Information processing device, information processing method, program, and drug evaluation method

Abstract

The information processing device is provided with a processor. The processor acquires a plurality of unknown waveform data, the judgement result of inferiority or superiority of which is unknown based on similarity with an ideal waveform, judges the inferiority or superiority of each of the plurality of unknown waveform data based on a plurality of teacher waveform data linked to judgement results of inferiority or superiority, and outputs the inferiority or superiority of the plurality of unknown waveform data in a comparable manner.

Description

[Technical Field]

[0001] The technology of the present disclosure relates to an information processing device, an information processing method, a program, and a drug evaluation method. [Background technology]

[0002] To improve the efficiency of new drug development, a method for evaluating toxicity using cells such as cardiomyocytes generated from iPS (induced pluripotent stem) cells has been developed (see, for example, International Publication No. 2019 / 131806). Toxicity evaluation is performed by evaluating the responsiveness of cells to drugs.

[0003] For example, a well plate with multiple wells is used as a culture vessel for cardiomyocytes. A microelectrode array (MEA) consisting of multiple microelectrodes is provided on the bottom surface of each well. Such a well plate is called an MEA plate. Each microelectrode of the microelectrode array outputs a waveform indicating electrophysiological changes in the cells cultured in the well (for example, a myocardial waveform indicating the pulsation of cardiomyocytes). Toxicity assessment is performed by measuring changes in the waveform in response to a drug.

[0004] The waveforms output from each of the multiple microelectrodes installed in a well vary. For this reason, for each well, one microelectrode that outputs the most ideal waveform is selected as the target electrode for toxicity evaluation. This selection of the target electrode is also called the selection of the golden channel (see, for example, Reference 1). For example, in the case of cardiomyocytes, the waveform closest to the healthy state is selected from the myocardial waveforms output from each microelectrode based on known myocardial waveforms representing a healthy state (e.g., a state free of diseases such as arrhythmia) obtained from previous measurements. Reference 1: Seeding and culturing iCell® cardiomyocytes 2.0 on MED plates, [Retrieved January 27, 2021], Internet<https: / / alphamedsci.com / download / protocols / dissociated5_190625%20(J).pdf> Summary of the Invention [Problem to be solved by the invention]

[0005] Since there is no clear definition of the ideal waveform that serves as a standard for selecting target electrodes, and only a rough shape is determined, it is difficult to mechanically select target electrodes based on the waveform output from each microelectrode. For this reason, target electrodes are currently selected by human sensory evaluation, in which the waveform output from each microelectrode is compared with a known ideal waveform.

[0006] However, such selection of target electrodes based on sensory evaluation has problems such as the evaluation taking time and variations in the evaluation depending on the experience of the evaluator, etc. The above-mentioned document 1 describes selecting target electrodes based on the peak amplitude of the waveform output from each microelectrode, but with such a simple mechanical method, it is difficult to accurately select a waveform that is close to the ideal waveform.

[0007] Therefore, there is a need for a method that can select a waveform close to the ideal waveform in a short time while maintaining the accuracy of waveform selection achieved by conventional sensory evaluation. The problem of selecting a waveform close to the ideal waveform is not limited to the field of toxicity evaluation, but exists in various fields.

[0008] The technology disclosed herein aims to provide an information processing device, an information processing method, and a program that enable a waveform close to an ideal waveform to be selected with high accuracy in a short time. [Means for solving the problem]

[0009] In order to achieve the above object, the information processing device of the present disclosure includes a processor that acquires multiple unknown waveform data whose superiority / inferiority judgment results based on similarity to an ideal waveform are unknown, and judges the superiority / inferiority of each of the multiple unknown waveform data based on multiple teacher waveform data to which the superiority / inferiority judgment results are linked, and outputs the superiority / inferiority of the multiple unknown waveform data in a manner that allows comparison.

[0010] It is preferable that the processor clusters a set including a plurality of teacher waveform data and a plurality of unknown waveform data, and, as a result of the clustering, determines the probability that the unknown waveform data is superior for each cluster that includes at least one of the plurality of unknown waveform data, thereby determining superiority or inferiority.

[0011] The processor preferably calculates the probability for each cluster based on the number of teacher waveform data with a good decision and the number of teacher waveform data with a good decision and a poor decision.

[0012] The probability is expressed as the number of teacher waveform data judged as good divided by the number of teacher waveform data judged as good and bad, and it is preferable that the processor outputs a comparative ranking of the merits of multiple unknown waveform data based on the probability.

[0013] The processor preferably performs clustering using the k-medoids or k-means method, although other clustering algorithms may be used.

[0014] The processor preferably performs a filtering process to remove unknown waveform data that does not meet the evaluation criteria from the set or to lower the ranking of the unknown waveform data.

[0015] Preferably, the processor inputs the unknown waveform data into a neural network that has undergone machine learning based on teacher waveform data, and determines the superiority or inferiority based on the results output from the neural network.

[0016] It is preferable that the processor inputs unknown waveform data into the encoder of an autoencoder that has undergone machine learning based on superior-determined teacher waveform data, and then determines whether the data is superior or inferior based on the difference between the waveform data restored by the decoder and the unknown waveform data input to the encoder.

[0017] The unknown waveform data is preferably a pulse signal output by a cell.

[0018] Examples of cells include nerve cells, cardiac muscle cells, skeletal muscle cells, and smooth muscle cells, with cardiac muscle cells being preferred.

[0019] The drug evaluation method according to the technology of the present disclosure is a drug evaluation method that evaluates drugs based on information output from the above-mentioned information processing device, and it is preferable that unknown waveform data with a high ranking in terms of merit is used for drug evaluation.

[0020] The information processing method according to the disclosed technology acquires multiple unknown waveform data whose superiority / inferiority judgment results based on similarity to an ideal waveform are unknown, and uses machine learning to judge the superiority / inferiority of each of the multiple unknown waveform data based on multiple teacher waveform data to which the superiority / inferiority judgment results are linked, and outputs the superiority / inferiority of the multiple unknown waveform data in a manner that allows comparison.

[0021] The program relating to the technology of the present disclosure causes a computer to execute an acquisition process for acquiring multiple unknown waveform data whose superiority / inferiority judgment results based on similarity to an ideal waveform are unknown, a judgment process for judging the superiority / inferiority of each of the multiple unknown waveform data using machine learning based on multiple teacher waveform data to which the superiority / inferiority judgment results are linked, and an output process for outputting the superiority / inferiority of the multiple unknown waveform data in a manner that allows them to be compared. [Effects of the Invention]

[0022] According to the technology of the present disclosure, it is possible to provide an information processing device, an information processing method, and a program that enable a waveform close to an ideal waveform to be selected with high accuracy in a short time. [Brief explanation of the drawings]

[0023] [Figure 1] FIG. 1 is a diagram schematically illustrating an electrical function evaluation system. [Figure 2] FIG. 1 is a perspective view showing an example of an MEA plate. [Figure 3] FIG. 1 is a diagram showing an example of a well. [Figure 4] FIG. 1 shows an example of a microelectrode array. [Figure 5] FIG. 2 is a block diagram illustrating an example of a hardware configuration of the electrical function evaluation system. [Figure 6] FIG. 2 is a block diagram illustrating an example of a functional configuration of the information processing device. [Figure 7] FIG. 1 is a diagram showing an example of a myocardial waveform. [Figure 8] FIG. 10 is a diagram showing an example of unknown waveform data output from a microelectrode array. [Figure 9] 10 is a flowchart illustrating an example of a determination process. [Figure 10] FIG. 10 is a diagram showing an example of a set including a plurality of unknown waveform data and a plurality of teacher waveform data. [Figure 11] FIG. 10 is a diagram illustrating an example of clustering. [Figure 12] FIG. 1 is a diagram illustrating a clustering technique. [Figure 13] FIG. 10 is a diagram illustrating a method for calculating a score. [Figure 14] FIG. 10 is a diagram showing an example of the ranking of unknown waveform data ranked based on scores. [Figure 15] FIG. 10 is a diagram showing an example of final rankings after re-clustering. [Figure 16] FIG. 10 is a diagram showing an example of displaying a plurality of unknown waveform data in a manner that allows comparison of their merits. [Figure 17] FIG. 10 is a diagram showing a set including a plurality of teacher waveform data and a plurality of unknown waveform data prepared in the embodiment. [Figure 18] FIG. 10 is a diagram showing the results of the first clustering performed in the example. [Figure 19]FIG. 10 is a diagram showing the results of the second clustering performed in the example. [Figure 20] FIG. 10 is a diagram showing the final ranking of unknown waveform data obtained in an example. [Figure 21] FIG. 1 is a diagram illustrating a neural network learning method. [Figure 22] FIG. 10 is a diagram showing an example in which unknown waveform data similar to an ideal waveform is input to a trained neural network. [Figure 23] FIG. 10 is a diagram showing an example in which unknown waveform data that is not similar to an ideal waveform is input to a trained neural network. [Figure 24] FIG. 1 is a diagram illustrating a learning method of an autoencoder. [Figure 25] FIG. 10 is a diagram illustrating an example in which unknown waveform data similar to an ideal waveform is input to a trained autoencoder. [Figure 26] FIG. 10 is a diagram illustrating an example in which unknown waveform data that is not similar to an ideal waveform is input to a trained autoencoder. [Figure 27] 10 is a flowchart showing a determination process according to a modified example. [Figure 28] 10 is a flowchart illustrating an example of performing a filtering process during execution of clustering. [Figure 29] FIG. 10 is a diagram illustrating an example of a filtering process. DETAILED DESCRIPTION OF THE INVENTION

[0024] Hereinafter, embodiments of the technology of the present disclosure will be described with reference to the drawings.

[0025] FIG. 1 schematically shows an electrical function evaluation system 2 that measures the electrical activity of cells. The electrical function evaluation system 2 shown in FIG. 1 is composed of a cell culture device 10 and an information processing device 20. The cell culture device 10 makes it possible to measure waveforms that indicate electrophysiological changes in cells (e.g., myocardial waveforms that indicate the pulsation of cardiomyocytes) while culturing the cells. The cell culture device 10 also controls the culture environment (e.g., temperature, carbon dioxide concentration).

[0026] An MEA plate 30 is used for culturing cells. The cell culture device 10 is provided with a culture chamber 11 that houses the MEA plate 30. The cell culture device 10 is also provided with a sliding lid 12 for opening and closing the culture chamber 11. The MEA plate 30 is mounted in the culture chamber 11 with cells seeded therein. The culture chamber 11 functions as an incubator, enabling long-term cell culture.

[0027] In this embodiment, cardiomyocytes produced from iPS cells are cultured as cells in the cell culture device 10. The cell culture device 10 also measures the extracellular potential representing the myocardial waveform of the cardiomyocytes seeded on the MEA plate 30 using a multipoint measurement method, and outputs the waveform data obtained by the measurement to the information processing device 20. The waveform data represents the pulse signal output by the cardiomyocytes.

[0028] The information processing device 20 is configured by a general computer such as a personal computer. Software for analyzing waveform data input from the cell culture device 10 is installed in the information processing device 20. The information processing device 20 has a display unit 21 and an input unit 22. The display unit 21 is a display device such as a liquid crystal display or an organic EL (Electro Luminescence) display. The input unit 22 is an input device such as a keyboard, a touchpad, or a mouse. The information processing device 20 is connected to the cell culture device 10 by wire or wirelessly. The display unit 21 and the input unit 22 may be configured as external devices connected to the information processing device 20.

[0029] The information processing device 20 calculates the field potential duration (FPD), interspike interval (ISI), etc. based on the input waveform data. The FPD corresponds to the QT interval (the time from the start of the Q wave to the end of the T wave) in an electrocardiogram, and is therefore used as an indicator of arrhythmia. A prolonged QT interval indicates the possibility of causing arrhythmia. Based on the FPD, etc., the user can perform toxicity assessment to evaluate the responsiveness of cells to drugs.

[0030] Fig. 2 shows an example of an MEA plate 30. The MEA plate 30 is a multi-well plate in which a plurality of culture wells (hereinafter simply referred to as wells) 32 are arranged on a substrate 31. The MEA plate 30 shown in Fig. 2 has 48 wells 32. Note that the number of wells 32 provided in the MEA plate 30 is not limited to 48, and may be 24, 96, or the like.

[0031] FIG. 3 shows an example of a well 32. The well 32 is to opening 33 is formed The well 32 is a roughly cylindrical container with a cylindrical shape. Cells are seeded so that they adhere to the bottom 34 of the well 32. The well 32 is filled with a culture solution containing a medium. A microelectrode array 40 (see FIG. 4) is embedded in the bottom 34 of the well 32.

[0032] FIG. 4 shows an example of a microelectrode array 40. The microelectrode array 40 has a plurality of electrodes 41. In the example shown in FIG. 4, the microelectrode array 40 has 16 microelectrodes (hereinafter simply referred to as electrodes) 41 arranged in a 4×4 square. The electrodes 41 are exposed at the bottom 34 of the well 32 and come into contact with the seeded cells. Each of the electrodes 41 is connected to a potential measurement circuit 50 (described later) via wiring 42. Hereinafter, the electrodes 41 may be referred to as channels CH. The 16 electrodes 41 are distinguished by being referred to as channels CH1 to CH16.

[0033] 5 shows an example of the hardware configuration of the electrical function evaluation system 2. The cell culture device 10 has a culture chamber 11, a potential measurement circuit 50, and a communication I / F (interface) 51. The potential measurement circuit 50 measures the extracellular potential of cardiomyocytes cultured on an MEA plate 30 housed in the cell culture device 10. Specifically, the potential measurement circuit 50 measures the extracellular potential via each electrode 41 of a microelectrode array 40 provided in each well 32. That is, the potential measurement circuit 50 measures 16 myocardial waveforms for each well 32.

[0034] The potential measurement circuit 50 transmits the measured myocardial waveform as waveform data to the information processing device 20 via the communication I / F 51. If the number of wells 32 formed in the MEA plate 30 is 48 and the number of microelectrode arrays 40 provided in each well 32 is 16, 768 pieces of waveform data are transmitted from the potential measurement circuit 50 to the information processing device 20.

[0035] The information processing device 20 includes a processor 23, a memory 24, an input unit 22, a display unit 21, a communication I / F 25, and a bus 26. The processor 23 is a computer that realizes various functions by reading out a program 28 and various data stored in the memory 24 and executing processing. The processor 23 is, for example, a CPU (Central Processing Unit).

[0036] The memory 24 is a storage device that stores the program 28 and various data used when the processor 23 executes processing. The memory 24 includes, for example, a random access memory (RAM), a read-only memory (ROM), or a storage. The RAM is, for example, a volatile memory used as a work area or the like for the processor 23. The ROM is, for example, a non-volatile memory that holds the program 28 and various data. The ROM is, for example, a flash memory. The storage is, for example, a large-capacity storage device such as a hard disk drive (HDD) or a solid state drive (SSD), and stores an operating system (OS), various data, etc. The memory 24 may be configured as an external device connected to the information processing device 20.

[0037] The memory 24 also stores teacher waveform data TD. The teacher waveform data TD is waveform data associated with a judgment result of whether the waveform is close to a known ideal waveform. This judgment of superiority or inferiority is made, for example, by a human sensory evaluation in which the waveform is compared with a known ideal waveform. The ideal waveform is a waveform that represents a healthy state (for example, a state free of diseases such as arrhythmia) obtained through previous measurements.

[0038] The waveforms output by the multiple electrodes 41 included in the microelectrode array 40 provided in the well 32 vary. For toxicity evaluation, it is necessary to use a waveform that is close to an ideal waveform. As will be described in detail later, the information processing device 20 performs a process of selecting, from the multiple electrodes 41 included in the microelectrode array 40, the electrode 41 that outputs a waveform that is most similar to the ideal waveform, as a target electrode for toxicity evaluation. This selection of the target electrode is also called selection of a golden channel (hereinafter referred to as GC). The information processing device 20 performs the GC selection process using multiple waveform data transmitted from the cell culture device 10 and multiple teacher waveform data TD pre-stored in the memory 24.

[0039] Fig. 6 shows an example of the functional configuration of the information processing device 20. The functions shown in Fig. 6 are realized by the processor 23 of the information processing device 20 executing processes based on the program 28. These various functions may also be realized by hardware.

[0040] The processor 23 functions as a data acquisition unit 60, a determination unit 61, and an output unit 62. The data acquisition unit 60 performs an acquisition process to acquire waveform data transmitted from the cell culture device 10. Since the similarity of the waveform data transmitted from the cell culture device 10 to the ideal waveform is unknown, hereinafter, the waveform data acquired by the data acquisition unit 60 from the cell culture device 10 is referred to as unknown waveform data UD. The result of the superiority / inferiority judgment of the unknown waveform data UD based on the similarity to the ideal waveform is unknown.

[0041] The determination unit 61 performs a determination process to determine the superiority or inferiority of each of the plurality of unknown waveform data UD acquired by the data acquisition unit 60 in terms of similarity to an ideal waveform, based on the plurality of teacher waveform data TD stored in advance in the memory 24. As will be described in detail later, in this embodiment, the determination unit 61 performs the determination of superiority or inferiority using a clustering technique, which is a type of machine learning algorithm.

[0042] The output unit 62 performs an output process to output the superiority or inferiority of the plurality of unknown waveform data UD determined by the determination unit 61 so as to be comparable. For example, the output unit 62 causes the display unit 21 to display the superiority or inferiority of the plurality of unknown waveform data UD so as to be comparable.

[0043] FIG. 7 shows an example of a myocardial waveform. The myocardial waveform shown in FIG. 7 is a measured waveform of an extracellular potential. In FIG. 7, FPD is the above-mentioned extracellular potential duration, and ISI is the beat interval. A1 is the amplitude of the first peak P1, and A2 is the amplitude of the second peak P2. These values ​​vary depending on the state of the cell or various factors at the time of measurement.

[0044] The quality of the teacher waveform data TD is determined by human evaluation using the FPD, ISI, maximum potential of P1 (hereinafter referred to as P1max), minimum potential of P1 (hereinafter referred to as P1min), A2 value, and the overall shape of the waveform as evaluation criteria. For example, a waveform that satisfies the following formulas (1) to (4) and meets the evaluation criteria that the overall shape is close to the ideal waveform is judged to be "excellent" (i.e., GC), and a waveform that does not meet the evaluation criteria is judged to be "poor." P1max≧200μV (1) P1min≦-200μV (2) A2≧15μV (3) FPDcF≧340msec (4) where FPDcF is the FPD to ISI 1 / 3 This is the value divided by .

[0045] The above determination result is added to the teacher waveform data TD (see FIG. 10).

[0046] Fig. 8 shows an example of unknown waveform data UD output from a plurality of electrodes 41 included in the microelectrode array 40. Fig. 8 shows unknown waveform data UD corresponding to each of channels CH1 to CH16. The determination unit 61 determines the relative merits of each piece of unknown waveform data UD through a determination process described below, and selects the electrode 41 that provides the waveform closest to the ideal waveform as the target electrode (i.e., GC).

[0047] Fig. 9 shows an example of the determination process by the determination unit 61. This example of the determination process will be described with reference to the flowchart shown in Fig. 9. First, the determination unit 61 creates a set including a plurality of unknown waveform data UD acquired by the data acquisition unit 60 and a plurality of teacher waveform data TD stored in the memory 24 (step S10).

[0048] FIG. 10 is an example of a set created in step S10. The unknown waveform data UD and teacher waveform data TD are time-series data for a period including one interbeat interval (ISI) (see FIG. 7). t1 to tn represent n times. For example, n=7250. T1 to Tm are data names for identifying m pieces of teacher waveform data TD. For example, m=1044. U1 to U16 are data names for identifying 16 pieces of unknown waveform data UD corresponding to channels CH1 to CH16.

[0049] The teacher waveform data TD is linked to the result of a superiority / inferiority judgment based on human evaluation. "1" indicates that the waveform is judged to be similar to the ideal waveform (i.e., GC). "0" indicates that the waveform is judged to be dissimilar to the ideal waveform (i.e., not GC). In other words, "1" indicates that the result of the superiority / inferiority judgment based on the similarity to the ideal waveform is a "superior judgment," and "0" indicates that the result of the superiority / inferiority judgment based on the similarity to the ideal waveform is a "poor judgment."

[0050] Next, the determination unit 61 combines the multiple pieces of unknown waveform data UD and multiple pieces of teacher waveform data TD included in the created set, and performs clustering using the k-medoids method (step S11). As an example, as shown in Fig. 11, the determination unit 61 performs clustering with the number of clusters k set to 3. As a result, the unknown waveform data UD and the teacher waveform data TD are each assigned to one of three clusters CL1 to CL3.

[0051] Figure 12 explains the clustering technique. As shown in Figure 12, clustering is performed by treating each of the multiple unknown waveform data UD and multiple teacher waveform data TD as points plotted in n-dimensional space. The n axes representing the n-dimensional space are the above-mentioned t1 to tn. For the sake of simplicity, Figure 12 is a diagram in which the unknown waveform data UD and teacher waveform data TD are plotted in a three-dimensional space with t1 to t3 as the axes.

[0052] The well-known k-medoids method is used for clustering. In the k-medoids method, first, k points are randomly selected as medoids in n-dimensional space. Next, each point is assigned to the cluster of the closest medoid. Then, within each cluster, a new medoid is set so that the sum of the distances to all other points within the cluster is minimized. After this, the process is repeated until there are no more changes in the medoids. In this embodiment, k=3, so three clusters CL1 to CL3 are generated. Note that the k-means method can also be used instead of the k-medoids method. In the k-means method, the center of gravity of the points within a cluster is calculated, and the process is repeated until there are no more changes in the center of gravity.

[0053] Next, the determination unit 61 identifies clusters that include the unknown waveform data UD from the multiple clusters generated by clustering (step S12). In the example shown in Fig. 11, all of the clusters CL1 to CL3 include the unknown waveform data UD, so all of the clusters CL1 to CL3 are identified as clusters that include the unknown waveform data UD.

[0054] Next, the determination unit 61 determines whether each of the unknown waveform data UD included in the cluster identified in step S12 is superior or inferior (step S13). Specifically, the determination unit 61 determines whether each of the unknown waveform data UD in each of the identified clusters is superior or inferior by calculating the probability that the unknown waveform data UD is superiorly determined to be "1" (hereinafter referred to as the score SC). More specifically, the determination unit 61 determines the score SC for each of the identified clusters based on the number N1 of teacher waveform data with a superior determination of "1" and the number NT of teacher waveform data with a superior determination of "1" and a poor determination of "0". NT corresponds to the number of teacher waveform data TD included in the cluster.

[0055] 13, the score SC is calculated by dividing the number N1 of teacher waveform data with a superior judgment of "1" by the number NT of teacher waveform data with a superior judgment of "1" and a poor judgment of "0" (i.e., N1 / NT). In the example shown in FIG. 13, the score SC of cluster CL1 is 0.125, the score SC of cluster CL2 is 0.021, and the score SC of cluster CL3 is 0.250. Therefore, the unknown waveform data U6, U8, and U9 included in cluster CL3 have the highest scores SC.

[0056] Next, the determining unit 61 ranks the unknown waveform data UD based on the results of the superiority / inferiority determination (step S14). unknown Waveform Data UD Based on the score SC, which represents the result of the evaluation, the items are sorted in descending order of SC score. In order 14, unknown waveform data U6, U8, and U9 included in cluster CL3 are ranked first.

[0057] Next, the determination unit 61 determines whether there are multiple pieces of first-ranked unknown waveform data UD (step S15). If the determination unit 61 determines that there are multiple pieces of first-ranked unknown waveform data UD (step S15: YES), the determination unit 61 re-clusters the cluster including the first-ranked unknown waveform data UD (step S16). The clustering method is the same as in step S11. In the example shown in FIG. 14, the three pieces of teacher waveform data TD, unknown waveform data U6, U8, and U9, are first, so the cluster CL3 (see FIG. 11) including them is re-clustered.

[0058] After step S16, the determination unit 61 returns the process to step S12. Thereafter, the determination unit 61 executes the processes of steps S12 to S15 on the multiple sub-clusters obtained by clustering cluster CL3. If the determination unit 61 determines that there is not multiple first-ranked unknown waveform data UD (i.e., there is only one first-ranked unknown waveform data UD) (step S15: NO), the determination process ends.

[0059] 15 shows an example in which unknown waveform data U6, U8, and U9 are ranked as a result of performing the processes of steps S12 to S15 after clustering cluster CL3. In the example shown in FIG. 15, only unknown waveform data U9 is ranked first. That is, electrode 41 (i.e., channel CH9) from which unknown waveform data U9 was obtained is selected as the target electrode (GC) for toxicity evaluation.

[0060] The output unit 62 makes it possible to compare the superiority or inferiority of multiple pieces of unknown waveform data UD by, for example, displaying a table shown in Fig. 15 on the display unit 21. The output unit 62 may also display a GC selected from channels CH1 to CH16 in an identifiable manner, as shown in Fig. 16. The display example shown in Fig. 16 is an example of "a mode of outputting the superiority or inferiority of multiple pieces of unknown waveform data in a manner that allows comparison" according to the technology of the present disclosure.

[0061] As described above, the technology disclosed herein is characterized by determining the superiority of each of a plurality of unknown waveform data sets based on a plurality of teacher waveform data sets to which the superiority judgment results are linked, and outputting the superiority of the plurality of unknown waveform data sets so that they can be compared, thereby enabling a waveform that is close to the ideal waveform to be selected with high accuracy in a short time.

[0062] Furthermore, it is possible to evaluate drugs based on the superiority / inferiority determination information output from the information processing device 20. Of the multiple unknown waveform data, the unknown waveform data with the highest ranking of superiority determined by the information processing device 20 may be used for drug evaluation.

[0063] [Example] An example of the process for determining superiority will be described below.

[0064] In this example, a MAESTRO768PRO manufactured by Axion BioSystems was used as the cell culture device 10, and a multiwell plate with 24 wells 32 was used as the MEA plate 30. Cardiomyocytes were cultured in each well 32 of the cell culture device 10 while measuring myocardial waveforms. Waveform data was generated by extracting data from the waveform data measured between 150 and 135 seconds from the start of measurement, with the highest voltage point designated as the zero point on the time axis, and extracting 250 points in the negative direction from the zero point and 7000 points in the positive direction. This waveform data represents a myocardial waveform including one beat interval.

[0065] Based on the waveform data acquired by the cell culture device 10, teacher waveform data TD for three plates was prepared, to which the results of superiority judgments made by humans in the past were linked. That is, the number of teacher waveform data TD prepared was 1,152. In addition, based on the waveform data acquired by the cell culture device 10, unknown waveform data UD for one well was prepared, for which new superiority judgments were made. That is, the number of unknown waveform data UD prepared was 16.

[0066] FIG. 17 shows a plurality of teacher waveform data TD and a plurality of unknown Waveform Data UD The data name of the teacher waveform data TD indicates the plate number and the serial number of the waveform data within the plate. For example, T2-383 indicates that the plate number is 2 and the serial number is 383.

[0067] Next, the multiple unknown waveform data UD and multiple teacher waveform data TD included in the set were combined, and clustering (first clustering) was performed using the k-medoids method using MATLAB (registered trademark), a numerical analysis software made by MathWorks (registered trademark). Here, k = 10. Figure 18 shows the results of clustering the multiple unknown waveform data UD and multiple teacher waveform data TD included in the set shown in Figure 17. The 10 clusters generated by clustering are numbered 1 to 10 with no overlaps.

[0068] Next, a score SC was calculated for each cluster containing unknown waveform data UD, and the unknown waveform data UD were ranked in descending order of score SC based on the calculated score SC. Figure 18 shows the results of ranking the unknown waveform data UD. The first place cluster contained three pieces of unknown waveform data U3, U6, and U11. Therefore, cluster CL6, which contained the first place unknown waveform data U3, U6, and U11, was subjected to clustering again (second clustering). Here, since the number of first place unknown waveform data UDs was three, k was set to 3.

[0069] Figure 19 shows the results of clustering cluster CL6 again. As shown in Figure 19, a score SC was calculated for each of three sub-clusters CL6-1, CL6-2, and CL6-3 generated by clustering cluster CL6 again. Then, as shown in Figure 20, the unknown waveform data U3, U6, and U11 included in cluster CL6 were ranked based on the score SC. As a result, the score SC of unknown waveform data U11 was the highest, and only unknown waveform data U11 was ranked first, so channel CH16 was selected as GC.

[0070] As described above, it has been confirmed that by using a clustering technique, it is possible to accurately select a waveform that is close to an ideal waveform from multiple unknown waveform data in a short period of time.

[0071] [Variations] Next, various modifications of the above embodiment will be described.

[0072] In the above embodiment, the teacher waveform data TD is associated with binary data of "1" or "0" as the result of the superiority / inferiority judgment, but it may also be associated with data of three or more values. In other words, the result of the superiority / inferiority judgment is not limited to being expressed by two values, but may also be expressed by three or more values.

[0073] In the above embodiment, the judgment unit 61 judges the superiority or inferiority of the unknown waveform data UD by clustering, but the method is not limited to clustering, and the superiority or inferiority of the unknown waveform data UD may also be judged by using a neural network that has undergone machine learning.

[0074] 21 schematically illustrates a learning method for the neural network 70 used by the judgment unit 61. In the learning phase, the neural network 70 is trained using the time series data t1 to tn of the teacher waveform data TD as explanatory variables and the result of the judgment of superiority or inferiority linked to the teacher waveform data TD as the objective variable. The result of the judgment of superiority or inferiority is the "label" in so-called labeled machine learning.

[0075] In the learning phase, when time series data t1 to tn are input to the input layer of the neural network 70, the output value from the output layer is input to the adjustment unit 71. The result of the superiority / inferiority judgment as a label is also input to the adjustment unit 71. The adjustment unit 71 compares the output value with the inferiority judgment result, and adjusts the weights and biases of the neural network 70 based on the difference between the output value and the inferiority judgment result.

[0076] As shown in Figures 22 and 23, the judgment unit 61 judges the superiority or inferiority of the unknown waveform data UD using a trained neural network 70A that has undergone machine learning using a plurality of teacher waveform data TD. As shown in Figure 22, when unknown waveform data UD similar to an ideal waveform is input to the neural network 70A, an output value of "1" is output. An output value of "1" indicates that the corresponding electrode 41 is a GC. On the other hand, as shown in Figure 23, when unknown waveform data UD not similar to the ideal waveform is input to the neural network 70A, an output value of "0" is output. An output value of "0" indicates that the corresponding electrode 41 is not a GC.

[0077] In this modification, the teacher waveform data TD used for learning by the neural network 70 is linked to binary data of "1" or "0" as the result of the superiority / inferiority judgment, but data of three or more values ​​may be linked. In other words, the result of the superiority / inferiority judgment is not limited to being expressed by two values, but may be expressed by three or more values. In this case, the output value from the neural network 70 is expressed by three or more values.

[0078] The determining unit 61 may also determine the superiority or inferiority of the unknown waveform data UD using an autoencoder that has undergone machine learning.

[0079] FIG. 24 illustrates a learning method for the autoencoder 80 used by the determination unit 61. The autoencoder 80 has an encoder and a decoder. The autoencoder 80 extracts features by reducing the dimensions of the input data using the encoder, and the decoder reconstructs and outputs the input data based on the extracted features. In the learning phase of the autoencoder 80, edge weights are adjusted so that the input and output match. Through this learning, important information required for reconstruction is extracted from the data, and a network for efficiently reconstructing the original data is formed.

[0080] As shown in Figure 24, in the learning phase, teacher Waveform Data TD The autoencoder 80 is trained using only the trained autoencoder 80A. As a result, even if a waveform dissimilar to the ideal waveform is input, the trained autoencoder 80A (see Figures 25 and 26) outputs a waveform similar to the ideal waveform. Therefore, by comparing the input waveform and output waveform of the trained autoencoder 80A, the similarity of the input waveform to the ideal waveform can be determined. After inputting unknown waveform data UD to the encoder of the trained autoencoder 80A, the determination unit 61 determines the superiority or inferiority of the unknown waveform data UD based on the difference between the waveform data restored by the decoder and the unknown waveform data UD input to the encoder.

[0081] 25 and 26, the judgment unit 61 judges the superiority of the unknown waveform data UD using a trained autoencoder 80A that has undergone machine learning using multiple pieces of teacher waveform data TD linked to superiority judgment results, and a comparison unit 81 that compares input and output data. The comparison unit 81 calculates the difference between the input and output data, and outputs "1" if the difference is less than a certain value, and "0" if the difference is equal to or greater than the certain value.

[0082] As shown in FIG. 25, when unknown waveform data UD similar to the ideal waveform is input to the autoencoder 80A, the autoencoder 80A outputs waveform data similar to the ideal waveform. In this case, the comparison unit 81 outputs "1." An output value of "1" from the comparison unit 81 indicates that the corresponding electrode 41 is a GC. On the other hand, as shown in FIG. 26, even when unknown waveform data UD not similar to the ideal waveform is input to the autoencoder 80A, the autoencoder 80A outputs waveform data similar to the ideal waveform. In this case, the comparison unit 81 outputs "0." An output value of "0" from the comparison unit 81 indicates that the corresponding electrode 41 is not a GC.

[0083] In this modification, the comparison section 81 outputs binary data of "1" or "0" according to the difference value, but is not limited to this and may be configured to output data of three or more values.

[0084] Next, a modified example of clustering will be described. The ranking of unknown waveform data UD by clustering described in the above embodiment is performed based on the similarity of the shape with the ideal waveform. In order to improve accuracy in determining minute differences in peak values ​​in the waveform, it is preferable to perform a filtering process using the evaluation criteria expressed by the above equations (1) to (4) to exclude unknown waveform data UD that do not satisfy the evaluation criteria from the multiple unknown waveform data UD to be determined before clustering is performed.

[0085] Fig. 27 shows a determination process according to a modified example. As shown in Fig. 27, in this modified example, the determination unit 61 executes step S10, which creates a set (see Fig. 10) including a plurality of unknown waveform data UD and a plurality of teacher waveform data TD, and then performs a process of excluding unknown waveform data UD that does not satisfy the evaluation criteria from the set (step S20). Specifically, the values ​​of P1max, P1min, A2, and FPDcF are measured for each of the plurality of unknown waveform data UD, and unknown waveform data UD that does not satisfy at least one of equations (1) to (4) is excluded from the set and is therefore not subject to clustering. Steps S11 to S16 are the same as those described in the above embodiment.

[0086] In this way, by performing a filtering process to exclude unknown waveform data UD that do not satisfy the evaluation criteria before clustering, the unknown waveform data UD that do not satisfy the evaluation criteria are prevented from being ranked higher by clustering, thereby improving the accuracy of the determination process by the determination unit 61.

[0087] In the modification shown in Fig. 27, filtering is performed before clustering, but filtering may also be performed during clustering. For example, as shown in Fig. 28, the determination unit 61 performs step S14 of ranking multiple unknown waveform data UD based on the score SC, and then performs a process of lowering the ranking of unknown waveform data UD that does not satisfy the evaluation criteria (step S30). The method of identifying unknown waveform data UD that does not satisfy the evaluation criteria is the same as above.

[0088] For example, as shown in Fig. 29, the determination unit 61 performs filtering processing on multiple unknown waveform data UD. The determination unit 61 determines unknown waveform data UD that does not satisfy the evaluation criteria as a poor determination, and unknown waveform data UD that satisfies the evaluation criteria as a good determination, and lowers the ranking of the unknown waveform data UD that is determined to be poor (for example, to be last). In Fig. 29, values ​​that do not satisfy the evaluation criteria are marked with an asterisk.

[0089] In the example shown in Figure 29, the clustering results show that unknown waveform data U6, U8, and U9 are all tied for first place, but the P1min value of unknown waveform data U6 does not meet the evaluation criteria, and A2 of unknown waveform data U8 does not meet the evaluation criteria. Therefore, in this example, unknown waveform data U6 and U8 are judged to be inferior and are lowered in rank. As a result, only unknown waveform data U9 is ranked first, and electrode 41 (i.e., channel CH9) from which unknown waveform data U9 was obtained is selected as the target electrode (GC) for toxicity evaluation.

[0090] The filtering process in step S30 does not necessarily have to be performed immediately after step S14, but may be performed after it is determined in step S15 that there are multiple pieces of first-ranked unknown waveform data UD. In this case, the filtering process may be performed only on the multiple pieces of first-ranked unknown waveform data UD.

[0091] The evaluation criteria used in the filtering process are not limited to the above-mentioned formulas (1) to (4) and can be changed as appropriate.

[0092] In the above embodiment, cardiomyocytes are used as the cells, but it is also possible to use cells such as nerve cells instead of cardiomyocytes.

[0093] In the above embodiment, the hardware structure of the processing units that execute various processes, such as the data acquisition unit 60, the determination unit 61, and the output unit 62, is the following various processors.

[0094] Various types of processors include CPUs, programmable logic devices (PLDs), dedicated electrical circuits, etc. As is well known, a CPU is a general-purpose processor that executes software (programs) and functions as various processing units. A PLD is a processor whose circuit configuration can be changed after manufacturing, such as an FPGA (Field Programmable Gate Array). A dedicated electrical circuit is a processor with a circuit configuration designed specifically to execute specific processes, such as an ASIC (Application Specific Integrated Circuit).

[0095] A single processing unit may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, multiple FPGAs, or a combination of a CPU and an FPGA). Also, multiple processing units may be configured with a single processor. As an example of configuring multiple processing units with one processor, first, there is a form in which one processor is configured by combining one or more CPUs and software, and this processor functions as multiple processing units. Second, there is a form in which a processor is used that realizes the functions of an entire system including multiple processing units with one IC chip, as typified by a system on chip (SoC). In this way, various processing units are configured using one or more of the above-mentioned various processors as a hardware structure.

[0096] Furthermore, the hardware structure of these various processors is, more specifically, an electric circuit made up of a combination of circuit elements such as semiconductor elements.

[0097] The present invention is not limited to the above-described embodiment, and various configurations can be adopted without departing from the spirit of the present invention. Furthermore, the present invention extends to a computer-readable storage medium that non-temporarily stores a program in addition to the program.

Claims

1. A computer system comprising a memory and a processor, The memory stores a plurality of teacher waveform data pieces, each of which is associated with a superiority / inferiority judgment result based on the degree of similarity to an ideal waveform; The processor: The result of the superiority / inferiority judgment is unknown, and a plurality of unknown waveform data output from a plurality of electrodes included in a microelectrode array is acquired from the cell culture device; determining the superiority or inferiority of each of the plurality of unknown waveform data based on the plurality of teacher waveform data; selecting an electrode from the plurality of electrodes that outputs a waveform most similar to the ideal waveform as a target electrode; outputting the plurality of unknown waveform data so that the superiority or inferiority of the data can be compared; Information processing device.

2. The processor: clustering a set including the plurality of teacher waveform data and the plurality of unknown waveform data; determining the superiority or inferiority of each of the clusters containing at least one of the plurality of unknown waveform data as a result of the clustering by calculating the probability that the unknown waveform data is superior; The information processing device according to claim 1 .

3. The processor: calculating the probability for each of the clusters based on the number of teacher waveform data items with a good decision and the number of teacher waveform data items with a good decision and a bad decision; The information processing device according to claim 2 .

4. the probability is expressed as a value obtained by dividing the number of the teacher waveform data with a good judgment by the number of the teacher waveform data with a good judgment and a bad judgment, the processor outputs a comparative ranking of the plurality of unknown waveform data in terms of merits based on the probabilities. The information processing device according to claim 3 .

5. The processor performs the clustering using a k-medoids method or a k-means method. The information processing device according to any one of claims 2 to 4.

6. the processor performs a filtering process to remove unknown waveform data that does not satisfy an evaluation criterion from the set or to lower the ranking of the unknown waveform data in terms of relative merit. The information processing device according to claim 5 .

7. The processor: The unknown waveform data is input to a neural network that has undergone machine learning based on the teacher waveform data, and the superiority or inferiority is determined based on the results output from the neural network. The information processing device according to claim 1 .

8. The processor: The unknown waveform data is input to an encoder of an autoencoder that has undergone machine learning based on the teacher waveform data of the superiority judgment, and then the superiority or inferiority is judged based on the difference between the waveform data restored by a decoder and the unknown waveform data input to the encoder. The information processing device according to claim 1 .

9. The unknown waveform data is a pulse signal output by a cell. The information processing device according to any one of claims 1 to 8.

10. the cells are cardiomyocytes; The information processing device according to claim 9 .

11. A drug evaluation method for evaluating a drug based on information output from the information processing device according to claim 9 or 10, comprising: The unknown waveform data having a high ranking of superiority or inferiority is used for drug evaluation. Drug evaluation methods.

12. A plurality of unknown waveform data output from a plurality of electrodes included in a microelectrode array, the results of which are determined based on the degree of similarity to an ideal waveform, are unknown, and the unknown waveform data is acquired from a cell culture device; determining the superiority or inferiority of each of the plurality of unknown waveform data based on the plurality of teacher waveform data linked to the superiority or inferiority determination results; selecting an electrode from the plurality of electrodes that outputs a waveform most similar to the ideal waveform as a target electrode; outputting the plurality of unknown waveform data so that the superiority or inferiority of the data can be compared; Information processing methods.

13. an acquisition process for acquiring, from the cell culture device, a plurality of unknown waveform data output from a plurality of electrodes included in the microelectrode array, the results of which are determined based on the degree of similarity to the ideal waveform and are unknown; a determination process for determining the superiority or inferiority of each of the plurality of unknown waveform data based on the plurality of teacher waveform data linked to the superiority or inferiority determination results; a selection process for selecting, as a target electrode, an electrode that outputs a waveform that is most similar to the ideal waveform from the plurality of electrodes; an output process for outputting the plurality of unknown waveform data in a manner that allows comparison of their merits; A program that causes a computer to execute the following.

Images