Goal as "Microbial Reduction" and specifies the Approach as "Non- Surgical." The plan indicates that no antibiotics are recommended, and a "Heightened" self-care routine is advised. The Retest Timing is set at 3-Months. The report includes context for the recommendations .

[00363] In some implementations, intervention recommendations may be determined using approaches that integrate multiple dimensions of data and disease characterization. These approaches may include a step of comprehensive disease state characterization. For example, the characterization of peri-implant mucositis may include defining the condition as a reversible inflammatory state limited to soft tissues surrounding an implant. It may be defined as analogous to gingivitis in natural dentition. The definition may include that oral microbiome bacterial relative abundance signatures, determined via methods described herein, may indicate an "infection" and/or a peri-implant mucositis disease state. The characterization may define that a treatment goal may focus on reducing inflammation and microbial load without invasive procedures. In some cases, molecular and cellular markers of inflammation may be considered. In another example, for peri-implantitis the characterization may define peri-implantitis as an advanced, destructive condition involving inflammation and bone loss around the implant and that it may require more intensive treatment modalities. The definition may include that oral microbiome bacterial relative abundance signatures, determined via methods described herein, may indicate an "infection" and a peri-implantitis disease state classification. The characterization may include that treatment approaches may involve a combination of mechanical, antimicrobial, and surgical interventions. In some implementations, a quantitative assessment of bone loss using advanced imaging techniques may be included.

[00364] In the next step, a holistic approach to data collection and analysis may be implemented. For example, for microbiome analysis the collection may include high-throughput sequencing of the implant site microbiome, quantification of relative abundance of bacterial species, and functional analysis of microbial communities. In another example, for host factor analysis data collection may include consideration/collection of genetic susceptibility markers, immune response profiling, and systemic health indicators. In another collection may include collection of implant-specific factors. Factors such as implant surface characteristics, biomechanical stress analysis, time since implant placement, and treatment history may be taken into account/collected. In another example, collection may include collection/consideration of clinical parameters, such as probing depth, bleeding on probing, suppuration, and radiographic bone loss.

[00365] The next step of the method for determining intervention recommendations may include the implementation of advanced machine learning algorithms for data analysis and intervention assignment. In one example, the step may include supervised learning for models. Models may be trained on historical data of successful treatments. The system may use this training to predict treatment outcomes based on input parameters. The model may analyze factors such as microbial composition, patient demographics, and implant characteristics to suggest the most effective intervention strategy. In another example, the step may include unsupervised learning techniques. Clustering algorithms may be employed to group similar cases and identify patterns that may not be immediately apparent. This approach may provide for the discovery of novel associations between microbiome profiles and treatment outcomes. The method may uncover previously unknown relationships between specific bacterial communities and the success rates of certain interventions. In another example, the step may include training deep learning networks. Neural networks may be utilized to analyze complex, non-linear relationships betw een multiple factors. Deep learning netw orks may predict optimal intervention strategies that take into account a wide range of variables. [00366] The next step of the method for determining intervention recommendations may include dynamic evidence synthesis. In one example, the step may include automated literature mining. Natural language processing techniques may be employed for real-time analysis of new publications and provide for continuous updating of the evidence base, ensuring that the system's recommendations are always informed by the latest research findings. In another example, the step may include real-world evidence integration where electronic health records are analyzed to incorporate real-world treatment outcomes as well as the integration of post-market surveillance data that provides insights into the long-term effectiveness and safety of various interventions. In another example, the step may include Bayesian network analysis. This probabilistic approach may be used to provide dynamic updating of treatment efficacy probabilities. The system may consider evolving resistance patterns, adjusting its recommendations based on the latest data on antibiotic resistance and treatment effectiveness.

[00367] The next step of the method for determining intervention recommendations may include personalized risk assessment (or reference patient risk of progression scoring). In one example, the step may include multi-factorial risk calculation microbiome, host, and implant-specific factors are used to generate a comprehensive risk profile. Machine learning insights may be used to weight risk factors. The calculation may account for historical treatments and outcomes, providing a more accurate prediction of future risk. In another example, the step may include temporal risk proj ection. The method may predict disease progression rates, allowing for proactive intervention and may include projections of how the risk profile may change over time with or without intervention. [00368] The next step of the method for determining intervention recommendations may include adaptive intervention recommendation. In one example, the step may include multi-criteria decision analysis. The analysis may consider multiple criteria when recommending interventions. This may include evaluating the efficacy of treatments based on the unique microbial signature measured at the implant site. The efficacy assessment may be adjusted to account for confounding risk factors specific to the patient. Cost-effectiveness may also be factored into the decision-making process, allowing for the recommendation of treatments that provide the best value. Additionally, the system may take into account patient preferences, which may be gathered through questionnaires or previous treatment history. In another example, the step may include treatment sequence optimization. This optimization may consider the potential synergistic effects of combining different treatment modalities. The recommendation engine may adaptively update its suggestions based on the patent's response to previous treatments. For example, if a patient shows limited improvement with a conservative approach, the system may suggest escalating to more intensive interventions. In another example, the step may include a precision intervention approach where interventions may be tailored to individual patient profiles, taking into account factors such as age, overall health status, and lifestyle. The method may consider potential drug interactions and contraindications when recommending treatments. This may involve cross-referencing the patient's current medications with a comprehensive drug interaction database. The precision approach may also factor in the patient's genetic predisposition to certain conditions or their likelihood of responding to specific treatments. [00369] The next step of the method for determining intervention recommendations may include continuous learning and refinement with a feedback loop for continuous improvement. In one example, the step may include treatment outcome tracking. The step may include data collection from regular follow-up assessments of the patient's oral health status, including microbiome analysis and clinical examinations. The system may identify factors associated with treatment success or failure by analyzing patterns in patient outcomes. This information may be used to refine future treatment recommendations and improve predictive models. In another example, the step may include model retraining. Machine learning models within the system may undergo periodic retraining with new data. This process may incorporate newly acquired patient data, treatment outcomes, and emerging research findings. The decision algorithms may be adjusted based on emerging patterns identified during the retraining process. In another example, the step may include explainable integration for executing machine learning techniques to provide transparency. This approach may allow clinicians to understand the rationale behind intervention recommendations. The system may generate explanations for its recommendations, highlighting the key factors that influenced the decision.

[00370] Fig. 35 depicts a table showing the effectiveness of various antibiotics against different bacterial species that may be part of a report. The rows of the table list different bacterial species. The columns represent different antibiotics or antibiotic combinations. Each cell in the table contains a number from 2 to 5, indicating the level of effectiveness of the antibiotic against the corresponding bacterial species (with 5 being the most effective). The color coding of the cells ranges from light to dark, with darker shades representing higher effectiveness. This table provides a comprehensive overview of antibiotic susceptibility for various oral bacteria species. The values in the table may be used to determine an antibiotic suitability index. The index may then be aggregated to provide an overall assessment of the bacterial population's antibiotic susceptibility. In some implementations, the system may calculate a weighted average of the effectiveness scores for each antibiotic, taking into account the relative abundance of each bacterial species in the patient's sample. This weighted average may serve as the antibiotic suitability index for that particular antibiotic in the context of the patient's specific oral microbiome composition.

[00371] The table may, in some instances, be provided to a user as part of the graphical report. This visual representation may allow healthcare providers to see the underlying data on which the antibiotic suitability index was computed. The color-coding and numerical values in the table may offer an intuitive w ay for clinicians to quickly assess which antibiotics might be most effective against the specific bacterial community present in their patient's oral cavity. In some cases, the system may use this data to generate recommendations for antibiotic treatment, suggesting combinations of antibiotics that may be most effective based on the patient's unique oral microbiome profile. These recommendations may be presented alongside the table, providing clinicians with both the raw data and interpreted results to inform their treatment decisions. [00372] In some implementations, the approach to rating antibiotic efficacy against bacterial genera may include a multi-step analysis process. This process may begin with a comprehensive analysis of the target bacterial genera. The analysis may include identification of common characteristics for each genus. This may involve determining Gram-stain properties, which could be positive or negative. The oxygen relationships of the bacteria may also be assessed, categorizing them as aerobic, anaerobic, or facultative anaerobe. Additionally, the typical habitats of the bacteria may be identified, such as the oral cavity or gastrointestinal tract. An assessment of clinical relevance maybe conducted as part of the analysis. This may involve evaluating each genus's association with peri- implantitis and considering the genus's role in other oral infections. The process may also include identifying genera commonly linked to specific infection types. Genomic profiling may be performed as part of the bacterial genera analysis. This may involve analyzing whole genome sequences for each genus and identifying generic markers associated with antibiotic resistance. [00373] Rating antibiotic efficacy may also include a review of the antibiotic spectrum of activity. This may involve examining each antibiotic's known spectrum of activity and analyzing its pharmacological properties. The process may also integrate pharmacokinetic and pharmacodynamic (PK/PD) data to model antibiotic behavior in vivo. In embodiments, a scoring matrix that accounts for predicted antibiotic efficacy, resistance risk, and host factor influence may be generated, weighted scores for each dimension may be assigned and used to determine the Antibiotic Suitability- Index for each antibiotic-genus combination.

[00374] In some implementations, the efficacy may also be computed using additional approaches. In one example, machine learning and/or enhanced empirical data evaluation may be incorporated into the process. This approach may use machine learning algorithms to comprehensively evaluate empirical data. The process may include systematic review and data extraction from various sources such as clinical studies, systematic reviews, meta-analyses, and anonymized electronic health records. Machine learning algorithms, which may include random forests or support vector machines, may be applied to analyze large datasets of antibiotic sensitivity patterns, identify complex relationships between bacterial characteristics and antibiotic efficacy, and predict antibiotic sensitivity based on multiple factors. The model may be continuously refined through active learning techniques, incorporating new clinical data as it becomes available. In another example, a genomics- informed resistance pattern assessment may be conducted, integrating genomic data. This may involve an in-depth assessment of resistance patterns, including evaluation of resistance trends for each genus and antibiotic. The assessment may include analysis of genetic markers associated with antibiotic resistance, such as identification of known resistance genes and prediction of potential resistance based on genomic similarities. The potential for horizontal gene transfer within the oral microbiome may also be assessed.

[00375] In another example, host factor integration may be incorporated into the sensitivity estimation. This approach may involve analysis of host immune status and its impact on antibiotic efficacy, consideration of site-specific microbiome composition, and evaluation of host-specific factors that may influence antibiotic effectiveness. These factors may include pH and oxygen levels in specific oral niches.

[00376] In another example, a temporal sensitivity analysis may be incorporated into the sensitivity analysis to track changes in antibiotic sensitivity over time. This approach may involve analyzing historical antibiotic sensitivity data to identify trends and patterns in resistance development. By examining data from multiple time points, the system may detect shifts in bacterial susceptibility to specific antibiotics or antibiotic classes. This temporal analysis may reveal emerging resistance patterns or the re-emergence of sensitivity to previously ineffective antibiotics. The systems and methods may utilize statistical methods and machine learning algorithms to analyze these temporal trends. Time series analysis techniques may be applied to identity’ cyclical patterns, seasonal variations, or long-term trends in antibiotic sensitivity. In some cases, the system may employ change point detection algorithms to identity significant shifts in sensitivity patterns over time. Based on the identified trends and patterns, the system may develop predictive models to forecast future changes in antibiotic sensitivity. These models may take into account current antibiotic use patterns, known mechanisms of resistance development, and other relevant factors. The predictive capabilities may allow healthcare providers to anticipate potential changes in antibiotic efficacy and adjust treatment strategies proactively.

[00377] In yet another example, the system may integrate clinical guideline information and establish a feedback loop to continuously refine and update its antibiotic efficacy assessments. This process may begin with a comprehensive review of guidelines from authoritative bodies such as professional dental associations, infectious disease societies, and public health organizations. The system may extract relevant recommendations and best practices related to antibiotic use in dental implant procedures and peri-implant disease management. The antibiotic efficacy data generated by the system may be cross-referenced with these clinical guidelines to ensure alignment with recommended practices. This comparison may help identity' any discrepancies between the system's assessments and established clinical recommendations. In cases where discrepancies are found, the system may flag these instances for further review by healthcare professionals. A feedback mechanism may be implemented to incorporate real-world treatment outcomes into the model. This may involve collecting data on the effectiveness of prescribed antibiotic regimens, including information on treatment success rates, adverse effects, and any instances of antibiotic resistance encountered. Healthcare providers may input this information into the system, allowing it to learn from actual clinical outcomes and refine its efficacy predictions. The feedback loop may also include periodic updates based on new research findings, updated clinical guidelines, and emerging best practices in antibiotic stewardship. This continuous learning process may help ensure that the system's antibiotic efficacy assessments remain current and aligned with the latest clinical evidence and recommendations.

[00378] The methods and systems described herein may be deployed in part or in whole through a machine that executes computer software on a server, client, firewall, gateway, hub, router, or other such computer and/or networking hardware. The software program may be associated with a server that may include a file server, print server, domain server, internet server, intranet server and other variants such as secondary server, host server, distributed server, and the like. The server may include one or more of memories, processors, computer readable media, storage media, ports (physical and virtual), communication devices, and interfaces capable of accessing other servers, clients, machines, and devices through a wired or a wireless medium, and the like. The methods, programs, or codes as described herein and elsewhere may be executed by the server. In addition, other devices required for execution of methods as described in this application may be considered as a part of the infrastructure associated with the server.

[00379] The server may provide an interface to other devices including, without limitation, clients, other servers, printers, database servers, print servers, file servers, communication servers. distributed servers, and the like. Additionally, this coupling and/or connection may facilitate remote execution of program across the network. The networking of some or all of these devices may facilitate parallel processing of a program or method at one or more locations without deviating from the scope of the disclosure. In addition, any of the devices attached to the server through an interface may include at least one storage medium capable of storing methods, programs, code, and/or instructions. A central repository may provide program instructions to be executed on different devices. In this implementation, the remote repository may act as a storage medium for program code, instructions, and programs.

[00380] The software program may be associated with a client that may include a file client, print client, domain client, internet client, intranet client and other variants such as secondary client, host client, distributed client, and the tike. The client may include one or more of memories, processors, computer readable media, storage media, ports (physical and virtual), communication devices, and interfaces capable of accessing other clients, servers, machines, and devices through a wired or a wireless medium, and the like. The methods, programs, or codes as described herein and elsewhere may be executed by the client. In addition, other devices required for execution of methods as described in this application may be considered as a part of the infrastructure associated with the client.

[00381] The client may provide an interface to other devices including, without limitation, servers, other clients, printers, database servers, print servers, file servers, communication servers, distributed servers, and the like. Additionally, this coupling and/or connection may facilitate remote execution of program across the network. The networking of some or all of these devices may facilitate parallel processing of a program or method at one or more locations without deviating from the scope of the disclosure. In addition, any of the devices attached to the client through an interface may include at least one storage medium capable of storing methods, programs, applications, code, and/or instructions. A central repository may provide program instructions to be executed on different devices. In this implementation, the remote repository’ may act as a storage medium for program code, instructions, and programs.

[00382] The methods and sy stems described herein may be deployed in part or in whole through network infrastructures. The network infrastructure may include elements such as computing devices, servers, routers, hubs, firewalls, clients, personal computers, communication devices, routing devices and other active and passive devices, modules and/or components as known in the art. The computing and/or non-computing device(s) associated with the network infrastructure may include, apart from other components, a storage medium such as flash memory, buffer, stack, RAM. ROM, and the like. The processes, methods, program codes, instructions described herein and elsewhere may be executed by one or more of the network infrastructural elements.

[00383] The methods, program codes, and instructions described herein and elsewhere may be implemented on a cellular network having multiple cells. The cellular network may either be frequency division multiple access (FDMA) network or code division multiple access (CDMA) network. The cellular network may include mobile devices, cell sites, base stations, repeaters, antennas, towers, and the like. The cell network may be a GSM, GPRS, 3G, EVDO, mesh, or other network types.

[00384] The methods, programs codes, and instructions described herein and elsewhere may be implemented on or through mobile devices. The mobile devices may include navigation devices, cell phones, mobile phones, mobile personal digital assistants, laptops, palmtops, netbooks, pagers, electronic books readers, music players and the like. These devices may include, apart from other components, a storage medium such as a flash memory, buffer, RAM, ROM and one or more computing devices. The computing devices associated with mobile devices may be enabled to execute program codes, methods, and instructions stored thereon. Alternatively, the mobile devices may be configured to execute instructions in collaboration with other devices. The mobile devices may communicate with base stations interfaced with servers and configured to execute program codes. The mobile devices may communicate on a peer-to-peer network, mesh network, or other communications network. The program code may be stored on the storage medium associated with the server and executed by a computing device embedded within the server. The base station may include a computing device and a storage medium. The storage device may store program codes and instructions executed by the computing devices associated with the base station.

[00385] The computer software, program codes, and/or instructions may be stored and/or accessed on machine readable media that may include: computer components, devices, and recording media that retain digital data used for computing for some interval of time; semiconductor storage known as random access memory (RAM); mass storage typically for more permanent storage, such as optical discs, forms of magnetic storage like hard disks, tapes, drums, cards and other types; processor registers, cache memory, volatile memory, non-volatile memory; optical storage such as CD, DVD; removable media such as flash memory (e.g. USB sticks or keys), floppy disks, magnetic tape, paper tape, punch cards, standalone RAM disks. Zip drives, removable mass storage, off-line, and the like; other computer memory such as dynamic memory, static memory, read/write storage, mutable storage, read only, random access, sequential access, location addressable, file addressable, content addressable, network attached storage, storage area network, bar codes, magnetic ink, and the like. [00386] The methods and systems described herein may transform physical and/or or intangible items from one state to another. The methods and systems described herein may also transform data representing physical and/or intangible items from one state to another.

[00387] The elements described and depicted herein, including in flow charts and block diagrams throughout the figures, imply logical boundaries between the elements. However, according to software or hardware engineering practices, the depicted elements and the functions thereof may be implemented on machines through computer executable media having a processor capable of executing program instructions stored thereon as a monolithic software structure, as standalone software modules, or as modules that employ external routines, code, sen ices, and so forth, or any combination of these, and all such implementations may be within the scope of the present disclosure. Examples of such machines may include, but may not be limited to, personal digital assistants, laptops, personal computers, mobile phones, other handheld computing devices, medical equipment, wired or wireless communication devices, transducers, chips, calculators, satellites, tablet PCs, electronic books, gadgets, electronic devices, devices having artificial intelligence, computing devices, networking equipment, servers, routers and the like. Furthermore, the elements depicted in the flow chart and block diagrams, or any other logical component may be implemented on a machine capable of executing program instructions. Thus, while the foregoing drawings and descriptions set forth functional aspects of the disclosed systems, no particular arrangement of software for implementing these functional aspects should be inferred from these descriptions unless explicitly stated or otherwise clear from the context. Similarly, it will be appreciated that the various steps identified and described above may be varied, and that the order of steps may be adapted to particular applications of the techniques disclosed herein. All such variations and modifications are intended to fall within the scope of this disclosure. As such, the depiction and/or description of an order for various steps should not be understood to require a particular order of execution for those steps, unless required by a particular application, or explicitly stated or otherwise clear from the context.

[00388] The methods and/or processes described above, and steps thereof, may be realized in hardware, software or any combination of hardware and software suitable for a particular application. The hardware may include a general-purpose computer and/or dedicated computing device or specific computing device or particular aspect or component of a specific computing device. The processes may be realized in one or more microprocessors, microcontrollers, embedded microcontrollers, programmable digital signal processors or other programmable device, along with internal and/or external memory. The processes may also, or instead, be embodied in an application specific integrated circuit, a programmable gate array, programmable array logic, or any other device or combination of devices that may be configured to process electronic signals. It will further be appreciated that one or more of the processes may be realized as a computer executable code capable of being executed on a machine-readable medium.

[00389] The computer executable code may be created using a structured programming language such as C, an object oriented programming language such as C++, or any other high-level or low- level programming language (including assembly languages, hardware description languages, and database programming languages and technologies) that may be stored, compiled or interpreted to run on one of the above devices, as well as heterogeneous combinations of processors, processor architectures, or combinations of different hardware and software, or any other machine capable of executing program instructions.

[00390] Thus, in one aspect, each method described above, and combinations thereof may be embodied in computer executable code that, when executing on one or more computing devices, performs the steps thereof. In another aspect, the methods may be embodied in systems that perform the steps thereof and may be distributed across devices in a number of ways, or all of the functionalities may be integrated into a dedicated, standalone device or other hardware. In another aspect, the means for performing the steps associated with the processes described above may include any of the hardware and/or software described above. All such permutations and combinations are intended to fall within the scope of the present disclosure.

[00391] While the invention has been disclosed in connection with the preferred embodiments shown and described in detail, various modifications and improvements thereon will become readily apparent to those skilled in the art. Accordingly, the spirit and scope of the present invention is not to be limited by the foregoing examples but is to be understood in the broadest sense allowable by law. [00392] All documents referenced herein are hereby incorporated by reference in their entirety.

Claims

CLAIMS What is claimed is:

1. A post-dental implant treatment method comprising: collecting biological samples from an oral cavity of a subject to provide a biome dataset; identifying the subject as at a first risk of oral disease based on a risk score that is generated from relative abundance data of bacteria in the biome dataset by an Al model that traverses a weighted tree data structure to analyze the relative abundance of bacteria in the biome dataset to provide the risk score; and generating a plan for an antibiotic regimen for the subject at the first risk of oral disease after dental implant surgery, wherein the antibiotic regimen comprises a plurality of antibiotics selected based on the relative abundance of the bacteria and antibiotic resistance of the bacteria.

2. The method of claim 1, wherein the Al model is a random forest model or a decision tree model.

3. The method of claim 1, wherein the oral disease includes at least one of gingivitis, periodontitis, peri-implant mucositis, or peri-implantitis.

4. The method of claim 1, further comprising processing the biological samples to provide the biome dataset by: extracting and purifying DNA from the biological samples; performing DNA sequencing on the purified DNA; and analyzing the DNA sequencing to generate the biome dataset.

5. The method of claim 1, wherein the relative abundance is at a family level.

6. The method of claim 1 , wherein the relative abundance is at a genus level.

7. The method of claim 1, wherein the first risk of oral disease is above a threshold risk score value.

8. The method of claim 1, further comprising administrating the plan for the antibiotic regimen to the subject.

9. The method of claim 1 , further comprising: collecting second biological samples from the oral cavity of the subject to provide a second biome dataset during the antibiotic regimen; identifying the subject as at a second risk of oral disease based on a risk score that is generated from relative abundance data in the second biome dataset by the Al model that traverses the weighted tree data structure to analyze the relative abundance of bacteria in the second biome dataset to provide the risk score; and modifying the plan for the antibiotic regimen.

10. The method of claim 9, wherein modifying the plan for the antibiotic regimen comprises pausing the antibiotic regimen.

1 1. The method of claim 9, wherein the second risk of oral disease is below a threshold risk score value.

12. The method of claim 1, further comprising: collecting third biological samples from the oral cavity of the subject to provide a third biome dataset during the antibiotic regimen; identify ing the subject as at a third risk of oral disease based on the risk score that is generated from relative abundance data in the third biome dataset by the Al model that traverses the weighted tree data structure to analyze the relative abundance of bacteria in the third biome dataset to provide the risk score; and modifying the plan for the antibiotic regimen and implementing a different treatment.

13. The method of claim 12, wherein the third risk of oral disease is above a second threshold risk score value.

14. The method of claim 12, wherein the different treatment comprises at least one treatment selected from: sahva increasing chewing gum, xylitol chewing gum, surgical intervention, gum grafting, implant removal, partial implant removal, chemical debridement, mechanical debridement, acid etchant cleaning, custom tray delivery⁷ of antimicrobial or peroxide medicaments, implant surface polishing, hard tissue grafting, soft tissue grafting, surgical soft tissue resection, or gingival pocket irrigation or disinfection.

15. The method of claim 1 , wherein the relative abundance of bacteria includes a relative abundance measure of early bacteria, bridge bacteria, and late pathogenic bacteria.

16. The method of claim 1 , further comprising: determining a peri-implant disease-associated taxa percentage from the relative abundance of bacteria and wherein the Al model is trained to provide a risk score based on an input that includes the peri-implant disease-associated taxa percentage.

17. The method of claim 1, further comprising: determining an anaerobe score from the relative abundance of bacteria and wherein the Al model is trained to provide a risk score based on an input that includes the anaerobe score.

18. The method of claim 1 , further comprising: determining a gram stain profile from the relative abundance of bacteria and wherein the Al model is trained to provide a risk score based on an input that includes the gram stain profile.

19. The method of claim 1 , further comprising: determining an alpha diversity from the relative abundance of bacteria and wherein the Al model is trained to provide a risk score based on an input that includes the alpha diversity.

20. A treatment method comprising: collecting biological samples from an oral cavity' of a subject to provide a biome dataset; identifying the subject as having a first risk score for a disease based on a risk score that is generated from relative abundance data in the biome dataset by a trained Al model that predicts scores based on training examples; and generating a treatment plan for the subject having the first risk score for disease, wherein the treatment plan is based on the first risk score and the relative abundance data.

21. The method of claim 20. wherein the Al model is a random forest model or a neural network model.

22. The method of claim 20, further comprising administering the treatment plan to the subject.

23. The method of claim 20, wherein the disease includes at least one of an oral disease, cancer, cognitive decline, rheumatoid arthritis, or Alzheimer's.

24. The method of claim 20. further compnsing processing the biological samples to provide the biome dataset by: extracting and purifying DNA from the biological samples; performing DNA sequencing on the purified DNA; and analyzing the DNA sequencing to generate the biome dataset.

25. The method of claim 20, wherein the relative abundance is at a family level.

26. The method of claim 20, wherein the relative abundance is at a genus level.

27. A system for determining a post-dental implant treatment method, comprising: a collection device configured to be introduced into an oral cavity and to collect biological samples from surfaces; a storage unit, containing a storage medium that preserves the collected sample by stabilizing the sample for subsequent analysis; a processing module configured to extract biological material from the collected sample to isolate bacteria; an analysis unit, configured to analyze the isolated bacteria material to identify and quantify microbial species; a data recording module integrated with the analysis unit, configured to record a relative abundance of the microbial species; a model module configured to correlate the recorded relative abundance with an oral health score; a treatment identification unit configured to identify one or more treatment regimens based on the oral health score and the relative abundance of bacteria; and a user interface configured to provide a platform for users to access the recorded relative abundance and the one or more treatment regimens.

28. The system of claim 27. wherein the oral health score includes at least one of gingivitis, periodontitis, peri-implant mucositis, or peri-implantitis risk score.

29. The system of claim 27, wherein the model module comprises a random forest model.

30. The system of claim 27, wherein the analysis unit is further configured to: perform DNA sequencing on the isolated bacteria material; and analyze the DNA sequencing to quantify' microbial species.

31. The system of claim 27, wherein the relative abundance is at a family level.

32. The system of claim 27, wherein the relative abundance is at a genus level.

33. The system of claim 27, wherein the model module is further configured to: identify a high risk of oral disease based on the oral health score generated from relative abundance data by an Al model traversing a weighted tree data structure.

34. The system of claim 27, wherein the one or more treatment regimens comprises at least one treatment selected from: saliva increasing chewing gum, xylitol chewing gum, surgical intervention, gum grafting, implant removal, partial implant removal, chemical debridement, mechanical debridement, acid etchant cleaning, custom tray delivery of antimicrobial or peroxide medicaments, implant surface polishing, hard tissue grafting, soft tissue grafting, surgical soft tissue resection, or gingival pocket irrigation or disinfection.

35. The system of claim 27, wherein the relative abundance of the microbial species includes a relative abundance measure of early bacteria, bridge bacteria, and late pathogenic bacteria.

36. The system of claim 27, wherein; the data recording module is further configured to determine a peri-implant disease- associated taxa percentage from the relative abundance of microbial species; and the model module is further configured to provide the oral health score based on an input that includes the peri-implant disease-associated taxa percentage.

37. The system of claim 27, wherein; the data recording module is further configured to determine an anaerobe score from the relative abundance of microbial species; and the model module is further configured to provide the oral health score based on an input that includes the anaerobe score.

38. The system of claim 27, wherein; the data recording module is further configured to determine a gram stain profile from the relative abundance of microbial species; and the model module is further configured to provide the oral health score based on an input that includes the gram stain profile.

39. The system of claim 27, wherein; the data recording module is further configured to determine an alpha diversity from the relative abundance of microbial species; and the model module is further configured to provide the oral health score based on an input that includes the alpha diversity.

40. A method for analysis of oral health, the method comprising: receiving a sample comprising bacteria from an oral cavity⁷ of a subject; analyzing the sample to generate an analysis comprising a relative abundance of bacteria in the sample; determining, using a trained machine model, a classification of the relative abundance indicative of an oral health issue; and generating a report based on the classification, wherein the report comprises personalized intervention recommendations.

41 . The method of claim 40, wherein the sample comprises a saliva sample.

42. The method of claim 40, wherein the sample comprises a plaque sample.

43. The method of claim 40, wherein the relative abundance is at a family level.

44. The method of claim 40, wherein the relative abundance is at a genus level.

45. The method of claim 40, w herein the trained machine model comprises a random forest structure.

46. The method of claim 40, wherein the trained machine model comprises a decision tree structure.

47. The method of claim 40. further comprising: determining a current state of oral health based on the classification; and determining a probability of transitioning into a second stage of oral health based on the classification.

48. The method of claim 47, further comprising: applying a qualifier to the analysis prior to the use of the trained machine model, wherein the qualifier is configured to screen for high-risk pathogenic profiles.

49. The method of claim 47, further comprising: adjusting the probability of transitioning based on subject demographic data.

50. A method of using a trained machine learning model to identify a treatment based on oral biome composition, comprising: training, by a computer, a machine model based on input data and a selected training algorithm to generate a trained machine model, wherein the selected training algorithm includes a computation of a loss function; detecting one or more health-adverse biome compositions using the trained machine model; determining that the one or more health-adverse biome compositions are associated with one or more oral diseases; identifying, based on the detected health-adverse biome compositions, a type of treatment for the one or more oral diseases; and generating, based on the biome composition, a regimen for the type of treatment.

51. The method of claim 50, wherein the regimen comprises antibiotic dosage.

52. The method of claim 50, wherein the regimen comprises a ty pe of antibiotic.

53. The method of claim 50, wherein the biome composition comprises a relative abundance of bacteria.

54. The method of claim 53, wherein the relative abundance is at a family level.

55. The method of claim 53, wherein the relative abundance is at a genus level.

56. The method of claim 53, wherein the relative abundance includes a relative abundance measure of early bacteria, bridge bacteria, and late pathogenic bacteria.

57. The method of claim 50, wherein the trained machine model is a random forest model.

58. The method of claim 50, wherein the trained machine model is a neural network model.

59. The method of claim 57, wherein training further comprises: selecting a random subset of the input data; and generating a tree structure based on the subset of the input data.

60. The method of claim 59, wherein training further comprises: generating a plurality of tree structures based on different random subsets of the input data.

61. The method of claim 60, wherein detecting one or more health-adverse biome compositions using the trained machine model comprises at least one of voting or averaging outputs of each tree.

62. The method of claim 50, wherein the loss function is a mean absolute error comprising a measure of an average absolute difference between observed and predicted values.

63. A computer-implemented method of training a machine learning model for biome analysis comprising: collecting a set of biome data for a set of patients from a database; applying one or more transformations to each biome data to create a modified set of biome data; creating a first training set comprising the modified set of biome data and a set of patient data of the set of patients; training the machine model in a first stage using the first training set; creating a second training set for a second stage of training comprising the first training set, the modified set of biome data, and biome classification data from the machine learning model after the first stage of training; and training the machine learning model in the second stage using the second training set.

64. The method of claim 63. wherein training the machine model in the first stage comprises training random forest machine learning methods.

65. The method of claim 63, wherein training the machine model in the second stage comprises training decision tree categorization methods.

66. The method of claim 63, wherein training the machine model in the first stage comprises training a neural network model.

67. The method of claim 66, wh erein training the machine model in the second stage comprises fine- tuning the neural netw ork model.

68. The method of claim 66, w herein the machine learning model is trained to identify one or more health-adverse biome compositions using the trained machine model.

69. The method of claim 63, wherein the machine learning model is trained to identify one or more health-adverse biome compositions using the trained machine model.

70. The method of claim 63, w herein applying one or more transformations comprises determining a relative abundance of bacteria at a family level or a genus level.

71. A computer-implemented method of training a machine learning model for predicting progression of an oral disease using biome analysis, the method comprising: collecting a set of biome data for a set of patients from a database; applying one or more transformations to each biome data to create a modified set of biome data; creating a first training set comprising the modified set of biome data and a set of patient data of the set of patients: training the machine model in a first stage using the first training set; creating a second training set for a second stage of training comprising the first training set, the modified set of biome data, and biome classification data from the machine learning model after the first stage of training; and training the machine learning model in the second stage using the second training set.

72. A method for processing biome data and generating personalized user reports, comprising: receiving, by a computer system, raw biome data collected from oral biological samples; preprocessing the raw biome data to remove artifacts and normalize the data; storing the preprocessed biome data in a centralized database; extracting features from the preprocessed biome data; analyzing the extracted features using one or more machine learning models to generate biome classifications; storing the biome classifications in the centralized database; receiving a request for biome analysis for a subject; retrieving relevant biome data and classifications for the subject from the centralized database; retrieving personal data of the subject, wherein the personal data includes at least one of demographic information, medical history, lifestyle factors, or dental care habits; generating a personalized biome report for the subject based on the retrieved data, classifications, and personal data; and transmitting the personalized biome report to a user device.

73. The method of claim 72, wherein the personalized biome report comprises an identification of an oral disease for the subject.

74. The method of claim 73, wherein the personalized biome report comprises a treatment plan for an oral disease.

75. The method of claim 72, wherein the request for the biome analysis is received from the user device and the personalized biome report is generated in real time based on the request.

76. A system for multi-site biome sampling and analysis, comprising: a first collection device configured to be introduced into an oral cavity and to collect a first biological sample at a first site of a subject; a second collection device configured to be introduced into the oral cavity and to collect a second biological sample at a second site of the subject; a processing module configured to extract biological material from the collected biological samples to isolate bacteria; an analysis unit, configured to analyze the isolated bacteria material to identify and quantify microbial species at the first site and the second site; a data recording module integrated with the analysis unit, configured to record a relative abundance of the microbial species at the first site and the second site; a model module configured to: correlate, using a first model, the recorded relative abundance at the first site with a first oral health score; correlate, using a second model, the recorded relative abundance at the second site with a second oral health score; analyze, using a third model, the first oral health score and the second oral health score to determine a third oral health score; a treatment identification unit configured to identify one or more treatment regimens based on the third oral health score; and a user interface configured to provide a platform for users to access the one or more treatment regimens.

77. The system of claim 76, wherein the first model and the second model comprise a random forest model.

78. The system of claim 76. wherein the third oral health score includes at least one of gingivitis, periodontitis, peri-implant mucositis, or peri-implantitis risk score.

79. The system of claim 76, wherein the third model comprises a neural network model.

80. The system of claim 76, wherein the first oral health score includes at least one of gingivitis or periodontitis risk score and the second oral health score includes a score for peri-implant mucositis or peri-implantitis risk .

81. The system of claim 76, wherein the first model module is further configured to: identify a first oral health score based on a relative abundance data by an Al model traversing a weighted tree data structure.

82. The system of claim 76, wherein the one or more treatment regimens comprises at least one treatment selected from: saliva increasing chewing gum, xylitol chewing gum, surgical intervention, gum grafting, implant removal, partial implant removal, chemical debridement, mechanical debridement, acid etchant cleaning, custom tray delivery of antimicrobial or peroxide medicaments, implant surface polishing, hard tissue grafting, soft tissue grafting, surgical soft tissue resection, or gingival pocket irrigation or disinfection.

83. A method for multi-site biome sampling and analysis, comprising: introducing one or more collection devices into an oral cavity and collecting biological samples from a first site and a second site within the oral cavity of a subject; extracting biological material from the collected samples to isolate bacteria; analyzing the isolated bacteria to identify and quantity’ microbial species at the first site and the second site; recording a relative abundance of the microbial species at the first site and the second site; correlating the recorded relative abundance at the first site with a first oral health score using a first model; correlating the recorded relative abundance at the second site with a second oral health score using a second model; analyzing the first and second oral health scores using a third model to determine a third oral health score; identifying one or more treatment regimens based on the third oral health score; and providing a platform through a user interface for users to access the identified treatment regimens.

84. A post-dental implant treatment method comprising: collecting a first set of biological samples, at a first time, from an oral cavity of a subject to provide a first biome dataset; collecting a second set of biological samples, at a second time, from the oral cavity' of the subj ect to provide a second biome dataset; determining a risk level of the subject for oral disease based on a risk score that is generated from changes in a relative abundance of bacteria between the first biome dataset and the second biome dataset by an Al model that traverses a weighted tree data structure to analyze the changes in the relative abundance of bacteria to provide the risk score; and determining a first treatment plan for the subject, wherein the first treatment is based on the changes of the relative abundance of the bacteria, the risk level, and a relative abundance of bacteria in the second biome dataset.

85. The method of claim 84, wherein administering the first treatment comprises: administering an antibiotic regimen to the subject according to the first treatment plan of oral disease after dental implant surgery, wherein the antibiotic regimen comprises a plurality' of antibiotics selected based on the relative abundance of the bacteria in the second biome dataset and antibiotic resistance of the bacteria.

86. The method of claim 84. wherein the changes in relative abundance comprise changes in the relative abundance of early bacteria, bridge bacteria, and late pathogenic bacteria.

87. The method of claim 84, wherein the second set of biological samples is collected after a second treatment, and wherein the first treatment is different from the second treatment.

88. The method of claim 87, further comprising: determining adverse effects of the second treatment based on the changes in the relative abundance of bacteria.

89. The method of claim 87, further comprising: determining an effectiveness of the second treatment based on the changes in the relative abundance of bacteria.

90. A system for post-dental implant treatment, comprising: a first module configured to collect a first set of biological samples at a first time from an oral cavity of a subject, and to generate a first biome dataset from these samples; a second module configured to collect a second set of biological samples at a second time from the oral cavity of the subject, and to generate a second biome dataset from these samples; an analysis module comprising an Al model that traverses a weighted tree data structure to analyze changes in a relative abundance of bacteria between the first biome dataset and the second biome dataset, and to generate a risk score for identifying the subject as being at a first risk level of oral disease; and a treatment module configured to determine a treatment plan for the subject identified as being at the first risk level of oral disease after dental implant surgery, wherein the treatment is determined based on the changes in the relative abundance of bacteria between the first and second biome datasets, as well as the relative abundance of bacteria in the second biome dataset.