WO2012121618A1

WO2012121618A1 - Multisensor and method for evaluating taste characteristics of analytes

Info

Publication number: WO2012121618A1
Application number: PCT/RU2011/000145
Authority: WO
Inventors: Andrey Vladimirovich LEGIN; Alisa Mikhailovna RUDNITSKAYA; Dmitry Olegovich KIRSANOV
Original assignee: Individual
Current assignee: Individual
Priority date: 2011-03-09
Filing date: 2011-03-09
Publication date: 2012-09-13
Anticipated expiration: 2013-09-09

Abstract

The present invention relates to a multisensor for evaluating taste characteristics of an analyte and a method for carrying out the same which can be employed, for example, in pharmaceutical industry, food industry, waste environmental control and other fields where such evaluation may be necessary or desirable. In particular, the present invention relates to a multisensor equipped with several chemical sensors exhibiting cross-sensitivity to components of a sample under evaluation for quantitative and qualitative assessment of some basic taste properties such as bitterness, saltiness, sweetness, sourness or umami or any combination thereof as well as some specific tastes well known in organoleptic gustatory analysis.

Description

MULTISENSOR AND METHOD FOR EVALUATING

TASTE CHARACTERISTICS OF ANALYTES

FIELD OF THE INVENTION

The present invention relates to a multisensor for evaluating taste characteristics of an analyte and a method for carrying out the same which can be employed, for example, in pharmaceutical industry, food industry, waste environmental control and other fields where such evaluation may be necessary or desirable. In particular, the present invention relates to a multisensor equipped with several chemical sensors exhibiting cross- sensitivity to components of a sample under evaluation for quantitative and qualitative assessment of some basic taste properties such as bitterness, saltiness, sweetness, sourness or umami or any combination thereof as well as some specific tastes well known in organoleptic gustatory analysis. BACKGROUND OF THE INVENTION

Assessment of the taste and flavor of oral drug preparations is of major interest to the pharmaceutical industry, for example for research-based companies. Typical tasks include evaluation of taste changes caused by aging or during development of masking of unpleasant (usually bitter) taste of active substances and selection of the least bitter- tasting molecules from a number of newly obtained chemical entities.

Traditionally taste assessment of pharmaceutical preparations is carried by a taste panel comprised of a certain number of specially trained volunteers within well-controlled procedures. Despite representing one of the most reliable ways for taste evaluation this approach have a number of drawbacks as being slow, expensive, subjective and, in some cases, poorly reproducible. The tasting is also complicated by ethical restrictions due to the fact that the taste panel has to be exposed to active drugs while being healthy, even at levels considerably lower than the therapeutic dose as in the case of "rinse-and-spit" type studies. Moreover, drugs can be taken by the panel provided that the toxicological profile has been established which undoubtedly limits early-stage development efforts.

Although it is imperative that any pharmaceutical compound has the appropriate activity, selectivity and ADME (absorption, distribution, metabolism, elimination) characteristics it is also important that its formulations are acceptable to the patients in need and hence consumed by them. No matter how effective the active moiety in a pharmaceutical product is, this cannot be therapeutically beneficial unless it is actually taken (and often repeatedly) by the patient. This may even cause financial losses once two or more products with similar API efficiency and safety profiles but different palatability are on a market. Hence appearance, smell, taste and texture of pharmaceutical products are of great importance and should be given enough consideration prior to commercialization. Yet the same is taste evaluation in field of veterinary.

As alternative to a human taste panel certain types of taste evaluating multisensor array systems also known as electronic tongues for measuring taste characteristics, especially bitterness, of different pharmaceutical formulations have been recently developed worldwide. The main groups of electronic tongues comprise potentiometric sensors including lipid membrane taste sensors (LMTSs), ion-sensitive field effect transistors (ISFETs), voltammetric electronic tongues, electronic tongues equipped with optic-based sensors and combinations thereof in one sensor system (see Vikas Anand et al, Drug Discovery Today, Vol. 12, Numbers 5/6, March 2007 incorporated herein by reference). These types of sensor systems will be briefly described below.

LMTS capitalize upon the properties of lipids which participate in the natural process of taste. The sensors are formed by dispersing the lipid compound responsible for transducing the signal on to a polymeric matrix that is normally non-conduciing, such as a polyvinyl chloride (PVC). Such sensors analyze in a non-specific manner detected signals and hence can extract the inherent taste characteristics of substances.

In particular, two typical well-known examples of LMTSs are Taste Tasting Systems SA401 and SA402 which have been developed by Anritsu Corporation together with researchers at Kyushu University in Japan (see e.g. US 5302262, US 5482855, JP 5099896, JP 6174688). The detecting sensor part of the systems consists of seven (SA401 ; Anritsu Co. , Ltd., Japan) or eight (SA402; Intelligent Sensor Technology, Inc. , Japan http:// www.insent .co.jp) electrodes (channels) made of lipid-polymer membranes. Different types of lipid are used for preparing the membrane. Each lipid is mixed in a test tube containing polyvinyl chloride and a particular plasticizer, dissolved in tetrahydrophuran, and dried on a glass plate at 30 °C to form a transparent thin film, almost 200 pm thick. Lipid or polymer membranes are fitted on a multichannel electrode that acts as the detecting electrode. The detecting electrode of each channel is made up of silver wires plated with Ag/AgCI which is kept in holes filled with 3 M KCI solution. The electrode is connected to a scanner through high-input impedance amplifiers. The voltage difference between the multichannel detecting electrode and an Ag/AgCI reference electrode is measured with active and placebo formulations. The potentiometric response data from all probes for active formulations and placebo formulations are compared using principal component analysis (PCA) mapping. Thus lipid membranes immobilized with a polymer act as transducers converting the taste sensation into an electronic potential pattern. Presently, two other similar Taste Tasting Systems SA402B and TS-5000Z based on the same lipid material are commercially available (Yoshikazu Kobayashi et al, Sensors 2010, 10, 341 1 -3443).

More specifically the above-indentified sensor systems have been applied for evaluating taste attributes of food products, beverages and several pharmaceuticals including amino acids (see Yohko Miyanaga et al, "Prediction of the bitterness of single, binary- and multiple- component amino acid solutions using a taste sensor" International Journal of Pharmaceutics 248 (2002) 207-218 and Kikkawa et al., Discrimination of taste of amino acids with a multichannel taste sensor, Jpn. J. App. Phys. 32, 5731-5739, 1993), various commercial drugs (see Takahiro Uchida et al, "A new method for evaluating bitterness of medicines by semi-continuous measurement of adsorbtion using a taste sensor " , Chem. Pharm. Bull 49 (10) 1336-1339 (2001) or Atsu Tanigake et al, "The bitterness Intensity of Clarithromycin evaluated by a taste sensor", Chem. Pharm. Bull 51 (1 1) 1241-1245 (2003)) in individual and multi-component aqueous solutions and for studying bitter taste masking techniques as well showing adequate results which are in good agreement with data of sensation analysis obtained by volunteer taste panels (see e.g. Sou Takagi et al, "Quantification of Suppression of Bitterness using an Electronic Tongue", J. Pharm. Science, vol. 90, no. 12, December 2001).

Ion-sensitive field effect transistors (ISFET) are prepared by pasting artificial lipid-polymer membranes of the same composition, as in LMTS above, on to the gate of FET. The FET taste sensor has the same sensitivity to taste substances as LMTS, but the potential reproducibility is less than that for LMTS and the lifetime is shorter for miniaturized devices. ISFETs detect ions in a solution through the use of selective membranes containing dispersed amorphous semi-conductors (PVC membranes containing dispersed semiconductors) and conventional electrodes. Such sensors do not permit the detection of non-polar substances, such as coffee, for example or those that do not form electrolytes, such as sacharose. In particular, one of the well known taste evaluating systems based on ISFETs is Astree electronic tongue developed by Alpha M.O.S. (http://www.alpha-mos.com), the instrument being equipped with a seven-sensor probe assembly for qualitative and quantitative analysis. It is fully automated, with 16 to 48 positions for formulation samples. The probes consist of a silicon transistor with proprietary organic coating, which modifies the physical properties of the sensor, resulting in potential variations. The measurement is potentiometric, with readings taken against Ag/AgCI electrode. Each probe is cross- selective to enable coverage of the full taste profile. The system samples, quantifies, digitizes, records and process potentiometric readings with multivariate statistical tools integrated with software.

Advantageously the system can be combined with a solid foam dynamic analyzer (S-FDA) to measure, dynamically, the taste of drug in solid dosage forms during their dissolution phase for simulation of the buccal dissolution mechanism, to measure tablet coating dissolution and to check the effectiveness and homogeneity of the coating on different dosage forms when dissolving. Thus, the system provides methods for obtaining taste and dissolution data simultaneously. In particular a patent application US 20040191918 discloses a method and an apparatus for quantifying the bitterness of a sample comprising an active drug by utilizing electronic tongue device equipped with a plurality of electrochemical liquid sensors having overlapping sensitivities, connected to the processing system for data acquisition by means of multivariative analysis. Preferably, the electronic tongue device represents a Astree ISFET-based electronic tongue system as indicated above.

The voltammetric electronic tongue developed by S-Sonse consists of four working metal electrodes made of gold, iridium, platinum and rhodium, an Ag/AgCI reference electrode and a stainless steel counter electrode. A relay box enables the working electrodes to be connected consecutively, to form four standard three-electrode configurations. The potential pulses/steps are applied by a potentiostat which is controlled by personal computer (PC). The PC is used to set and control the pulses, measure and store current responses, and to operate the relay box Voltage pulses are applied to the working electrode and the resulting current is measured. A hybrid electronic tongue has also been developed based on the combination of the measurement techniques potentiometry, voltametry and conductivity. It has many useful applications, including the continuous monitoring of milk in the dairy industry, in the water- cleaning process, in dishwater process, to follow fermentation during yoghurt fabrication, as an instrument to detect trace amounts of, for example, cadmium, lead or copper in soil and for the continuous monitoring of the chemical oxygen demand (COD)

For example, US 6,290,838 (Alpha MOS) discloses an apparatus for characterizing liquids characterized in that it comprises at least one electrode, equipped with at least two types of sensor, such as lipid sensor, quartz microbalance (QMB), surface acoustic device (SAW device), ISFET, MEMFET, optic-based sensor etc. , the respective physical or chemical reactions of which when soaked in a liquid are of different nature, the sensors of the measuring electrode(s) being non-specific sensors intended to generate respective output signals emanating from the sensors; and units for capturing and processing signals.

Another type of taste-sensing systems is electronic tongue developed by the University of Texas and Vusion, Inc. The electronic tongue initially developed by the University of Texas consists of a light source, a sensor array and a detector. The light source shines onto chemically adapted polymer beads arranged on a small silicon wafer, which is known as a sensor chip. These beads change color on the basis of the presence and quantity of specific chemicals. The change in color is captured by a digital camera and the resulting signal converted into data using a video capture board and a computer. The technology can be applied to the measurement of a range of chemical compounds, from simple such as calcium carbonate in water through to complex organic compounds such as haemoglobin in blood and proteins in food. Moreover, it is helpful in discriminating mixtures of analytes, toxins and/or bacteria in medical, food/beverage and environmental solutions. As a result, the electronic tongue has many potential uses in the food, beverage, chemical and pharmaceutical industries. Vision, Inc. is developing a chemical analyzer and sensor cartridge, based upon the electronic tongue technology of University of Texas, that can instantly analyze complex chemical solutions. The analyzer consists of a customized housing into which the sensor cartridge can be placed and exposed to liquid chemicals within a process plant. However, the majority of known electronic tongues have several disadvantages. In particular, most sensor systems comprise standard sensors regardless of a system under analysis that leads to a low level of discrimination capacity in some particular cases. With respect to LMTSs and ISFET sensors which are the types of sensors most similar to the sensors used in the present invention it would be worth noting that such e-tongues comprise a number of globally sensitive sensors wherein each sensor (or a group of sensors) is responsible for a particular taste characteristic, i.e. each of them corresponds to classic taste stimuli. For example, the already mentioned Taste Tasting System SA402B includes seven lipid sensors, three of which are bitterness sensors and the other four are umami, saltiness, sourness and astringency sensors accordingly. Although global sensitivity is a key feature for covering a broad spectrum of pharmaceutical compositions, or any solution in general, in terms of evaluation of their taste properties such e-tongues can still be useless for certain analytes since taste attributes in different formulations are often due to ingredients of different nature.

Furthermore, LMTS electronic tongues were reported to be susceptible to minor changes in analytical conditions such as small deviations in room temperature (up to 1 to 10 degrees) and age or history of sensors. In other words an output signal of a particular sensor may be drifting within time or ambient temperature when being immersed into the same solution.

Thus, although the known multisensor array systems have acquired considerable popularity and commercial success, there has been continuing need of multisensor array systems which would have better properties in terms of taste evaluation of various multi- component compositions.

Furthermore, it would be advantageous to expand the range of analytical systems which can be easily assessed in terms of their taste characteristics.

In view of the foregoing, it is an object of the present invention to provide a multisensor, which is on one hand, is suitable for taste evaluation of wide range of compositions comprising various chemical species, whilst on the other hand, allows generation of stable reproducible data signals irrespective of ageing of the sensors. Particularly, it is an object of the invention to provide a multisensor for taste evaluation which can be exploited in different fields of research and technology including pharmaceutical industry, waste environmental control, food industry, alcohol industry, etc.

Yet another object would be a particular sensor array which can be employed in the multisensor of the above-indentified type.

A still another object of the present invention is to provide a simple method for taste evaluation of any liquid sample using the above-indentified multisensor. Finally, it is an object of the present invention to provide a method for evaluating taste- masking efficiency of a sample.

SUMMARY OF THE INVENTION

The present invention is based on the discovery of a multisensor for evaluating taste properties of various liquid samples or analytes, the multisensor comprising newly developed cross-sensitive chemical sensors. Further a method for evaluating taste properties of various analytes is disclosed.

In its first aspect, the present invention provides a multisensor for assessment of taste characteristics of a liquid sample comprising one or more components, the multisensor comprising

an array of a plurality of at least two measuring sensors comprising an active sensing compound in a matrix, wherein each sensor exhibits cross-sensitivity to the components of the sample; and

a device for registration of signals generated by said at least two sensors ; wherein at least one of the at least two measuring sensors comprises as its active compound a compound of a general formula (I):

(I)

where Ri , Rz, R3, 4, 5 may be same or different and are independently hydrogen; halogen; hydroxyl;

d e alkyl optionally substituted with hydroxyl, C-,,₆alkoxy or at least one halogen atom; C₃.₁₀cycloalkyl optionally substituted with hydroxyl, d-ealkyl, d ealkoxy or at least one halogen atom; aryl optionally substituted with C-i ₆alkyl, at least one halogen atom, hydroxyl, d -ealkoxy;

5-6-membered oxygen-containing heterocyclyld ealkyl; 5-6-membered heteroaryl comprising 1 -3 heteroatoms selected from O, N or S; -C(=0)OH; -C( -0)OR₆ where R₆ is C_{1 6} alkyl optionally substituted with hydroxyl or halogen;

with the proviso that

at least one of R_{1 t} R₂, R₃, R₄, R₅ is -C(=0)NR₇R₈ where

R-_/, R₈ are independently hydrogen, d-y alkyl optionally substituted with hydroxyl, Ci.. alkyl, d.₆alkoxy or one or more halogens; aryl optionally substituted with C_{1 7}alkyl, one or more halogens, hydroxy!, d ₆aikoxy; or

Ry, R₈ may together with the nitrogen atom to which they are attached form a 5-6-membered heterocyclic group optionally substituted with d _/alkyl, d ealkoxy, one or more halogen atom, or phenyl;

and/or any mixtures thereof.

In its second aspect the present invention provides a method for evaluating taste characteristics of a liquid sample, the method comprising the following steps:

(a) providing a multisensor according to claim 1 ;

(b) washing each measuring sensor with a washing solution suitable for the sensors and the anaiyte being measured;

(c) soaking each measuring sensor with a conditioner suitable for a particular measuring sensor;

(d) immersing each measuring sensor within the sample subjected for the taste assessment,

(e) collecting data in the form of output signals generated by the at least two measuring sensors;

(f) processing the data.

Yet in its third aspect the present invention provides an array comprising at least two cross-sensitive sensors comprising an active sensing compound in a matrix, wherein at least one of the said at least two sensors comprises as its active compound a compound of a general formula (I).

Yet in its forth aspect the present invention provides a membrane for use in a measuring sensor, the membrane comprising a plasticized polymer matrix, and an active compound in the matrix, wherein said active compound represents a compound of a general formula (I)^.

Yet in its fifth aspect the present invention provides a method for evaluating taste masking efficiency of a sample using the above-identified multisensor.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates (a) a scheme and (b) a photo of a multisensor according to a particular embodiment.

FIG. 2 illustrates various types of chemical sensors which can be used in particular embodiments of a multisensor according to the present invention. FIG. 3 represents three SEM images demonstrating evolution of the surface layer of Cu- Ag-As-Se chalcogenide glass membrane. Fig 3 (b) shows early formation of the surface sensing layer as obtained by soaking the sensor in distilled water. Fig. 3(c) shows grained-type structure of the surface membrane layer after soaking the sensor in a particular conditioner.

FIG. 4 represents SEM images of chalcogenide glass (Ag-As-S) sensor (a) without conditioning, (b) after 10080 minutes in 1 mmol/l CuCI_? solution, (c) after 12600 minutes in 1 mmol/l H₂0₂. FIG. 5 represents a 3D SEM image showing topology of a part of a chalcogenide glass sensor.

FIG. 6 (a) illustrates discrimination between bitter, sweet and salty substances based on the data obtained from a multisensor according to one preferable embodiment using LDA; (b) classification of bitter and sweet tasting substances based on the data obtained from said multisensor.

FIG. 7 illustrates

(a) - (b) discrimination of different substances having the same taste (salty or sweet) by the multisensor according to one preferable embodiment. Data were processed by using PCA;

(c) identification of different bitter tasting substances by said multisensor. Data were processed by PCA. The arrow on the plot shows the supposed direction of bitterness intensity increase.

FIG. 8 illustrates quantitative evaluation of the content of drug A by the multisensor according to one preferable embodiment. Data were processed by PCA. FIG.9 illustrates

(a) distinguishing effect of the multisensor according to one preferable embodiment when subjected for the taste assessment of compositions comprising the same quantity of drug A and different flavorings: fresh peach, aged peach, orange and strawberry. Data were processed by PLS2;

(b) recognition of quinine samples with and without flavorings: 1 unflavoured quinine (reagent grade, Mecrk), 2 unflavoured commercial quinine tablet, 3 and 4 flavoured commercial quinine tablets, 3 poorly masked bitter taste, 4 well-masked bitter taste. The arrow on the plot shows the direction of bitterness intensity increase. Data were processed by PCA.

FiG.10 illustrates

(a) evaluated masking effect of bitter taste of drug substances (drug A, quinine and drug C) by the sweetener aspartame using the multisensor according !o one preferable embodiment wherein the content of aspartame is / 0.5 mmol/L, // 1 rnmol/L; /// 2mmol/L. Data were processed by PLS ;

(b) evaluation of mixing effect of bitter substance (caffeine) with sweet one (aspartame) by using the multisensor according to one preferable embodiment. Data were processed by PLS2.

DETAILED DISCRETION OF THE INVENTION Herein below embodiments of the present invention will be described as non-limiting examples in connection with the accompanied drawings. According to the first aspect, the present invention relates to multisensor for assessment of taste characteristics of a liquid sample comprising one or more components, the multisensor comprising

an array of a plurality of . at least two measuring sensors comprising an active sensing compound in a matrix, wherein each sensor exhibits cross-sensitivity to the components of the sample; and

(I)

where

R-i , R₂, R₃, R₄, R₅ may be same or different and are independently hydrogen; halogen; hydroxyl;

Ci„₆ alkyl optionally substituted with hydroxyl, C-, ₆alkoxy or at least one halogen atom; C_{3 l}ocycloalkyl optionally substituted with hydroxyl, Ci ₆alkyl, d ₆alkoxy or at least one halogen atom; aryl optionally substituted with C_{1 -6}alkyl, at least one halogen atom, hydroxyl, 0·, ₆alkoxy;

5-6-membered oxygen-containing heterocyclic^ _ealkyl; 5-6-membered heteroaryl comprising 1 -3 heteroatoms selected from O, N or S; -C(=0)OH; -C(^~0)OR₆ where R₆ is C-i 6 alkyl optionally substituted with hydroxy! or halogen;

with the proviso that

at least one of R₂, R₃, R₄, R₅ is -C(=0)NR₇R₈ where R₇, R₈ are independently hydrogen, C alkyl optionally substituted with hydroxyl, d ₇alkyl, d.₆alkoxy or one or more halogens; aryl optionally substituted with C_1-7alkyl, one or more halogens, hydroxyl, C_r.₆alkoxy; or

R₇, R₈ may together with the nitrogen atom to which they are attached form a 5-6-membered heterocyclic group optionally substituted with d ₇alkyl, Ci_-6alkoxy, one or more halogen atom, or phenyl;

and/or any mixtures thereof.

According to some embodiments the multisensor consists of several main parts connected to each other to the extent so as to work as a unified apparatus. FIG. 1 (a) depicts a scheme for one possible embodiment of a multisensor according to the present invention. One part of the multisensor is a sensor array 2 comprising at least two measuring sensors exhibiting cross-sensitivity to several chemical entities. The second part is a device for capturing a signal originating from each sensor such as, for example, a voltmeter 5. The third part is a computer 6 or any other electronic device suitable for data storage and data acquisition. In addition to these main parts the multisensor may optionally comprise a reference electrode 3 and a broader sub-group of sensors which one can select to use as measuring sensors for evaluating taste properties of a certain sample (not shown in Fig. 1 (a)).

Fig 1 (b) is a photograph of the multisensor according to a one preferable embodiment. A sample subjected for the taste assessment is filled into a container which is optionally placed on a magnetic stirrer 3 providing homogenization of the sample contents. Preferably the magnetic stirrer includes a hot plate or other means for controlling temperature of the sample. The sensor array 2 linked to the multichannel high impedance voltmeter 1 is immersed into the container for an period of time enough for achieving equilibrium conditions and an output signal emanating from each sensor is collected. The registration device must be capable of being connected to a computer which functions as a device where signals can be stored and processed by the methods known in the art.

According to certain applications, the term "a liquid sample" represents an aqueous solution of at least one chemical entity. Although non-aqueous solutions are rarely met in practice, it was proved to be useful to evaluate their taste characteristics as well by using the multisensor according to the present invention. In particular, the liquid sample may include emulsions, suspensions, dispersions, slurries and other mixtures of various chemical entities. Preferably, the liquid sample is an aqueous solution of at least one pharmaceutical compound such as an aqueous solution of a drug or a multi-component aqueous solution such a solution of a pharmaceutical composition. In some instances, the present multisensor can be employed for assessing taste properties of various analytes and liquid samples as could be found in food industry, milk industry, alcohol industry, waste environmental control, water control, cosmetics and other fields where such assessment may be necessary or desirable.

In general, the sub-group of sensors can be quite broad and comprise from 1 to 100 sensors which are not necessarily used each time but rather specifically selected depending on a type of a system such as, for example, a liquid solution of a pharmaceutical composition under evaluation. Preferably said group includes from 1 to 50, and more preferably from 1 to 30 sensors. Preferably, the array of a plurality of measuring sensors, which is optionally selected from the broader sub-group of sensors, comprises from 2 to 100, more preferably from 2 to 50, and the most preferably from 2 to 30 measuring sensors. According to another embodiment the number of the measuring sensors can be even less and be in the range, for example, from 2 to 10 or 2 to 5 sensors. Furthermore, one skilled in the art will appreciate that a whole sub-group of available sensors can be applied each time as measuring sensors regardless of a system under evaluation provided that not all signals are taken into account when the taste properties of the system are assessed.

The array of sensors which can be used as measuring sensors in the multisensor according to some embodiments is thoroughly compiled to provide evaluation of taste properties of any desirable sample. Furthermore, it should be understood that the multisensor per se may or may not comprise the sub-group of sensors which can be further employed as measuring sensors during analysis, however the measuring sensors are always selected from said sub-group with respect to a certain application. For example, a potential developer of different drug preparations constantly dealing with new compounds may require a multisensor with the whole sub- group of available sensors in order to be at the liberty of choosing any measuring sensor suitable for his or her purpose, whilst when it comes to routine taste assessment in a certain type of a composition the sensors might be selected from said sub-group only once, for example, when the multisensor is purchased. Hence, a type of a sample subjected to analysis shall be always taken into consideration when selecting measuring sensors. A procedure as proposed by the authors of the present invention which one may employ for selecting measuring sensors based on several parameters such as for example mean slope S, stability factor K and factor of non-selectivity F, will be described below in greater detail. From now on the term "sensor", unless specified, will refer to any sensor the multisensor can be equipped with regardless whether it is a measuring sensor or any sensor included into the sub-group.

Advantageously each sensor is highly reproducible that means that the standard deviation of a signal emanating from the sensor when soaked in a particular solution is not more than approximately 5% to 10% when measured daily throughout at least a month. Furthermore each measuring sensor is non-specific, i.e. it exhibits cross-sensitivity to several components of a sample under evaluation. The term cross-sensitivity when used with respect to a sensor in the present description corresponds to a chemical sensor which is not sensitive only to a particular component/ingredient but rather simultaneously to the group of the components which are present in a sample subjected to analysis.

According to some embodiments, said chemical sensors represent membrane sensors. In general, a membrane sensor consists of a support which usually functions as a transducer and a membrane as a sensitive element deposited thereon by any suitable procedure. The transducer transforms a signal emanating on the surface layer of the membrane and transfers it to a registration unit through any standard means well known in the art to a skilled technician. Although a transducer can transfer the signal generated by a sensor, it cannot make any difference between the sources of this signal (selectivity) or its intensity (sensitivity) or whatever else. A transducer cannot somehow improve a signal of a sensor it can only make it less detectable. Therefore one should carefully select an appropriate transducer for a specific application. The membrane in turn is a true sensing element of a sensor. It is responsible for differentiating between various components which are present in the system and shall be non-specific, i.e. cross-sensitive as was already mentioned above.

It should be understood that a composition of a membrane is of great importance for the purposes of the present invention due to the fact that overall performance of the multisensor depends on the exchanging process taking place on the surface of the membrane in use. In this respect one is always free to choose any suitable transducer known in the art to which the membrane can be attached depending on conditions of a certain application. Thus, the examples of the membrane sensors, which are not intended to limit the scope of the present invention, include one of the following, a potentiometric electrode, a field effect transistor (FET), light-addressable potentiometric sensor (LAPS), quartz piezoelectric device (or quartz microbaiance, QMB), acoustic wave device sensor (SAW), etc.

However, it should be understood that certain types of chemical sensors represent uniform structures which are not dividable into any simpler parts. For example, a potentiometric electrode as well as a sensor with a plasticized membrane comprising an active compound, such as for example ionophore, are the sensors wherein the sensing elements, i.e. membranes, are uniformly coupled with transducers.

On the contrary, FET, QMB and SAW devices are transducers which cannot adequately change their properties in response to a certain chemical environment unless a sensing layer (a membrane) is attached to them. Later throughout the present description one should appreciate that the terms F ET sensor, LAPS sensor, QMB sensor or SAW sensor are the sensors with sensing elements (membranes) attached to the corresponding transducers. Various methods of producing the above-indentified sensors are known to a person skilled in the art and can be found in one of the following sources (a) Bratov A. et al, Analytica Chemica Acta, Volume 678, Issue 2, 30 September 2010, Pages 149-159; (b) Jimenez-Jorquera et al, Sensors, Volume 10, Issue 1, January 2010, Pages 61 -83; (c) Schoning M.J., Kloock J. P., Electroanalysis, Volume 19, Issue 19-20, October 2007, Pages 2029-2038. A few non-limiting types of membrane sensors are depicted in Fig . 2. Fig 2 illustrates (a) a potentiometric electrode, (b) a light-addressable potentiometric sensor and (c) an ion- sensitive field effect transistor. Technically they look very similar to sensors known in the art however substantially differs from known analogues by a composition of a sensing element which is not selective to a particular ion but rather exhibits cross-sensitivity.

According to some embodiments a compound of a general formula (I) is introduced into a matrix of said sensing element.

According to another embodiment, the membrane of a chemical sensor is prepared from organic and/or inorganic material. In particular, the membrane includes an active compound spread within the membrane matrix. According to some embodiments the active compound can be absorbed, adsorbed, dissolved, dispersed, bonded, embedded or otherwise introduced into/onto the matrix. Thus the membrane can represent one of the following an oxide glass membrane; a chalcogenide glass membrane; a membrane based on single or (poly)crystalline compounds; a membrane comprising a polymer, a plasticizer and an active compound. The active compound can represent an ion-exchanger also known as an ionophore or a neutral carrier. The ion-exchanger binds positively or negatively charged species as organic and inorganic ions, whilst a neutral carrier can bind uncharged molecules of several components.

Compounds of the general formula (I) have been surprisingly found to be well suited active compounds for evaluating taste properties of different complex solutions. Preferably, in formula (I) each RT and R₆ are -C(0)NR₇R₈ where R_/, R₈ are independently hydrogen; C_{1 7} alkyl optionally substituted with hydroxyl, C_{1 Y}alkyl, C< ₆alkoxy or at least one halogen; aryl optionally substituted with C-i _/alkyl, at least one halogen, hydroxyl, C-i . ₆alkoxy; or R₇, R₈ may together with the nitrogen atom to which they are bonded form a 5-6-membered heterocyclic group optionally substituted with C₁ alkyl, C_{1 6}alkoxy at least one halogen atom or phenyl. Non-limiting examples of the active compounds which fall under the scope of the general formula (I) include numerous derivatives of dipicolinic acid including Ν,Ν',Ν,Ν'- tetraethylpyridine-2,6-dicarboxamide, N,N',N, N'-tetrabutylpyridine-2,6-dicarboxamide (TBDPA); N, N'-ditolyl-N,N'-diethylpyridine-2,6-dicarboxamide (TDPA); N. N'-diphenyl-N. N'- dimethylpyridine-2,6-dicarboxamide, N,N'-diphenyl-N ,N'-dibenzylpyridine-2,6- dicarboxamide; N,N'-dipropyl-N, N'-di(4-butylphenyl)pyridine-2,6-dicarboxamide, 4-Chloro- N, N', N, N'-tetraethylpyridine-2,6-dicarboxamide, 4-(2-Oxo-2phenylethyl)-N, N',N, N'- tetraethylpyridine-2,6-dicarboxamide, 4-[2-(Biphenyl-4-yl)- 2-oxoethyl]-N, N',N , N'- tetraethylpyridine-2,6-dicarboxamide, 6~diethylcarbamoyl-4-(2-hydroxy-2- phenylvinyl)pyridine-2-carboxylic acid methyl ester, 2,2^'-dipyridyl-6,6'~dicarboxylic acid diamides and the like. These compounds are commercially available or can be synthesized in a laboratory by well known procedures to a skilled technician.

A chemical sensor comprising a compound of the general formula (I) as an active compound surprisingly has been shown to provide high reproducibility and appropriate cross-sensitive properties to various components when employed in the multisensor disclosed herein when used for evaluating taste properties of various analytes. The membrane of this type can be prepared by mixing a polymer with a plasticizer and an active compound in a suitable solvent at ambient conditions. The so obtained mixture is allowed to dry for an enough amount of time producing a thin membrane which can be mechanically modified as needed and later attached to a suitable transducer in order to form a chemical sensor. The polymer which serves as a main membrane matrix ingredient can be selected from polyvinyl chloride (PVC), a various polyurethans (PUR), cellulose including acetyl cellulose and other suitable polymers well known in the art. A plasticizer aims at providing good mechanical properties of the membrane including appropriate flexibility and can be also selected from any suitable plasticizer including for example one of the following dioctyi phenyiphosphonate (DOPP), 2-nitrophenyl octyl ether (NPOE), bis-(l-butylpentyl) adipate (BBPA), phosphoric acid trix(2-ethylhexyl) ester (PTEH), diethylene glycol dibutyl ether (DGDE), bis-(2-ethylhexyl) sebacate (BEHS), 3- (trimethoxysilyl)propyl methacrylate (TMSPM), trioctyl trimellitate (TOTM), tributyl O- acetylcitrate and others. The multisensor equipped with the above-indentified sensor shows excellent correlation of its output data with human perception for numerous compositions, in particular liquid solutions of drugs. According to some embodiments, the multisensor additionally comprises a chemical sensor comprising a crystalline (single or poly) membrane. Preferably, the crystalline (single or poly) membrane comprises one or more metal chalcogenides of a general formula M_xC_y wherein M is a metal selected from an alkali metal, an alkaline-earth metal, a transition metal, a lanthanide or actinide, such as, for example, Na, Ca, Ag, Pb, Cu, Cd, Cr, etc ; C is a non-metal selected from a chalcogenes and/or halogens, such as, for example S, Se, Te, F, CI, Br, I; x is from 1 to 3 and y is from 1 to 7 Preferably the crystalline (single or poly) membrane additionally comprises one or more metals such as Pt, Au, Ag, Sb, Pd, Rh, Cr, Mo or the like. According to some embodiments, the membrane represents a non-specific glass membrane, such as an oxide glass membrane comprising one or more metal oxides of a general formula M_xO_y and/or one or more non-metal oxides of a general formula E,O_b and/or any mixtures thereof wherein M is a metal selected from an alkali metal, an alkaline-earth metal, a lanthanide, Al, Ga, In, Sc, Y and the like, x is from 1 to 2 and y is 1 to 3, and E is a non-metal selected from the group comprising a halogen, a chalcogen, N, P, As, B, C, Si and the like, z is from 1 to 2 and b is from 1 to 3. Particularly, a non-specific oxide glass membrane may comprise one of the following Na_?0, L\₂0, K_?0, MgO, CaO, SrO, Al_z0₃, SC₇O₃, Y?0₃ La₂0₃, Si0₂, As_?0₃, ZrO_?, o02, W0₃ and others. Yet according to another embodiment the membrane represents a non-specific glass membrane such as a chalcogenide glass membrane comprises one or more metal chalcogenides of a general formula M_xC_y and/or one or more non-metal chalcogenides of a general formula E₇C_b wherein M is a metal selected from an alkali metal, an alkaline- earth metal, a transition metal, a lanthanide or actinide; C is chalcogene selected from S, Se and/or Te; and E is non-metal selected from Si, As, P, Sb, Ge, Sn, I; x is from 1 to 3 and y is from 1 to 5; and z is from 1 to 3 and y is from 1 to 5; and/or any mixtures thereof. Preferably, the chalcogenide glass membrane comprises one or more of the following Na, Ca, Cu, Cd, Ag, Pb, Cr, As, Se, S or I. The non-limiting examples of a chalcogenide glass membrane include Cu-Ag-As-Se, Cd-Ag-S-I with each element being in different proportions.

Each of the above-indentified types of glass membranes when soaked in a liquid solution possesses cross-sensitivity to several components, such as several ions including cations, anions or neutral species. According to some embodiments in addition to non- specific sensors the multisensor can include up to several standard commercially available ion-selective sensors. For example, such commercially available sensors include sensors which are individually selective to one of the following ions H ' , Na' , K' , HV , Ca²' , Mg^?' , Cd^?\ Pb⁷' , Cu²' , Ag' , F , CI^", Br, I , NO, , CN , SCN , BF„ , CIO, , SO/ , COy", PC¹/ and others. Preferably the multisensor further comprises a pH glass electrode.

It has been found by the authors of the present invention that a glass membrane of a chemical sensor, when employed for evaluation of the taste properties of a sample has an improved surface layer with the thickness being at least of about 1 nm to several micrometers, preferably from about 50 nm to about 100 pm, more preferably from about 100 nm to 5 pm, and the most preferably from about from about 20 to 500 nm. The layer has a granular nanostructure which is formed during soaking in a liquid solution, the granules being several nanometers in size. This nanolayer is believed to be extremely crucial for overall performance of the multisensor as far as it provides high reproducibility and cross-sensitivity for a particular chemical sensor. While not wishing to be bound by any theory the authors' assumption is based on the idea that a layer which is formed on the surface of a membrane material accelerates a component-exchange process and/or adsorption process occurring on a phase border between the solution and the membrane. This results in increasing of reproducibility of the sensor and rises up sensitivity of the membrane at the same time, thus providing a desirable value of the cross-sensitive properties.

Although the key factor for a proper surface layer formation is membrane material itself, there are some other important conditions which should be taken into account. In particular, the layer might by optionally pre-formed on the surface of a membrane by soaking the sensor in a suitable conditioning solution for enough amount of time prior to be immersed in a solution which taste properties are to be assessed. A suitable conditioning solution or simply, a conditioner, is properly selected depending on the nature of a membrane of a particular chemical sensor and/or the system under analysis. The preferable non-limiting examples of a conditioning solution include 0 ⁴ to 10 ¹ M lithium sulfate, 10 ⁴ to 10 M potassium iodide, 10 ^b to 10 ^? M malic acid and/or any mixtures thereof. A period of time needed for layer formation is variable in each case and can be easily determined by a person skilled in the art. Usually a conditioner is used for storage purposes preserving undesirable processes on a surface of a membrane material. Therefore conditioning allows drastically increasing shelf-life of a sensor, i.e. a period of time when it exhibits stable and reproducible signals, and moreover provides facilitating the development of cross-sensitive properties to a particular sensor. This consequently allows achieving long durability of the present multisensor. In particular, shelf-life of each sensor described herein is not less than 6 months, more preferably not less than 12 months, and most preferable not less than 24 months during regular exploitation.

The layer formation on the surface of a chalcogenide glass membrane can be visually demonstrated by referring to Fig. 3(a)-(c). As can be seen the layer of the chalcogenide glass membrane forms by being immersed even in distilled water (Fig. 3(a)) However an overall grain size in this case is still rather big and should be desirably improved. This is achieved by soaking the membrane material in a suitable conditioner (Fig . 3(c)) which provides formation of the grains with the typical dimensions in the range from 20 to 500 nm. In addition to this organization, qualitative composition of the surface layer shown in Fig. 2 (c) can be dramatically different from that of an initial bulk material. Some of the components, for example Se, can vanish from the surface under oxidizing conditions which finally leads to the changes in physical and chemical properties of the surface layer compared to the bulk material. In other words, the surface becomes flexible to the external environment. For example, relative density of the surface layer was measured to be 20% to 30% of the density of the initial bulk material. Although Fig. 3 illustrates layer formation occurring on the surface of chalcogenide glass membrane other membranes including plasticized membranes are also undergone this process. However, surfaces of the latter are purely subjected to SEM analysis due to specific properties of the initial material itself. The analysis of the structure usually requires deep vacuuming of a sample whilst plasticized membranes are not merely stable at such conditions. Further, it has been found that shelf-life of a properly synthesized and conditioned sensor is not less than about 6 months, preferably not less than 12-24 months, during continuous intensive exploitation. This rather long period of time which is normally unusual for majority of chemical sensors is achieved mainly by the formation of nanostructure on the surface of the specifically selected or synthesized initial membrane material and its proper conditioning. The samples of the initial membrane material are not very large in size, preferably from at least 1 to 5 millimeters in diameter. At the same time this membrane materials provide desirable robustness and high reproducibility for the sensors.

Fig. 4 shows SEM images of the surface of Ag-As-S chalcogenide glass membrane material obtained (a) without conditioning; (b) after conditioning in 1 mmol/l CuCI_? solution for 168 hours; and (c) after conditioning in 1 mmol/l H_?O_z solution for 210 hours. It can be seen that the surface of the sample shown in Fig. 4(a) does not exhibit any visual structure organization at all. Nanostructure formed after conditioning the sample in the copper solution provided enhanced properties similar to those mentioned above for the glass material shown in Fig. 3. Fig. 4(c) shows the evolution of the surface in the presence of the strong oxidizing agent. The surface in this case becomes irregular and the membrane material irreversibly degrades which in turn was proved to sharply decrease the sensing properties of the sample. However, even after that irreversible process the surface of the membrane material can be renewed again by, for example, mechanical polishing followed by soaking the sample in a suitable conditioner provided that the sample is thick enough for being mechanically polished.

Not wishing to be bound by any theory the authors believe lhat the surface of a measuring sensor is not fully involved into interaction with an analyte. F ig. 5 illustrates three- dimensional SEM image showing small part of the surface of a chalcogenide glass membrane. As could be seen the surface is shown to have nano-sized protuberances which are believed to be the places where the exchanging process is very likely to happen. Said protuberances form a smaller part of the whole surface. Thus, if an area of the sensing surface is small the sensor's output signal will be weak (exchange current between material and solution) with its behavior being unstable. Thus, according to some embodiments the sensing elements of the chemical sensors are characterized by at least millimeter sized dimensions.

Cross-sensitivity evaluation

The present sensors are required to be cross-sensitive to various components of a liquid solution the taste of which is to be assessed. Another requirement is that the sensors were reproducible. T hese features are important for a good assessment of the taste properties and overall multisensor performance. Thus one is in need of any standard procedure which can be used for estimating whether a sensor is cross-sensitive and reproducible enough and therefore is useful in the present multisensor. In particular this procedure can be employed for the selection of a sensor array from a broader sub-group of sensors for use in the multisensor as described herein. Although the term cross-sensitivity is broadly used in literature there has not been proposed any unambiguous definition or any unified procedure for estimating this parameter so far. There are several other terms describing the same phenomenon as cross-sensitivity which are usually used interchangeably such as for example non- selectivity, cross-reactivity, low selectivity, partial sensitivity, global selectivity and so forth. There are several ways for estimating cross sensitivity of a sensor. According to some embodiments a method for estimating whether a particular sensor is appropriate is based on calculating three parameters: a mean slope (S), a reproducibility factor (K) and a factor of non-selectivity (F). Developed by the authors of the present invention the method is useful for any type of a sensor according to the present invention and will be described herein below.

The first and the most important parameter of cross sensitivity is the mean slope S which is determined though series of measurements taken in individual solutions of a given group of compounds by the following equation:

S = (1/n)^*∑S| where S, is a slope coefficient of an electrode function of a sensor in calibration solutions comprising individual compounds cross-selectivity to which is determined, and n is a number of the compounds in the solution. Thus, this parameter is a mean selectivity of a sensor to the group of compounds. The stability factor (K), the second parameter, is calculated by the following equation:

K = (1 /n)^*∑Kj = (1 /n)^*∑(S, Is- ) where S, is a slope coefficient of a sensor in calibration solutions and s, is a standard deviation of Si in individual solutions. Obviously, a standard deviation is calculated on several parallel measurements and characterizes reproducibility of a sensor in solutions under study. The factor of non-selectivity corresponds to a distribution of selectivity of a particular sensor to the group of compounds:

F = S/s^z where S is a mean slope coefficient and s is a standard deviation.

The higher the magnitude of the mean slope S, the higher is selectivity of a sensor to the group of compounds, i.e. the components of a solution. The optimal value is estimated in each individual case. For example, in the case of potentiometic electrodes used as sensors in the multisensor according to the present invention, S usually varies from 0, i.e. the electrode is absolutely not selective to 29 mV, i.e. the case when the signal is in compliance with Nernst equation at ambient conditions. Thus, the value of S in immediate proximity of 29 mV corresponds to the sensor which is equally selective to all the components. However, this is very rare in practice. Sometimes S is more than 30 mV which means that the sensor is highly selective to a one or more particular components whilst less selective to others. This would be the case where one should consider the value of the factor of non-selectivity, F, in order to make sure the sensor is well suited in terms of cross-sensitivity. If F is less than about 0.1 the sensor is typically highly selective to a particular component. The value of F being more than 0.1 corresponds to the sensor which has cross-sensitivity and thus can be used in the present multisensor. According to some embodiments F is preferably more than 0.5. The most preferably F is more than 1 which means that the sensor possesses relatively uniform cross sensitive distribution to all components in the solution. The stability factor K is used for estimation of a sensor reproducibility. Although the estimation of reproducibility is usually carried out in solutions T RU2011/000145

23

comprising individual components, it is obvious that the obtained values correlate well with reproducibility measured in mixed solutions comprising these individual components. The more the stability factor K, the more the reproducibility is. For the purposes of the present invention K is not less than about 1 , more preferably not less than 2. Further it means that the standard deviation of a signal emanating from the sensor when soaked in a particular standard solution is not more than 10% when measured daily throughout at least a month.

The above-indentified group of parameters conveniently provides reliable data for the selection of measuring sensors. Further, it would be worth mentioning that this method does not make a priori assumptions on type of the sensors and nature of their output signals. Thus, it can be implemented for the estimation of cross-sensitive properties of either potentiometric sensors or any type of chemical sensors such as a potentiometric electrode, FET sensor, LAPS sensor, quartz piezoelectical device, acoustic wave device sensor, etc. Further, it helps to establish the sensors which can be selected from the broad group of sensors as measuring sensors to be used in the multisensor according to the present invention.

(a) providing a multisensor according to claim 1 ;

(b) washing each measuring sensor with a washing solution suitable for the sensors and the analyte being measured;

(d) immersing each measuring sensor within the sample subjected for the taste assessment;

(f) processing the data.

Preferably, a multisensor which is employed in the method is the multisensor as described by any one of the embodiments above. In particular, it may comprise an array consisting of 2 to 100, preferably 2 to 50, the most preferably 2 to 30 measuring sensors which could be previously selected from a broader sub-group of chemical sensors. The measuring sensors possess adequate reproducibility and cross-sensitive properties to the components of a liquid sample which is subjected for the assessment of taste characteristics.

Preferably, the multisensor comprises

(I)

where

Ri, R₂, R3, 4, R5 may be same or different and are independently hydrogen; halogen; hydroxyl;

Ci_-6 alkyl optionally substituted with hydroxyl, Ci .₆alkoxy or at least one halogen atom; C₃.₁₀cycloalkyl optionally substituted with hydroxyl, C_{1 6}alkyl, ₆alkoxy or at least one halogen atom; aryl optionally substituted with ₆alkyl, at least one halogen atom, hydroxyl, C_{1 6}alkoxy;

5-6-membered oxygen-containing heterocyclylCi ₆alkyl; 5-6-membered heteroaryl comprising 1 -3 heteroatoms selected from O, N or S; -C(=0)OH; -C(=0)OR₆ where R₆ is C-i„6 alkyl optionally substituted with hydroxyl or halogen;

with the proviso that

at least one of R₂, R₃, R„, R₆ is -C(=0)NR₇R₈ where R₇, R₈ are independently hydrogen, C_{1 -Y} alkyl optionally substituted with hydroxyl, C, _yalkyl, C^alkoxy or one or more halogens; aryl optionally substituted with Ci_-7alkyl, one or more halogens, hydroxyl, Ci..₆alkoxy; or

R_Y, R_B may together with the nitrogen atom to which they are attached form a 5-6-membered heterocyclic group optionally substituted with _yalkyl, C_{1 6}alkoxy, one or more halogen atom, or phenyl;

and/or any mixtures thereof.

Selection of measuring sensors

Generally there is a standard approach which one can use to make a proper selection of a sensor array comprising measuring sensors from a broader sub-group of chemical sensors. The approach is completely empirical on its merits and provides only practical answers to questions what kind of chemical sensors should be taken for a certain task and how it could be achieved. Broadly, it can be described by the following steps. Firstly goes the formation of an original sensor array based on several general points including previous knowledge, availability of certain sensors and certain types of sensors, theoretical considerations (if any), the results of sensitivity and cross-sensitivity evaluations of the sensors, general knowledge about sensing mechanisms of the sensors and also about composition of a sample, and common sense. At the second step the measurements with the original broad sensor array (the sub-group of sensors) and elaboration of the precise procedures and specific methods of measurement are taken into consideration. Finally, the calculation of the results of the measurements in the terms of (i) recognition (classification, identification) of analytes; (ii) quantitative determination of individual parameters (for example, the determination of concentrations of inorganic and/or organic substances); (iii) correlation with and assessments of integral parameters (e.g. flavour, odour or taste in case of chemical sensors) (iv) general qualitative judgments such as "yes"-"no"; "good"-"bad"; "confirms to the standard"-"out of the standard", etc. Thus, only those sensors could be chosen which are well responsive to the taste characteristics (or other parameters) to the certain extent, whilst other irresponsive sensors can be excluded from the array. It should be also understood that this approach allows one to select several possible groups of measuring sensors and final selection is a matter of a skilled specialist's decision. According to some embodiments the multisensor additionally includes a chemical sensor comprising a glass membrane. Preferably, the glass membrane is an oxide glass membrane or a chalcogenide glass membrane. According to some embodiments the oxide glass membrane of the additional sensor comprises one or more metal oxides of a general formula M_xO_y and/or one or more non- metal oxides of a general formula E_zO and any mixtures thereof wherein M is a metal selected from an alkali metal, an alkaline-earth metal, a lanthanide, Al, Ga, In, Sc, Y and the like, x is from 1 to 2 and y is 1 to 3, and E is a non-metal selected from the group comprising a halogen, a chalcogen, N, P, As, B, C, Si and the like, z is from 1 to 2 and b is from 1 to 3. Preferably the oxide glass membrane comprises one of the following Na_zO, Li₂0, K₂0, MgO, CaO, SrO, Al₂0₃, Sc₂0₃, Y₂0₃ La_?0₃, Si0₂, As₂0₃ or any mixtures thereof. According to some embodiments the chalcogenide glass membrane comprises one or more metals, one or more chalcogens, and optionally one or more non-metals wherein the metal is selected from an alkali metal, an alkaline-earth metal, a lanthanide, Al, Ga, In, Sc, Y; the chalcogene is selected from S, Se and/or Te; and the optional non-metal is selected from Si, F, CI, Br, I, N, P, As, B, C; and/or any mixtures thereof.

Preferably the glass membrane has a sensitive surface layer with the thickness being in the range from at least about 1 nm, preferably from about 50 nm to about 100 pm, more preferably from about 100 nm to 5 μητι, and the most preferably from about from about 20 to 500 nm .

A step of soaking of a measuring sensor in a suitable conditioner is a preparation step prior to the step of measuring an output signal originating from the sensor This step allows achieving better properties for the sensors subjected to soaking for a long period of time or at least for a period of time than the measurement is carried out in terms of good reproducibility, high stability and appropriate cross-sensitivity finally resulting in high discrimination capabilities of the multisensor. From a physical point of view the conditioner changes constitution of the surface layer of a membrane therefore accelerating component exchange process on a boarder of a membrane and a liquid solution under evaluation. For example, the conditioner may be used for fulfilling ion content in a surface layer of a sensor. Soaking is deemed to be completed when the output signal originating from the sensor being conditioned stops changing or drifting with time, i.e. the sensor is at the equilibrium conditions with the sample. Usually, the soaking lasts from several minutes to several hours, preferably from about 1 to 60 minutes, more preferably from 1 to 30 minutes, and the most preferably from 1 to 5 minutes.

A type of a conditioner mainly depends on a type of a membrane. Among other factors it depends on a material of the membrane comprised in the sensor. Thus, for example, a conditioner for a measuring sensor with a plasticized membrane may be different from that of a sensor comprising a glass membrane. Furthermore, a liquid solution subjected to taste analysis shall be also taken into consideration when choosing a proper conditioner. According to some embodiments a conditioner is selected from one of lithium sulfate, potassium iodide, maleic acid. A concentration of the conditioner usually varies from Ι Ο^"4 to 10^"1 mol/l. After soaking each measuring sensor is immersed into a liquid sample for a certain amount of time which is enough for collecting a stable output signal from the sensor. Optionally, the signal can be taken as many times as needed to make sure it is not changing with time. Having collected the signals each sensor is optionally rinsed with a rinsing solution to clean the surface of a sensor and preserve it for a later use. This results in achieving longer self life of a sensor.

According to some embodiments a suitable rinsing solution is selected from tris(hydroxyl methyl)amino methane (TRIS) and/or sodium hydrocarbonate (NaHC0₃) in aqueous ethanol solution.

Although the chemical sensors and their properties are at the highest priority for the taste assessment, adequate methods for the data processing are also needed. A signal of a particular sensor having cross-sensitive properties in a multi-component sample is quite complex and includes information about different components which are present in the sample. Thus, one should analyze the outputs of the sensors taken in combination to retrieve said valuable information.

According to some embodiments an output signal of the multisensor is analyzed as the output of the sensor array as a whole using various methods of multivariate analysis. The methods of multivariate analysis include without limitation classical and non-classical methods, parametric and non-parametric methods, linear and non-linear methods such as, for example, principal component analysis (PCA), linear discriminate analysis (LDA), partial least-squares regression (PLS), artificial neural networks, support vector machines, soft independent modeling of class analogy (SIMCA), other various multivariate regressions and the like, which all allow one to deal with multivariate data. Preferably, the methods of multivariate analysis for some applications of the present invention include PCA, LDA and PLS.

Data processing. Methods of multivariate analysis.

The data processing consists of recognition, classification (identification) and multivariate calibration. According to some embodiments recognition is mainly performed by principal component analysis (PCA), classification by linear discriminant analysis (LDA) or by partial least-squares regression (PLS) whilst multivariate calibration is undertaken with the help of PLS. Linear discriminant analysis seeks to find a hyperplane in space (i.e. space of X variables or sensors' output in this case), which separates clusters of points. This hyperplane is defined by the linear discriminant function. When there are more than two groups, several discrimination functions can be computed. The maximum number of functions will be equal to the number of groups minus one, or the number of variables in the analysis, whichever is smaller. Values of these discriminant functions for samples are called scores or roots and can be used for results presentation Principal component analysis and PLS are the so-called projection methods that allow one to efficiently reduce the number of original variables and reject noise. PCA is an unsupervised method in which only a matrix of independent X variables is employed and no additional information is necessary to perform analysis (data decomposition). PCA consists of calculation of a set of new variables, principal components (PCs), which are orthogonal and non- correlated. The first PC corresponds to the direction of the largest variance in the data, etc. Principal components are linear combinations of the original variables with coefficients called loadings. Values that these new variable PCs take for each sample are called scores. Usually the first two or three PCs contain about 80% (or even more) of the variance of the data. Thus, in the majority of cases, experimental data consisting of dozens of variables can be displayed in the form of two- or three-dimensional score plots without significant lost of meaningful information. In PLS regression, both independent X and dependent Y variables are employed, and decomposition of X variables is guided by the structure of Y variable(s). In this way, mostly the information (variance) in X related to the phenomena of interest Y is retrieved. Owing to this feature, PLS can be very a powerful tool not only for multivariate calibration but also for recognition and classification.

Furthermore, the basics of the methods which can be used for the data processing can be found, for example, in one of the following documents Naes N, Martens H (1989) Multivariate calibration, Wiley, New York; Esbensen K (2001 ) Multivatiate analysis in practice, Camo ASA Baker M, Rayens W (2003) J Chernomotrics 17:166-173 incorporated herein in full entity by the reference. Below non-limiting examples of a multisensor according to the present invention and a method for evaluating taste characteristics of some particular samples are disclosed. It should be understood that the examples and embodiments described herein are for illustrative purposes only and that various modifications or changes in light thereof will be suggested to persons skilled in the art and are to be included within the spirit and purview of this application.

EXAMPLES

EXAMPLE 1 : Taste assessment in pharmaceutical compositions.

Materials.

Several types of the samples were studied: (1 ) individual APIs, (2) taste additives with bitter, sweet or salty tastes, and (3) commercial, oral drug preparations. All nonproprietary materials were either Analar or reagent grade. All the substances for assessment were from normal commercial batches that had been cleared for human use. A list of samples measured, with a short description of their tastes and concentrations, are shown in Table 1 .

All the samples were delivered in the solid form as tablets or powders. Aqueous solutions of the samples were prepared for measurement by the multisensor. Doubly distilled water from a single source was used throughout the experiments as sample diluent and for washing the sensors. The concentration range of solutions was chosen to be close to that typically used by taste panels. In Table 1 concentrations of measured solutions are shown in mmol^*L as is traditional for potentiometric sensors and in g^* L ¹ or mass percentages 00145

30

Direct measurements with the sensor array were performed in these solutions without any further pretreatment. At least three replicates of each sample were run in random.

Sensor array and measurement method.

The array of the multisensor used in the present Example consisted of 30 sensors with either modified chalcogenide glass or plasticized PVC membranes, chosen to have a wide ranging cross-sensitivity to multiple inorganic and organic substances. The sensors as used are listed in Table 2. A standard glass pH electrode was also included into the sensor array. All sensors, beside the reference and glass pH electrodes, were fabricated in-house. Potentiometric measurements were performed by using a high input impedance multichannel voltmeter. The measuring process was controlled by a computer, which also collected and stored the potential values of each sensor of the array versus the silver/silver chloride reference electrode. A drawing of one preferable version of the multisensor is shown in the Fig. 1. Measurement time in all solutions was 5 min, which allowed all sensors to reach equilibrium potential. The sensors were washed with distilled water for a few minutes between measurements in order to reach a typical background potential value.

Table 1. List of samples measured with the multisensor.

Samples Taste attributes Concentrations

Individual substances 10-20 mmol L^"1

Quinine 0.5-50 mmol L^"1

Bitter

Caffeine 0.07 mmol L^"1

Drug B (32 mg L^"1)

Aspartame 0.3-10 mmol L^"1

Acesulfam K Sweet 12 and 15 mmol L^"1

(0.2 and 0.24%)

Sucrose 6 and 1.5 mmol L^"1

(0.2 and 50%)

Sodium chloride 35 mmol L^"

Salty

(2.06 g L¹)

Sodium benzoate 15 mmol L

(2.12 g L^"1)

Drug D 10.6 mmol L^"

(2.09 g L^"1)

Binary mixtures 21 +28 mmol L^"1

Caffeine + aspartame (0.4 + 0.8%)

Bitter-sweet

43+14 mmol L^"1

(0.8 + 0.4%)

Quinine + aspartame 3 mmol L^"1 of d rug

3 + 0.1 mmol L^"1

3 + 0.2 mmol L^"1

3 + 0.4 mmol L ¹

Drug A + aspartame 3 mmol L^' of d rug

3 + 0.1 mmol L^"1

3 + 0.2 mmol L^"

3 + 0.4 mmol L ¹

Drug C + aspartame 3 mmol L^"1 of d rug

3 + 0.1 mmol L^"1

3 + 0.2 mmol L^"1

3 + 0.4 mmol L^"

Flavoured drug substances 0.09 mmol L^"1 (32 g L^" )

Quinine with three different 4 and 2 g L^"1

flavouring (1 1 .1 and 5.5 mmol L^"1)

Bitter taste masked

Drug A in tablets flavoured with: 4 g L^"1 (1 1.1 mmol L^"1)

by flavourings (to

peach (both fresh and aged) 4 g L^"1 (1 1.1 mmol L^"1)

different extent)

Strawberry Orange Concentration of flavourings and Placebo tablets with fresh matrix components

peach flavouring corresponding to 4 and 2 g L ¹ Table 2. List of sensors as used in the multisensor.

Composition of membrane

of a chemical Additional

N° material (matrix, active

sensor information compounds)

Oxide glass electrode 30SiCv 30Na_?O-40AI;,O₃ ^' pH glass electrode

2 Silver chloride electrode Ag/AgCI sat. KCI Reference electrode 3 Cross-sensitive (I) where Ri , R₅ are - Cross-sensitive to chemical sensor C(=0)NR₇R₈ , R₇,R₈ are C5 inorganic and

alkyl organic cations 5 %

NaTFPB 2 %

PVC 25 %

DOP 68 %

Cross-sensitive (I) where R-i , R_¾ are - Cross-sensitive to chemical sensor C(=0)NR₇R₈ , Ry,R₈ are C3 inorganic and

alkyl organic cations

5 %

NaTFPB 5%

PVC 25 %

DOP 65 %

Cross-sensitive (I) where R-,, R₅ are - Cross-sensitive to chemical sensor C(=0)NR₇R₈ , R_7lR₈ are C_¾ alkyl inorganic and

5 % organic cations

TDDA-TPB 5%

PVC 25 %

DOP 65 %

Cross-sensitive (I) where R-,, R_¾ are - Cross-sensitive to chemical sensor C(=0)NR_yR₈ , R₇,R₈ are C alkyl inorganic and

1 % organic cations TPB 3 %

PVC 25 %

TOP 71 %

Cross-sensitive (I) where R- R₅ are - Cross-sensitive to alkyl, R. inorganic and organic cations

Cross-sensitive to inorganic and organic cations

Cross-sensitive NaFTPB 3 % Cross-sensitive to chemical sensor PVC 27 % inorganic and

DOA 70 % organic cations

Cross-sensitive KTPB 3 % Cross-sensitive to chemical sensor PVC 27 % inorganic and

DOP 70 % organic cations I Cross-sensitive CCD 10 % Cross-sensitive to chemical sensor PVC 20 % inorganic and

DOP 70 % organic cations I Cross-sensitive TDDABr 3 % Cross-sensitive to chemical sensor PVC 27 % inorganic and

DOP 70 % organic anions

Cross-sensitive THDABr 3 % Cross-sensitive to chemical sensor PVC 27 % inorganic and

CA 70 % organic anions I Cross-sensitive TPP-TPB 2 % Cross-sensitive to chemical sensor PVC 28 % inorganic and

TOP 70 % organic substances I Cross-sensitive TDDA-TPB 2 % Cross-sensitive to chemical sensor PVC 27 % inorganic and

DOA 70 % organic substances I Cross-sensitive DMDODABr 2 % Cross-sensitive to chemical sensor PVC 27 % inorganic and

CA 70 % organic anions I Cross-sensitive CCD 10 % Cross-sensitive to chemical sensor PVC 20 % inorganic and

CA 70 % organic cations I Cross-sensitive PORPH 3 % Cross-sensitive to chemical sensor PVC 27 % inorganic and

TOP 70 % organic cations

Cross-sensitive PORPH 3 % Cross-sensitive to chemical sensor CCD 2 % inorganic and

PVC 25 % organic cations DOA 70 % 23 Cross-sensitive PORPH 3 % Cross-sensitive to chemical sensor TDDA-TPB 2 % inorganic and

PVC 25 % organic cations DOP 70 %

24 Selective chemical AgCI-Ag₂S Selective to chloride sensor ion in most of

solutions

25 Selective chemical Pt, metallic Sensitive to redox sensor activity in most of solutions

26 Selective chemical Ag₂S Sensitive to redox sensor activity in most of solutions

27 Selective chemical Sb, metallic Sensitive to redox sensor activity in most of solutions

28 Cross-sensitive Ag₁₈Cr₂As_3SSe₁₇Te_?8 ^* Cross-sensitive to chemical sensor heavy metals with enhanced selectivity to chromium(VI)

29 Cross-sensitive

Cross-sensitive to chemical sensor heavy metals with enhanced selectivity to mercury(ll)

30 Selective chemical Cu₁₀Ag₁₅As„_bSe₃ Sensitive to heavy sensor metals with

enhanced selectivity to copper(ll) composition is expressed in atomic %

"KTPB" = Potassium tetrakis(4-chlorophenyl)borate

"NaFTPB" = Sodium tetrakis(4-fluorophenyl)borate

"TDDA-TPB" = Tetradodecylammonium tetrakis(4-chlorophenyl)borate

"CCD" - chlorinated cobalt dicarbollide

"DOP" = Bis(2-ethylhexyl) phthalate "TOP" = Tris(2-ethylhexyl) phosphate

"DOA" = Bis(2-ethylhexyl)adipate

"CA" = Cetyl alcohol

"PVC" = poly(vinyl chloride)

"TDDABr" = tetradodecylammonium bromide

"THDABr" = tetrahexadecylammonium bromide

"TPP-TPB" = Tetraphenylphosphonium tetraphenylborate

"DMDODABr" = Dimethyldioctadecylammonium bromide

"PORPH" = 5, 10, 15,20-tetraphenylporphyrin

Data processing.

All calibration models, either for classification or quantitative determination, were validated by using test sets. This means that all experimental data were split in two parts: one of which (measured earlier) was used to produce calibration models; the other, usually measured later on during the experiment, was used as test (i.e. to evaluate obtained calibration models). Usually 2/3 of the experimental data were included in the calibration set and about 1/3 in the test set. All results below are shown for the test data set unless otherwise stated. The Unscrambler software (CAMO ASA, Norway, v.7.8) was used for the data processing. However a specialist skilled in the art should appreciate that any software program suitable for the data processing can be applied

Objectives The following tasks were addressed in the course of the study (highlighted in bold):

1. Discrimination between bitter, sweet and salty substances with the multisensor; examination of possible memory effects of bitter, sweet or salty substances on the sensors.

The multisensor was applied to the discrimination of substances with different taste modalities (i.e. bitter, sweet and salty). The measurements were made in individual solutions of sodium benzoate, sodium chloride, drugs A, B and D , acesulfam K (ASK), sucrose, aspartame, quinine and caffeine. Drugs A, B and D were the pharmaceuticals which taste modalities weren't intentionally disclosed before the analysis. Concentration of each substance was 10 mmol L- . Data were processed by using LDA. Two discriminant functions were calculated: the resulting scores or roots for all samples are shown in Fig. 6 (a). Groups of points on this plot (and plots below) visualise the classification of corresponding samples as "perceived" by the multisensor. The distances between points can be considered, in the first instance, as an approximate measure of the difference between samples. The distance between all salty samples and all other ones is relatively big, whereas bitter and sweet samples are lying somewhat closer to each other. The samples closest to the border between classes with sweet and bitter tastes appeared to be aspartame and caffeine, correspondingly. However each class can still be clearly distinguished.

Distinct discrimination of sweet and bitter substances can be further confirmed by numerical classification. A classification model was made in this case by using LDA. Two classes, bitter and sweet, were modelled. Modelling was performed with a part of the data, namely the calibration data set. The obtained model was validated by using the test data set Discrimination functions are used in LDA for calculation of the scores for each sample for each class. Scores calculated for both classes for all samples, including calibration and test data sets, are shown in the Fig. 6 (b). The sample is assigned to the class for which it has the smallest score. Thus, the border between two classes "bitter" and "sweet" can be shown as the line y = x. Prediction was correct in all cases. It is important to point out that this classification was performed for a wide number of different substances eliciting bitter and sweet taste (five bitter and three sweet samples) The multisensor was capable of correctly classifying all the samples, even though the substances with the same taste modality might have had different a chemical nature.

Measurements with the sensor array were made in random order during these experiments (i.e. salty or sweet samples were measured after bitter ones and vice versa). We observed that the order of measurements does not influence reproducibility of the response of the sensors or discrimination power of the array (Figs. 6 (a) and (b)). Thus, the sensors used in the array of the multisensor did not suffer any memory effects in any of the studied substances.

2. Discrimination between different substances eliciting the same taste (i.e. salty, sweet and bitter). Assessment of the capability of the multisensor to quantify the content of selected bitter and sweet substances and to identify correctly drug preparations containing active substance and corresponding placebo.

The next step was aimed at further recognition of different substances with the same basic taste modality. The multisensor was capable of discriminating three different salty substances (Fig. 7(a)). Three different substances eliciting sweet taste were also distinguished (Fig. 7(b)). On the latter score plot (Fig. 7(b)), about 98% of total variance is explained by the first component. This is mainly related to the fact that the "multisensor- perceived" difference between ASK and two other sweet substances is significantly bigger than between sugar and aspartame. However, the distance between the aspartame and sugar clusters on the score plot is still longer than between replicated samples of the same substance in each cluster. Thus, the multisensor can reliably distinguish not only ASK, but all three sweet substances. An advanced experiment was performed with four bitter materials having different intensities and types of bitterness. The results of the data processing are shown in Fig. 7(c) as a PCA score plot. The intensity of bitterness as perceived by the taste panel was as follows: drug B>quinine>caffeine»drug A. Drug A had relatively little bitterness. At first sight, neither of the first two principle components in Fig. 7 (c) coincide with the direc- tion of bitterness intensity, which appears to be somewhere in between PC1 and PC2. It is interesting that the bitterness character seems to be picked up on PC2: drug A and quinine both have fairly "clean" bitterness, which has only a slight lingering character; caffeine lingers longer; and drug B is quite short lived, albeit intense. 3. Assessment of capability of the MULTISENSOR to quantify the content of selected bitter and sweet substances and to identify correctly drug preparations containing active substance and corresponding placebo.

Caffeine is often used along with quinine as reference substance for sensory panels, and the bitter taste of other substances can be expressed in units of either of these standards. Aspartame is often used to render bitter-tasting drugs more palatable. Since application of the multisensor to quantification of bitterness and to evaluation of bitter taste-masking approaches is of major interest, the sensitivity of the sensor array to caffeine, quinine and aspartame was thoroughly studied. Firstly, measurements were made on the individual substances in solution at typical concentration ranges relevant to human use These were 0.5-50 mmol L ¹ for caffeine, 10-20 mmol L ¹ for quinine and 0.3-10 mmol L ¹ for aspartame. Numerical calibration models were made by using PLS1 . Parameters of predicted versus measured curves for both calibration and test samples are summarised in Table 3. Parameters of these curves are close to expected values (i.e. slopes and correlation coefficients are around unity and offsets are close to zero). Thus, the sensors of the array exhibit good enough sensitivity to bitter substances and to the sweet substance, ensuring multisensor capability not only to differentiate them, but also to quantify their content.

Quantitative response to the bitter-tasting drug A was also observed. The measurements were made in solutions prepared by dissolving commercial tablets that contained 250 and 500 mg of drug A and the same flavouring. The resulting concentrations of drug A in solutions were 5 and 10 mmol L ¹ , respectively. Solutions of placebo tablets comprising the same excipients and flavourings but no drug at all were also prepared The multisensor was capable of distinguishing samples with different content of drug A (Pig. 8) and the solution made from placebo tablets with zero concentration of drug A. Though there are too few points to make a representative calibration plot, the promise for quantification of the drug A content by using themultisensor is evident.

Table 3. Parameters of prediction versus measured curves produced with the help of the multisensor for quantitative determination of quinine, caffeine and aspartame

Substance Slope Off-set Corre

Quinine

Calibration 1 .00 0.00 1 .00

Prediction 1 00 0.09 1 .00

Caffeine

Calibration 0.99 -0.02 1 .00

Prediction 0.97 -0.09 0.97

Aspartame

Calibration 1 .00 -0.01 1 .00

Prediction 0.96 -0.12 0.99

4. Evaluation of different taste-masking approaches (e.g. how the additions of different quantities of sweeteners and flavourings helps to eliminate the bitter taste of a drug)

The possibility of applying the multisensor to evaluation of taste-masking effects was investigated Measurements with the multisensor were made in the solutions of drug samples containing various flavourings with the aim to evaluate whether the effect of flavouring could be identified by the instrument. The solutions were prepared by using commercial tablets, which had the same content of drug A and one of three flavourings: orange, strawberry, and peach. Peach flavours were supplied both freshly prepared and after ageing in a controlled temperature and humidity environment. Although it is known that the content of drug A in these tablets is not affected, the peach flavour, which originally had the most acceptable taste of the four, was found to have changed character during this storage. Samples with "aged peach" flavour were much less organoleptically acceptable to a human taste panel than any of the other flavoured samples. Classification of the data was performed in this case by using corresponding to four flavourings. A plot of resulting PLS2 scores is shown in Fig. 9(a). All flavourings can be reliably discriminated by the multisensor. Moreover, positions of samples on the score plot correlate with organoleptic properties of these formulations determined earlier by the sensory panel. Peach-flavoured tablets masked the bitter taste of drug A most effectively, but with time the peach flavour partly decomposes and loses its masking properties. Correspondingly, the samples with fresh peach flavouring are located further from the others on the score plot (Fig.9 (a)), whereas the sample with aged peach flavour is closer to the samples with orange and strawberry flavours, which both masked the bitterness of drug A less effectively.

A similar study was performed by using the multisensor to evaluate quinine solutions both with and without flavourings. The results are shown on a PCA score plot in Fig. 9(b). PC1 in this plot roughly coincides with the direction in which the change of bitterness of the sample occurs. The unflavoured quinine substance (coded 1 ) is located at the right edge of the plot; the less bitter quinine tablet without flavouring (coded 2) comes next. Sample 3 with poorly masked bitter taste is located further to the left, and sample 4, with well- masked bitter taste, appears on the opposite side of the plot. Thus, the relative positions of the samples on this plot are logical and correlate to their perceived bitterness. 5. Quantification of the effect of taste masking of bitter substances by sweet ones.

As discussed in the previous section, the masking of the bitter taste of drugs by sweeteners is an important part of the drug development process. For an analytical instrument to be useful in such a process, it should be capable of detecting not only bitterness intensity, but also the decrease of bitterness after the addition of sweetener. This means that an instrument should detect "apparent" or "perceived" bitterness (i.e. evaluate actual taste changes after addition of sweetener to bitter substance, when total concentration of bitter substance remains the same). To be able to accomplish this task an instrument should display cross-sensitivity to a wide range of bitter and sweet substances. The ability of the multisensor to "perceive" the effect of mixing of bitter and sweet substances was studied in mixed solutions of bitter substances: caffeine, quinine, drugs A and C and the sweetener aspartame. The capability of the multisensor to evaluate masking of bitter taste was also studied in binary solutions of three drugs with different added quantities of aspartame: 0.5, 1 or 2 mmol L^~1 of aspartame with constant content (15 mmol L^" ) of a drug— either quinine, drug A or drug C. The PLS score plot of these samples is shown in the Fig. 10(a). The samples of each bitter drug are located in a logical order from I to 111, reflecting the increase of the aspartame concentration or decrease in perceived bitterness. Location of the series of drug samples is correlated to bitterness intensity. It is the highest with drug C, whose bitter taste was not masked even by the highest concentration of aspartame, and the weakest with drug A, whose bitterness was already partly masked by the first addition of aspartame. Quinine samples exhibiting intermediate bitterness are located in between the drug C and drug A series. It is also worth emphasising that the spread of the samples containing drug A with the least bitterness is the smallest. The two preparations II and III with drug A are located close to each other on the score plot, since in mixture II (1 mmol L ¹ of aspartame with 5 mmol L. of drug) the bitter taste of the drug A is already well-masked and a further increase of aspartame content does not induce significant extra masking effect. It is worth noting that in this particular case, the direction of the first PC on the plot (Fig. 10(a)) roughly coincides with direction of bitterness intensity change.

Furthermore, an attempt to quantify suppression of bitterness by using the multisensor was performed in mixed binary solutions of quinine and caffeine with aspartame. Measurements with the multisensor were made in individual solutions of caffeine and aspartame and in mixed solutions of both substances. Data fitting was performed by using PLS2, concentrations of caffeine and aspartame being independent variables. The PLS score plot of these samples is shown in Fig. 10 (b). Caffeine and aspartame are clearly distinguished and each of their mixtures is located relatively close to the substance, which is present at higher concentration (e.g. the sample containing 0.2% of caffeine and 0. 1 % of aspartame is closer to pure caffeine samples).

The results of the multisensor calibration made by using the individual solutions of quinine hydrochloride were applied to assess the bitterness of binary mixtures of quinine with aspartame. Obviously, the mixtures of quinine with a sweetener (e.g. aspartame) exhibit less perceived bitterness than individual quinine solutions (i.e. "apparent" quinine concentration should be lower than the actual one). Calibration data obtained in individual quinine samples (Table 3) was used for predicting "apparent" or "perceived" quinine content in the mixed samples. The results are summarised in Table 4. As expected, the apparent concentration of quinine predicted by the multisensor for the mixed solutions is lower than the actual one and depends on the content of the aspartame, the higher the aspartame concentration, the lower the predicted value of apparent quinine concentration. Thus, apparent quinine concentration might be considered as a measure of perceived bitterness that depends, beside actual quinine presence, on the content of masking substance. Quinine is often used as a reference bitter substance for sensory panels, and bitterness of other substances evaluated by the multisensor might be expressed in quinine units. The possibility to use the multisensor in this way is very promising for formulation development. However, further efforts are necessary for building a generally applicable bitterness scale. In particular, measurements of a wider range of bitter substances whose bitterness character and intensity have been assessed and quantified by humans will be necessary for adequate calibration of the multisensor

Table 4. The multisensor prediction of "apparent" quinine concentration in binary mixed solutions of quinine and aspartame using calibration of the sensor array in the individual quinine solutions.

Actual composition of the samples "Apparent" concentration of bitter

substance (mmol L ¹ )

Predicted Standard deviation

Quinine 5 mmol L + 6 0,7

aspartame 2 mmol L ¹ (I

Quinine 1 5 mmol L ¹ + 7,7 0,6

aspartame 1 mmol L ¹ (I

Quinine 1 5 mmol L ¹ + 1 0 0,6

aspartame 0,5 mmol L ¹

Conclusions

The "multisensor" comprising an array of 30 different chemical sensors has been applied to analyse a variety of 41 individual substances and mixtures of particular interest for pharmaceutical research and development. The multisensor was capable of discriminating between substances with different taste modalities . It could also distinguish different substances eliciting the same basic taste (e.g . bitter) and shows promise in quantifying the content of each substance and nuances of the basic taste (e.g. lingering or short-lived). Different approaches to masking the bitter taste of drugs were evaluated by using the multisensor. I n all cases it was possible to establish correlation between masking effect/bitterness strength and the response of the sensor array of the multisensor. After calibration in individual quinine solutions the multisensor was successfully applied to the prediction of bitterness strength of binary mixtures with a sweetener in terms of "apparent" or "perceived" quinine content. Therefore, the multisensor proved to be a valuable tool for assessment and prediction of the taste of pharmaceuticals and related products. The system could potentially assist, or even replace, a sensory panel in certain types of routine analysis in pharmaceutical development and production This could have the benefit of providing the pharmaceutical formuiator with reliable data concerning the taste of the product quickly and without the need to ask volunteers to taste active pharmaceutical samples. Early development activities could be facilitated when human tasting is usually not possible in the absence of the required toxicological data.

Claims

1 . A multisensor for assessment of taste characteristics of a liquid sample comprising one or more components, the multisensor comprising

(I)

where

Ri , R_?, R3_> i, R.¾ ^may be same or different and are independently hydrogen; halogen; hydroxyl;

Ci .₆ alkyl optionally substituted with hydroxyl, C^alkoxy or at least one halogen atom; C₃_i₀cycloalkyl optionally substituted with hydroxyl, ₆alkyl, C ₆alkoxy or at least one halogen atom; aryl optionally substituted with Ci.₆alkyl, at least one halogen atom, hydroxyl, C_{1 6}alkoxy;

5-6-membered oxygen-containing heterocyclic _{l 6}alkyl; 5-6-membered heteroaryl comprising 1 -3 heteroatoms selected from O, N or S; -C(=0)OH; -C(-0)OR₆ where R₆ is Ci 6 alkyl optionally substituted with hydroxyl or halogen;

with the proviso that

at least one of R, , R₇, R₃, R₄, R_b is -C(=0)NR,R₈ where

R_{/ t} R₈ are independently hydrogen, Ci , alkyl optionally substituted with hydroxy!, Ci _yaikyl, ₆alkoxy or one or more halogens; aryl optionally substituted with C alkyl, one or more halogens, hydroxyl, C-i _Balkoxy; or R_y, R₈ may together with the nitrogen atom to which they are attached form a 5-6-membered heterocyclic group optionally substituted with Ci _/alkyl, Ci_₆alkoxy, one or more halogen atom, or phenyl;

and/or any mixtures thereof.

2. The multisensor according to claim 1 , wherein said array is selected from a broader sub-group of sensors adapted for a particular sample.

3. The multisensor according to claim 1 , wherein both R-, and R₅ are -C(0)NRyR₈ where R₇, R₈ are independently hydrogen; C _-7 alkyl optionally substituted with hydroxyl,

C-i.yalkyl, C-^alkoxy or at least one halogen; aryl optionally substituted with C alkyl, at least one halogen, hydroxyl, C^alkoxy; or

R₇, R₈ may together with the nitrogen atom to which they are bonded form a 5-6- membered heterocyclic group optionally substituted with d^alkyl, C, _f;alkoxy at least one halogen atom or phenyl.

4. The multisensor according to claim 1 , wherein the active component is selected from the group comprising N,N',N,N'-tetraethylpyridine-2,6-dicarboxamide;

N,N',N,N'-tetrabutylpyridine-2,6-dicarboxamide;

N,N'-ditolyl-N,N'-diethylpyridine-2,6-dicarboxamide ;

N,N'-diphenyl-N,N'-dimethylpy dine-2,6-dicarboxamide;

N,N'-diphenyl-N,N'-dibenzylpyridine-2,6-dicarboxamide;

N,N'-dipropyl-N,N'-di(4-butylphenyl)pyridine-2,6-dicarboxamide;

4-chloro-N, N' , N ,N'-tetraethylpyridine-2,6-dicarboxamide ;

4-(2-oxo-2-phenylethyi)-N,N',N,N'-tetraethylpyridine-2,6-dicarboxamide;

4-[2-(Biphenyl-4-yl)-2-oxoethyl]-N,N',N,N'-tetraethylpyridine-2,6-dicarboxamide;

6-diethylcarbamoyl-4-(2-hydroxy-2-phenylvinyl)pyridine-2-carboxylic acid methyl ester; 2,2'-dipyridyl-6,6'-dicarboxylic acid diamides; and/or

any mixtures thereof.

5. The multisensor according to claim 1 , wherein at least one of the at least two measuring sensors comprises as the matrix a glass membrane or a crystalline (single or poly) membrane.

6. The multisensor according to claim 5, wherein the glass membrane is selected from oxide glass membrane and chalcogenide glass membrane.

7. The multisensor according to claim 6, wherein each of the said oxide glass membrane and chalcogenide membrane has a sensitive surface layer with the thickness in the range from at least about 1 nm.

8. The multisensor according to claim 6, wherein the oxide glass membrane comprises one or more metal oxides of a general formula M_xO_y and/or one or more non-metal oxides of a general formula E,.O_b and any mixtures thereof, wherein

M is a metal selected from an alkali metal, an alkaline-earth metal, a lanthanide, Al, Ga, In, Sc, Y and the like,

x is from 1 to 2 and y is 1 to 3, and

E is a non-metal selected from the group comprising a halogen, a chalcogen, N, P,

As, B, C, Si and the like,

z is from 1 to 2 and b is from 1 to 3.

9. The multisensor according to claim 8 wherein the oxide glass membrane comprises one of the following oxides Na₇0, Li₂0, K_?0, MgO, CaO, SrO, Al_?0_{3 l} Sc_?0₃, Y_?0₃ La_?0_{3 l}

SiO_?, As₂0₃, Zr0₂, MoO_?, W0₃ or any mixtures thereof.

10. The multisensor according to claim 6, wherein the chalcogenide glass membrane comprises one or more metal chalcogenides of a general formula M_xC_y and/or one or more non-metal chalcogenides of a general formula E₇C_b wherein

M is a metal selected from an alkali metal, an alkaline-earth metal, a transition metal, a lanthanide or an actinide;

C is a chalcogen selected from S, Se and/or Te; and

E is a non-metal selected from Si, As, P, Sb, Ge, Sn, I;

x is from 1 to 3 and y is from 1 to 5; and

z is from 1 to 3 and b is from 1 to 5; and/or

any mixtures thereof

1 1 . The multisensor according to claim 10, wherein

is selected from Na, Ca, Cu, Cd, Ag, Pb, Cr, C is selected from As, Se, S; and

E is selected from I , Si.

12. The multisensor according to claim 5, wherein the crystalline (single or poly) membrane comprises one or more metal chalcogenides of a general formula M_xC_y wherein

M is a metal selected from an alkali metal, an alkaline-earth metal, a transition metal, a lanthanide or actinide;

C is a non-metal selected from a chalcogenes and/or halogens;

x is from 1 to 3 and

y is from 1 to 7.

13. The multisensor according to claim 12, wherein

M is selected from Na, Ca, Ag, Pb, Cu, Cd, Cr; and

C is selected from S, Se, Te, F, CI, Br, I.

14. The multisensor according to claim 12, wherein the crystalline (single or poly) membrane comprises additionally one or more metals such as Pt, Au, Ag, Sb, Pd, Rh, Cr, Mo or the like.

15. The multisensor according to claim 1 , wherein the said active sensing compound is absorbed, adsorbed, dissolved, dispersed, bonded, embedded or otherwise introduced into/onto the matrix.

16. The multisensor according to claim 1 , wherein at least one of the at least two sensors represents a potentiometric electrode, a field effect transistor, light-addressable potentiometric sensor, quartz piezoelectric devise, acoustic wave device sensor.

17. The multisensor according to claim 1 , wherein the taste characteristics are selected from the group comprising bitterness, saltiness, sourness, sweetness, umami and/or any combination thereof.

18. A method for evaluating taste characteristics of a liquid sample, the method comprising the following steps.

(a) providing a multisensor according to claim 1 ; (b) washing each measuring sensor with a washing solution suitable for the sensors and the analyte being measured;

(f) processing the data.

19. The method according to claim 18, wherein the multisensor additionally includes a sensor comprising a glass membrane or a sensor comprising a crystalline (single or poly) membrane.

20. The method according to claim 19, wherein the glass membrane is either oxide glass membrane or chalcogenide glass membrane.

21 . The method according to claim 20, wherein the glass membrane has a sensitive surface layer with the thickness being in the range from at least about 1 nm.

22. The method according to claim 18, wherein at least one of the at least two sensors represents a potentiometric electrode, a field effect transistor, light-addressable potentiometric sensor, quartz piezoelectric devise, acoustic wave device sensor.

23. The method according to claim 18, wherein the conditioner is selected from 10⁴ to 10 ¹ M lithium sulfate, lC to 10 M potassium iodide, 10 ^b to 0 ⁷ malic acid and/or any mixtures thereof.

24. The method according to claim 18, wherein it additionally comprises the step of rinsing a sensor with a rinsing solution.

25. The method according to claim 24, wherein the rinsing solution is selected from 0 to 10 ¹ M tris(hydroxymethyl)aminomethane, 10^" to 10 M NaHCO_a in aqueous ethanol solution.

26. The method according to claim 1 8, wherein water is at least doubly distilled (bidistilled) water.

27. The method according to claim 18, wherein the taste characteristics are selected from the group comprising bitterness, saltiness, sourness, sweetness, umami and/or any combination thereof.

28. An array comprising at least two cross-sensitive sensors comprising an active sensing compound in a matrix, wherein at least one of the said at least two sensors comprises as its active compound a compound of a general formula (I):

(I)

where

Ri , R_?, R3, -1 , R5 rnay be same or different and are independently hydrogen; halogen; hydroxyl;

C ₆ alkyl optionally substituted with hydroxyl, C-i ₆alkoxy or at least one halogen atom; C_{3 K}jcycioalkyl optionally substituted with hydroxyl, d , ₆alkyl, Ci.₆alkoxy or at least one halogen atom; aryl optionally substituted with C-, ₆alkyl, at least one halogen atom, hydroxyl, C-i ₆alkoxy;

5-6-membered oxygen-containing heterocyclic, ₆alkyl; 5-6-membered heteroaryl comprising 1 -3 heteroatoms selected from O, N or S; -C(^~0)OH; C(-^"0)OR₆ where R₆ is C, .ø alkyl optionally substituted with hydroxyl or halogen;

with the proviso that

at least one of R R₂, R₃, R₄, R_s is -C(=0)NR_yR₈ where

R , R₈ are independently hydrogen, C, _/ alkyl optionally substituted with hydroxyl, C-, ^a!kyl, ₆alkoxy or one or more halogens; aryl optionally substituted with Ci _/alkyi, one or more halogens, hydroxyl, C, _Galkoxy; or R₇, R₈ may together with the nitrogen atom to which they are attached form a 5-6-membered heterocyclic group optionally substituted with C-, _yalkyl, Ci_₆alkoxy, one or more halogen atom, or phenyl;

and/or any mixtures thereof.

29. A membrane for use in a measuring sensor, the membrane comprising a plasticized polymer matrix, and an active compound in the matrix, wherein said active compound represents a compound of a general formula (I)

Ri , Rz, R3, 4, Rs may be same or different and are independently hydrogen; halogen; hydroxyl;

C1.6 alkyl optionally substituted with hydroxyl, C-^alkoxy or at least one halogen atom; C₃-i₀cycloalkyl optionally substituted with hydroxyl, ₆aikyl, d _ealkoxy or at least one halogen atom; aryl optionally substituted with Ci .₆alkyl, at least one halogen atom, hydroxyl, C^alkoxy;

5-6-membered oxygen-containing heterocyclylC_{1 6}alkyl; 5-6-membered heteroaryl comprising 1 -3 heteroatoms selected from O, N or S; -C(=0)OH; -C(=0)OR₆ where R₆ is C^e alkyl optionally substituted with hydroxyl or halogen;

with the proviso that

at least one of R^ R₂, R₃, R₄, Rs is -C(=0)NR₇R₈ where

R₇, R₈ are independently hydrogen, Ci .₇ alkyl optionally substituted with hydroxyl, ^alkyl, C^ealkoxy or one or more halogens; aryl optionally substituted with Ci„₇alkyl, one or more halogens, hydroxyl, C_{1 B}a!koxy; or

R_y, R₈ may together with the nitrogen atom to which they are attached form a 5-6-membered heterocyclic group optionally substituted with C, _/alkyl, C-i ₆alkoxy, one or more halogen atom, or phenyl; and/or any mixtures thereof.

30. A method for evaluating taste-masking efficiency of a sample, the method comprising evaluating the taste characteristics of the original sample and the taste-masked sample using the multisensor according to claim 1.