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

Abstract

The present invention provides an information processing device that can more easily determine classification than conventional techniques based on the current signal from a nanopore when a substance passes through the nanopore. [Solution] An information processing device is provided that includes a conversion unit that converts the current signal generated when a substance passes through a nanopore into image data using a predetermined conversion process, and a discrimination unit that inputs the image data into a trained model generated by machine learning using the image data and discriminates the substance using the output from the trained model.

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] Nanopore sensing has attracted attention as a label-free and rapid single-molecule detection technology that uses nanoscale pores called nanopores. Patent Document 1 discloses a classification analysis method that performs classification analysis of detection signal data, which detects changes in current flow between electrodes caused by the passage of an analyte through a through-pore, by executing a computer-controlled program. Patent Document 2 discloses a technology that segments the current signal from each nanopore, reshapes and rescales the segmented events into square images, combines them to obtain stacked images, and provides them to a neural network to determine the classification of the sample. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2020-173259 [Patent Document 2] International Publication No. 2023 / 215406 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the conventional technology has room for improvement in determining a substance based on a current signal from a nanopore when the substance passes through the nanopore. Specifically, the conventional technology has room for improvement in the time required for determining a substance and the accuracy of the determination.

[0005] The present disclosure has been made in consideration of the above points, and aims to provide an information processing device, an information processing method, and an information processing program that can more easily determine classification based on the current signal from a nanopore when a substance passes through the nanopore, compared to conventional techniques. [Means for solving the problem]

[0006] According to one aspect of the present disclosure, there is provided an information processing device comprising: a conversion unit that converts a current signal generated when a substance passes through a nanopore into image data by a predetermined conversion process; and a discrimination unit that inputs the image data into a trained model generated by machine learning using the image data and discriminates the substance using the output from the trained model.

[0007] The conversion unit may convert the same current signal into one piece of image data by connecting a plurality of sub-image data pieces obtained by converting the same current signal based on a plurality of the conversion processes.

[0008] The conversion unit may convert a predetermined range of the current signal into the sub-image data as one of the plurality of conversion processes.

[0009] The conversion unit may convert the current signal into the sub-image data by extending a predetermined range of a tip portion of the current signal as one of the plurality of conversion processes.

[0010] The conversion unit may convert the current signal into the sub-image data by extending a predetermined range of a terminal end portion of the current signal as one of the plurality of conversion processes.

[0011] The conversion unit may convert the current signal into the sub-image data representing a change in frequency components over time based on a result of a Fourier transform of the current signal, as one of the plurality of conversion processes.

[0012] The conversion unit may convert the current signal into the sub-image data representing a plurality of pieces of statistical information as one of the plurality of conversion processes.

[0013] The substance may be a substance that can pass through the nanopore.

[0014] The substance may be a polynucleotide, a peptide or a protein.

[0015] According to another aspect of the present disclosure, an information processing method is provided in which a processor converts a current signal generated when a substance passes through a nanopore into image data using a predetermined conversion process, inputs the image data into a trained model generated by machine learning using the image data, and performs a process of identifying the substance using output from the trained model.

[0016] According to another aspect of the present disclosure, an information processing program is provided that causes a computer to convert a current signal generated when a substance passes through a nanopore into image data using a predetermined conversion process, input the image data into a trained model generated by machine learning using the image data, and execute a process of identifying the substance using output from the trained model. [Effects of the Invention]

[0017] According to the present disclosure, it is possible to provide an information processing device, an information processing method, and an information processing program that can more easily determine classification based on a current signal from a nanopore compared to conventional techniques. [Brief explanation of the drawings]

[0018] [Figure 1] FIG. 1 is a diagram illustrating an overview of an information processing device according to an embodiment of the disclosed technology. [Figure 2] FIG. 2 is a block diagram showing a hardware configuration of the information processing device. [Figure 3] FIG. 2 is a block diagram illustrating an example of a functional configuration of an information processing device. [Figure 4] 10 is a graph showing an example of the time course of the conductance of a current signal, excluding components of 1 Hz or higher, and the density of the presence of the current signal. [Figure 5]1 is a graph showing the current signal of a portion recognized as a single pore. [Figure 6] 1 is a graph showing the current signal of a portion recognized as a single pore. [Figure 7] This is a graph showing conductance data of a current signal and data obtained by removing fine components of 10,000 Hz or more from the current signal using a low-pass filter. [Figure 8] This is a graph showing data in which fine components of 10,000 Hz or higher of the current signal have been removed using a low-pass filter, one extracted signal, and the data density within the signal. [Figure 9] FIG. 2 is a diagram showing an example of sub-image data 1. [Figure 10] FIG. 10 is a diagram showing an example of sub-image data 2. [Figure 11] FIG. 10 is a diagram showing an example of sub-image data 3. [Figure 12] FIG. 10 is a diagram showing an example of sub-image data 4. [Figure 13] FIG. 10 is a diagram illustrating an example of image data. [Figure 14] 10 is a flowchart showing a flow of information processing by an information processing device. [Figure 15] 10 is a graph showing the accuracy of the results of identifying the type of miRNA using an information processing device based on the current signal generated when the miRNA passes through a nanopore. DETAILED DESCRIPTION OF THE INVENTION

[0019] An example of an embodiment of the present disclosure will be described below with reference to the drawings. The same reference numerals are used throughout the drawings to designate identical or equivalent components and parts. The dimensional proportions of the drawings are exaggerated for illustrative purposes and may differ from the actual proportions.

[0020] FIG. 1 is a diagram illustrating an overview of an information processing device according to this embodiment. The information processing device 10 according to this embodiment is a device that recognizes and identifies the type of a micro-object from a current signal obtained by a measuring device 20. The measuring device 20 is a device that obtains a current signal by nanopore sensing. Nanopore sensing is a sensing method that uses a current signal generated when a micro-substance passes through a fine through-hole (micropore, hereinafter collectively referred to as nanopore) on the micro- or nanometer scale, and is a molecular analysis method with high sensitivity and high time resolution. The substance to be identified is a substance that can pass through a nanopore, such as a polynucleotide, peptide, or protein. In this embodiment, the type of the substance to be identified is recognized and identified as miRNA (microRNA).

[0021] The measurement device 20 includes a measurement device fabricated using, for example, MEMS (Micro Electro Mechanical Systems). The measurement device 20 forms a lipid bilayer membrane on the measurement device using the droplet contact method, and then reconstitutes an α-hemolysin nanopore on the lipid bilayer membrane. The measurement device 20 places 4.7 μL of a miRNA solution adjusted to a final concentration of 10 μM, 1 M KCl, and 10 mM MOPS on the negative electrode of the measurement device, and 4.7 μL of a solution adjusted to a final concentration of 1 M KCl and 10 mM MOPS on the positive electrode of the measurement device. Measurements are performed at a sampling rate of 250 kHz while applying a voltage of 120 mV using an ammeter. A sampling rate of 250 kHz means that the current value is measured 250,000 times per second. Because miRNA is short and single-stranded, and does not undergo unzipping (where bases are peeled off one by one at the entrance of the pore) as occurs with double-stranded DNA / RNA, sufficient information can be obtained even when the miRNA passes through the nanopore rapidly.

[0022] Furthermore, by binding to the complementary strand of miRNA to form a double-stranded molecule, unzipping occurs at the entrance of the pore, extending the passage time and enabling detection even at lower sampling rates.

[0023] The information processing device 10 according to this embodiment acquires a current signal obtained when a solution of miRNA is passed through a nanopore in the measurement device 20, converts the acquired current signal into image data based on a predetermined conversion process, and uses the converted image data to recognize and identify the type of miRNA. When executing the process of recognizing and identifying the type of miRNA, the information processing device 10 uses a trained model 30, which is a neural network trained in advance using similarly converted image data. That is, the information processing device 10 inputs the image data converted from the current signal into the trained model 30 and obtains an output from the trained model 30 to identify the type of miRNA.

[0024] The information processing device 10 of this embodiment inputs image data converted from a current signal into a trained model 30 and obtains output from the trained model 30, thereby enabling the type of miRNA to be identified more easily and accurately than conventional discrimination methods.

[0025] FIG. 2 is a block diagram showing the hardware configuration of the information processing device 10. As shown in FIG.

[0026] 2, the information processing device 10 includes a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a storage 14, an input unit 15, a display unit 16, and a communication interface (I / F) 17. Each component is connected to each other via a bus 19 so as to be able to communicate with each other.

[0027] The CPU 11 is a central processing unit that executes various programs and controls each component. That is, the CPU 11 reads programs from the ROM 12 or storage 14 and executes the programs using the RAM 13 as a work area. The CPU 11 controls the above components and performs various arithmetic processing in accordance with the programs recorded in the ROM 12 or storage 14. In this embodiment, the ROM 12 or storage 14 stores an information processing program that recognizes and identifies the type of micro-object from the current signal obtained in the measuring device 20.

[0028] The ROM 12 stores various programs and various data. The RAM 13 temporarily stores programs or data as a working area. The storage 14 is configured with a storage device such as an HDD (Hard Disk Drive), SSD (Solid State Drive), or flash memory, and stores various programs including the operating system and various data.

[0029] The input unit 15 includes a pointing device such as a mouse and a keyboard, and is used to perform various inputs.

[0030] The display unit 16 is, for example, a liquid crystal display, and displays various information. The display unit 16 may also function as the input unit 15 by adopting a touch panel system.

[0031] The communication interface 17 is an interface for communicating with other devices such as the measurement device 20, and uses standards such as Ethernet (registered trademark), FDDI, and Wi-Fi (registered trademark).

[0032] When executing the above-mentioned location identification program, the information processing device 10 uses the above-mentioned hardware resources to realize various functions. The functional configuration realized by the information processing device 10 will be described.

[0033] FIG. 3 is a block diagram showing an example of the functional configuration of the information processing device 10. As shown in FIG.

[0034] 3, the information processing device 10 has, as functional components, a conversion unit 101 and a prediction unit 102. Each functional component is realized by the CPU 11 reading and executing an information processing program stored in the ROM 12 or the storage 14.

[0035] The conversion unit 101 converts the current signal measured by the measurement device 20, i.e., the current signal generated when a substance passes through a nanopore, into image data based on a predetermined conversion process. First, the measurement of the current signal by the measurement device 20 will be described.

[0036] To extract the current signal generated when a substance passes through the nanopore, the measuring device 20 extracts only the time period when only one pore is open from the measurement data. A low-pass filter is a function that removes high-frequency components from a waveform. The measuring device 20 uses a low-pass filter to remove fine components of a specific frequency, e.g., 1 Hz or higher, from the conductance of the current signal. In this embodiment, specifically, the measuring device 20 reduces noise generated when measuring the current using a low-pass filter that removes components above 100 kHz. The current data further reduces the frequency to 100 kHz, and then extracts the portion recognized as a single pore using a low-pass filter that removes components above 1 Hz. The portion recognized as a single pore refers to the portion recognized as a single nanopore when reconstituting an α-hemolysin nanopore on a lipid bilayer membrane, based on the conductance value. The measuring device 20 uses a low-pass filter to extract the portion recognized as a single pore of the conductance within a range of ±0.15 nS from the peak of the density throughout the entire data time.

[0037] FIG. 4 is a graph showing an example of the time course and presence density of the conductance of a current signal, excluding elements above 1 Hz. In the graph of FIG. 4, the portion where a long signal was present, such as around 150 seconds, jumps out of the range of the thresholds (max threshold and min threshold), so signals with lengths of less than 1 second are not considered. In the region recognized as a long single pore, such as around 130 to 170 seconds, there is a portion below the threshold (min threshold). However, we want to ignore the portion below the threshold and consider the entire region as a single pore, so if the threshold (min threshold) is exceeded within 1 second, it is included even if it is outside the threshold range. Also, in the graph of FIG. 4, the steep waveforms that appear when the current is reversed and the pore is unclogged, which appear around 215 to 230 seconds, pass within the threshold, albeit for a short time. To exclude these steep waveforms, which happen to fall slightly within the threshold despite not being recognized as a single pore, the measurement device 20 uses a low-pass filter to exclude waveforms with lengths of less than 1 second. By applying a 1-second low-pass filter and smoothing the data with a wide range, as shown in Figure 5, as shown in Figure 4, it becomes easier to determine whether the noise-free conductance value is within the threshold. The measurement device 20 calculates the median conductance value for the remaining range and excludes all outliers based on the median absolute deviation (MAD). The measurement device 20 then combines all of the measurement data for the same time domain as the remaining range before applying the low-pass filter, and determines the current signal for the area where only one α-hemolysin nanopore is recognized to have been reconstituted on the lipid bilayer membrane (the area recognized as a single pore).

[0038] Figure 5 is a graph showing the current signal from the portion recognized as a single pore, and the ranges indicated by symbols 21, 22, and 23 indicate the range remaining after excluding the group showing outliers by MAD. Figure 6 is a graph showing the current signal from the portion recognized as a single pore, and is a graph obtained by connecting all of the measurement data from the ranges indicated by symbols 21, 22, and 23 in the graph of Figure 5.

[0039] The measurement device 20 extracts the signal generated when a substance passes through the nanopore with conductance inhibited by a predetermined percentage, e.g., 40% compared to the open level. The open level is the most frequent conductance value (the value indicated by the apex of the central peak of the kernel density in Figure 4) when the pore is completely open (i.e., the portion of the conductance data that does not generate a signal when a substance passes through the nanopore) within the range of conductance data that can be specified as reference numerals 21, 22, and 23 in Figure 5, such as a conductance of around 1.4 in Figure 4. The 40% value is contained within the current noise, so if extracted as is, a large amount of noise would be judged as the signal. The reason for using a value 40% less than the open level is that, since the value would be buried in the noise range of data with a noise band above and below it, as indicated by reference numeral 24 in Figure 7 (described below), it is necessary to determine whether the signal falls below a value 40% less than the open level after reducing the noise, as indicated by reference numeral 25. To extract the signal, the measuring device 20 uses a low-pass filter to remove fine components above 10,000 Hz in order to distinguish the signal from noise during measurement. The measuring device 20 then extracts a range deeper than a predetermined percentage, for example, 40%, and obtains the unfiltered data within that range as a current signal. For all signals, the measuring device 20 converts the conductance data into depth (ion current inhibition rate) data by processing 1-(signal conductance / open level).

[0040] Figure 7 is a graph showing conductance data of a current signal and data obtained by removing fine components above 10,000 Hz from the current signal using a low-pass filter. Reference numeral 24 denotes the conductance data, and reference numeral 25 denotes the data obtained by removing fine components above 10,000 Hz from the current signal using a low-pass filter. Figure 8 is a graph showing data obtained by removing fine components above 10,000 Hz from the current signal using a low-pass filter, one extracted signal, and the data density within the signal. Reference numeral 26 denotes one extracted signal (the current signal generated when a substance passes through the nanopore), reference numeral 27 denotes data obtained by removing fine components above 10,000 Hz from the current signal using a low-pass filter, which was used to define the range of reference numeral 26, and reference numeral 28 denotes the data density within the signal calculated from the data indicated by reference numeral 27. The range indicated by reference numeral 26 is determined as a range in which the value indicated by reference numeral 27 is below a certain threshold.

[0041] The conversion unit 101 converts the current signal measured by the measurement device 20 into image data based on a predetermined conversion process, thereby visually representing the characteristics of the same current signal that occurs when a substance passes through the nanopore. In this embodiment, the current signal measured by the measurement device 20 is converted into image data through four conversion processes. Hereinafter, the image data converted by each conversion process will be referred to as sub-image data. In this embodiment, each sub-image data has a size of 256 pixels horizontally and 64 pixels vertically, but the size of each sub-image data is not limited to this example.

[0042] (Sub-image data 1, 2) The conversion unit 101 converts the current signal into sub-image data 1 and 2, which represent the current value in black and white and are expressed on a scale spanning a predetermined range, for example, 100 ms to 1 ms, so that the entire current signal and its details can be grasped. The sub-image data 1 is image data focusing on the leading edge of the current signal, and the sub-image data 2 is image data focusing on the trailing edge of the current signal. By including more types of information in the image, the sub-image data 1 and 2 enable the trained model 30 to make complex judgments.

[0043] First, the conversion to sub-image data 1 will be described. To align the data length to 100 ms, if the length of the current signal is 100 ms or more, the conversion unit 101 removes the data after 100 ms. If the length of the current signal is 100 ms or less, the conversion unit 101 fills the depth information from the end of the current signal to 100 ms with zeros. The conversion unit 101 reduces and inserts 100 ms of data for the top row of 64 pixels in sub-image data 1. When inserting the current signal data into sub-image data 1, the conversion unit 101 creates an image in which the depth closer to 0 appears black and the depth closer to 1 appears white. The conversion unit 101 gradually expands and inserts the data so that the first row is 100 ms and the 64th row (the bottom row) is 1 ms. The conversion unit 101 reduces and inserts 1 + (100 - 1) (62 / 63) = 98.4 ms of data from the left for the second row from the top. Furthermore, for the third row from the top, the conversion unit 101 reduces the data by 1+(100-1)(61 / 63)=96.9 ms from the front and inserts it. In this way, for the xth row (x is an integer from 1 to 64), the conversion unit 101 reduces the data by 1+(100-1)((64-x) / 63)=1 ms from the front and inserts it.

[0044] Next, the conversion to sub-image data 2 will be described. As with sub-image data 1, the conversion unit 101 aligns the data length to 100 ms. If the length of the current signal is 100 ms or more, the conversion unit 101 removes the data 100 ms or less from the end. If the length of the current signal is 100 ms or more, the conversion unit 101 aligns the end of the current signal with 100 ms and fills in the depth information from 0 ms to the start of the signal with 0. The conversion unit 101 reduces and inserts 100 ms of data for the top row of the 64 rows of pixels in sub-image data 2. When inserting the current signal data into sub-image data 2, the conversion unit 101 creates an image in which the closer the depth is to 0, the darker it appears, and the closer it is to 1, the whiter it appears. The conversion unit 101 gradually expands and substitutes the data so that the first row is 100 ms long and the 64th row (the bottom row) is 1 ms long. For the second row from the top, the conversion unit 101 reduces and inserts data by 1+(100-1)(62 / 63)=98.4 ms from the end. For the third row from the top, the conversion unit 101 reduces and inserts data by 1+(100-1)(61 / 63)=96.9 ms from the end. In this way, for the xth row (x is an integer from 1 to 64), the conversion unit 101 reduces and inserts data by 1+(100-1)((64-x) / 63)=1 ms from the end.

[0045] Fig. 9 is a diagram showing an example of sub-image data 1, and Fig. 10 is a diagram showing an example of sub-image data 2. By converting the current signal into sub-image data 1 and 2, the conversion unit 101 can generate image data that focuses on the leading and trailing edges of the current signal.

[0046] (Sub-image data 3) The sub-image data 3 is an image that represents the change over time of frequency components based on the result of Fourier transform of the current signal. In order to perform STFT (Short Time Fourier Transform), the transform unit 101 sets the width of the window for applying the Fourier transform to the current signal to 1 / 8 to 1 / 4 of the length of the current signal and the number of data points to 2. n Set the value to 2 seconds. n / 250000 [s]. The conversion unit 101 also sets the window movement width to 1 / 16 of the window width. The conversion unit 101 performs STFT between 0 and 10000 Hz, and obtains the result with the range of -100 to -30 dB as 0 to 1. In this case, all values ​​below -100 dB are set to 0, and all values ​​above -30 dB are set to 1. The conversion unit 101 stretches the STFT result to a size of 256 pixels horizontally and 64 pixels vertically before inserting it.

[0047] 11 is a diagram showing an example of the sub-image data 3. When inputting data into the sub-image data 3, the conversion unit 101 inputs the data so that the closer the value is to 0, the darker it appears, and the closer the value is to 1, the whiter it appears.

[0048] (Sub-image data 4) The sub-image data 4 is an image that represents various characteristics of the current signal. In this embodiment, the conversion unit 101 converts the current signal into sub-image data 4 that represents various statistical information related to the depth of the current signal, the voltage measured by the measurement device 20 when the current signal was obtained, the STFT window, the sampling rate when the current signal was obtained, the length of the current signal, etc.

[0049] For example, the conversion unit 101 converts the median depth of the current signal into one row of data consisting of 256 points. As an example, the conversion unit 101 assigns values ​​of -0.5, -0.492, -0.484, ..., 1.5 to each point, and the conversion unit 101 assigns a value of 1 to each point that is closer to the median depth of the current signal, and a value of 0 to each point that is farther away. The conversion unit 101 enters the resulting 256 pieces of data with values ​​between 0 and 1 into the top row of the sub-image data 4. When the data is entered into the sub-image data 4, the closer the value is to 0, the darker it will be, and the closer it is to 1, the whiter it will be.

[0050] The conversion unit 101 similarly creates data on the average value, maximum value, minimum value, standard deviation, first quartile, third quartile, asymmetry of the current distribution, fat tail of the current distribution, STFT window size, STFT window movement amount, length of the current signal, measured voltage, average true conductance value outside the signal, and sampling rate during measurement, and enters this data in the second, third, etc. rows from the top of the sub-image data 4. Table 1 shows the information for each row of the sub-image data 4. Since the sampling rate and length of the current signal can take on a wide range of values, they are expressed on multiple scales.

[0051] [Table 1]

[0052] The conversion unit 101 may convert the current signal into sub-image data 4 by enlarging the image data expressing each value in this manner so that it has 64 vertical pixels. Note that the information expressed as the sub-image data 4 is not limited to this example. Furthermore, the values ​​indicated at the left and right ends of each row of the sub-image data 4 are not limited to those shown in Table 1. For example, the conversion unit 101 may further add percentiles in 10% increments to the depth value. Furthermore, for example, the conversion unit 101 may further add the frequency value of a noise filter of the measurement device 20. Furthermore, for example, the conversion unit 101 may further add the frequency value of a filter used to extract the current signal of the measurement device 20.

[0053] The conversion unit 101 joins together the sub-image data generated through the above conversion to form a single image data. In this embodiment, each sub-image data is 256 pixels wide and 64 pixels high, so the single image data formed by joining the sub-image data has a size of 256 pixels wide and 256 pixels high.

[0054] In addition to the above-described sub-image data 1 to 4, the conversion unit 101 may also convert the current signal into sub-image data that represents the results of frequency analysis of the current signal using, for example, wavelet analysis. Wavelet analysis is an analysis method used in image processing, voice recognition, seismic wave analysis, etc., and, like STFT, can extract time-varying frequency components from the current signal.

[0055] The discrimination unit 102 inputs the image data generated by the conversion unit 101 into a trained model 30 that has been generated in advance by machine learning using the image data generated by the conversion unit 101 by converting from the current signal, and uses the output from the trained model 30 to discriminate the substance that has passed through the nanopore of the measuring device 20.

[0056] Here we present an example of machine learning using image data generated from current signals. When the target of judgment is miRNA, all images generated from data obtained from the results of measuring one type of miRNA are labeled with the name of the measured miRNA, and 80% are used for training and 20% are used for evaluation, and training is carried out using a convolutional neural network (CNN). Learning stops at the generation (epoch) when a predetermined judgment accuracy is achieved, and the completed trained model is designated as 30.

[0057] The discrimination unit 102 uses the trained model 30, in which the neural network has been trained in this manner, to discriminate substances that have passed through the nanopore of the measuring device 20, thereby being able to discriminate the type of miRNA more easily and accurately than conventional discrimination methods.

[0058] Next, the operation of the information processing device 10 will be described.

[0059] 14 is a flowchart showing the flow of information processing by the information processing device 10. The CPU 11 reads out an information processing program from the ROM 12 or the storage 14, loads it into the RAM 13, and executes it, thereby performing information processing.

[0060] In step S101, the CPU 11 acquires from the measurement device 20 a current signal obtained when a solution of miRNA is passed through the nanopore.

[0061] Following step S101, in step S102, the CPU 11 converts the acquired current signal into a plurality of sub-image data. In this embodiment, the CPU 11 converts the acquired current signal into sub-image data focusing on a predetermined range at the leading end of the current signal, sub-image data focusing on a predetermined range at the trailing end of the current signal, sub-image data based on the results of a Fourier transform of the current signal, and sub-image data based on an expression of the information on the current signal.

[0062] Following step S102, in step S103, the CPU 11 joins the sub-image data to generate image data. In this embodiment, the CPU 11 joins four sub-image data vertically to generate image data of 256 pixels vertically and 256 pixels horizontally.

[0063] Following step S103, in step S104, the CPU 11 inputs the image data generated in step S103 into the trained model 30. The trained model 30 is a model generated by training a neural network through machine learning using image data generated from a current signal obtained when a solution of miRNA from the measuring device 20 is passed through a nanopore in the same manner as in the processing of steps S102 and S103.

[0064] Following step S104, the CPU 11 acquires an output from the trained model 30 in step S105.

[0065] Following step S105, in step S106, the CPU 11 identifies the substance that has passed through the nanopore, for example, miRNA, based on the output from the trained model 30.

[0066] By executing a series of processes, the information processing device 10 can identify the substance that has passed through the nanopore of the measuring device 20, thereby making it possible to identify the type of substance more easily and accurately than conventional identification methods.

[0067] Figure 15 is a graph showing the accuracy of the results of identifying miRNA types using an information processing device based on the current signal generated when miRNA passes through a nanopore. The bar graphs r1, r2, r3, and r4 show the accuracy of the results of identifying miRNA types using a trained model 30 obtained by converting the current signal into sub-image data 1, sub-image data 2, sub-image data 3, and sub-image data 4, respectively, and performing 30 generations (epochs) of training using a convolutional neural network (CNN). The bar graphs r12, r13, and r14 show the accuracy of miRNA identification using a trained model 30 obtained by converting the current signal into sub-image data 1 and 2, sub-image data 1 and 3, and sub-image data 1 and 4, respectively, and then connecting the two sub-image data to form a single image data set, and performing similar training. Furthermore, r123, r124, r134, and r234 respectively indicate the accuracy of miRNA discrimination when using a trained model 30 obtained by converting a current signal into three types of sub-image data, connecting them together to form a single image data set, and performing similar training. r1234 indicates the accuracy of discrimination when using image data in which a current signal is converted into four types of sub-image data and connected together. As is clear from the results of Figure 15, the accuracy of discrimination is improved when using image data in which multiple sub-image data are connected together compared to when using one sub-image data set, and the accuracy of discrimination is highest when using image data in which four sub-image data are connected together.

[0068] Although the embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings, the technical scope of the present disclosure is not limited to such examples. It is clear that a person skilled in the art of the present disclosure can conceive of various modifications or alterations within the scope of the technical idea described in the claims, and it is understood that these modifications or alterations also naturally fall within the technical scope of the present disclosure.

[0069] Furthermore, the effects described in the above embodiments are explanatory or exemplary and are not limited to those described in the above embodiments. In other words, the technology according to the present disclosure may achieve other effects that are obvious to a person skilled in the art of the present disclosure from the description in the above embodiments, in addition to or instead of the effects described in the above embodiments.

[0070] In the above embodiments, the information processing performed by the CPU after reading the software (program) may be performed by various processors other than the CPU. Examples of such processors include programmable logic devices (PLDs) whose circuit configuration can be changed after fabrication, such as field-programmable gate arrays (FPGAs), and dedicated electrical circuits, such as application-specific integrated circuits (ASICs), which are processors with circuit configurations specifically designed to perform specific processing. The information processing may be performed by one of these processors, or by a combination of two or more processors of the same or different types (e.g., multiple FPGAs, or a combination of a CPU and an FPGA). The hardware structure of these processors is, more specifically, an electrical circuit that combines circuit elements such as semiconductor devices.

[0071] In addition, in each of the above embodiments, the information processing program is described as being pre-stored (installed) in a ROM or storage, but this is not limiting. The program may be provided in a form recorded on a non-transitory recording medium such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), or a USB (Universal Serial Bus) memory. The program may also be downloaded from an external device via a network. [Explanation of symbols]

[0072] 10. Information processing equipment 20 Measuring Equipment 30 pre-trained models

Claims

1. a conversion unit that converts a current signal generated when a substance passes through the nanopore into image data by a predetermined conversion process; a discrimination unit that inputs the image data into a trained model generated by machine learning using the image data and discriminates the substance using an output from the trained model; An information processing device comprising:

2. The information processing device according to claim 1 , wherein the conversion unit converts the same current signal into one image data by connecting a plurality of sub-image data pieces obtained by converting the same current signal based on a plurality of the conversion processes.

3. The information processing device according to claim 2 , wherein the conversion unit converts a predetermined range of the current signal into the sub-image data as one of the plurality of conversion processes.

4. The information processing device according to claim 3 , wherein the conversion unit converts the current signal into the sub-image data by extending a predetermined range of a tip portion of the current signal as one of the plurality of conversion processes.

5. The information processing device according to claim 3 , wherein the conversion unit converts the current signal into the sub-image data by extending a predetermined range of a terminal end portion of the current signal as one of the plurality of conversion processes.

6. The information processing device according to claim 2 , wherein the conversion unit converts the current signal into the sub-image data representing a change in frequency components over time based on a result of a Fourier transform of the current signal, as one of the plurality of conversion processes.

7. The information processing device according to claim 2 , wherein the conversion unit converts the current signal into the sub-image data representing a plurality of pieces of statistical information of the current signal as one of the plurality of conversion processes.

8. The information processing device according to claim 1 , wherein the substance is a substance that can pass through the nanopore.

9. The information processing device according to claim 8 , wherein the substance is a polynucleotide, a peptide, or a protein.

10. The processor: A current signal generated when a substance passes through the nanopore is converted into image data by a predetermined conversion process. The image data is input into a trained model generated by machine learning using the image data, and the substance is identified using an output from the trained model. A method for processing information.

11. On the computer, A current signal generated when a substance passes through the nanopore is converted into image data by a predetermined conversion process. The image data is input into a trained model generated by machine learning using the image data, and the substance is identified using an output from the trained model. An information processing program that executes processing.

Images