Information processing methods, programs, and information processing systems

Abstract

To analyze data that is increased by multicoloring.SOLUTION: An information processing device according to an embodiment includes: a dimensional compression unit that executes dimensional compression on each of multiple pieces of spectral data including a fluorescent component emitted from each of multiple particles labeled with one or more fluorescent dyes, respectively; an initial value determination unit that determines an initial value for each of multiple nodes on the basis of a result of the dimensional compression; and a clustering unit that executes clustering on the multiple pieces of spectral data using the initial value.SELECTED DRAWING: Figure 12

Description

Technical Field

[0001] The present disclosure relates to an information processing method, a program, and an information processing system.

Background Art

[0002] In fields such as medicine or biochemistry, it has become common to use a flow cytometer to quickly measure the characteristics of each of a large number of particles. A flow cytometer is a device that irradiates particles such as cells or beads flowing through a flow cell with a light beam, detects fluorescence or scattered light emitted from the particles, and optically measures the characteristics of each particle.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] In recent flow cytometers, the multicolorization of staining particles such as cells with a plurality of fluorescent dyes has been advanced. However, as the multicolorization progresses, the number of fluorescent substances that can be measured at one time increases, resulting in a combinatorial explosion. As a result, there is a problem that the amount of data to be processed increases and analysis becomes difficult.

[0005] Therefore, the present disclosure proposes an information processing apparatus, an information processing method, a program, and an information processing system capable of analyzing data increased by multicolorization.

Means for Solving the Problems

[0006] To solve the above problems, one embodiment of the information processing device according to the present disclosure includes: a dimensionality reduction unit that performs dimensionality reduction on each of a plurality of spectral data containing fluorescent components emitted from each of a plurality of particles, each labeled with one or more fluorescent dyes; an initial value determination unit that determines the initial value of each of the plurality of nodes based on the result of the dimensionality reduction; and a clustering unit that performs clustering on the plurality of spectral data using the initial values.

[0007] Furthermore, the effects described above are not necessarily limited, and any of the effects described herein, or any other effects that can be inferred from this specification, may be achieved in conjunction with or in lieu of the effects described above. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing the configuration of the information processing system according to the first embodiment. [Figure 2] This figure shows an example of a fluorescence spectrum. [Figure 3] This figure shows an example of a pre-processing parameter table. [Figure 4] This figure shows an example of a spectral plot. [Figure 5] This figure shows an example of clustering results. [Figure 6] This is a diagram illustrating the fluorescence separation process. [Figure 7] This figure shows an example of how the analysis results are displayed. [Figure 8] This is a flowchart showing the pre-processing flow by the pre-processing unit. [Figure 9] This flowchart shows the processing flow by the clustering unit. [Figure 10] This flowchart shows the flow of the fluorescence separation process performed by the fluorescence separation unit. [Figure 11] This is a schematic diagram illustrating the entire process from initialization to training in FlowSOM. [Figure 12]It is a block diagram showing the configuration of an information processing system according to a second embodiment. [Figure 13] It is a flowchart showing an operation example of a preprocessing / fluorescence separation unit according to a second embodiment. [Figure 14] It is a flowchart showing an example of clustering processing executed by a clustering processing unit according to a second embodiment. [Figure 15] It is a diagram for explaining the operation when initializing the representative node vectors of each representative node according to this embodiment, which was explained in step S222 of FIG. 14. [Figure 16] It is a flowchart showing an example of clustering by batch learning according to this embodiment, which was explained in step S223 of FIG. 14. [Figure 17] It is a diagram for explaining the operation when determining the number of clusters using consensus clustering, which was explained in step S224 of FIG. 14. [Figure 18] It is a block diagram showing an example of the hardware configuration of an information processing apparatus.

Embodiments for Carrying Out the Invention

[0009] Hereinafter, embodiments of the present disclosure will be described in detail based on the drawings. Further, the present disclosure will be described in accordance with the order of items shown below. 1. First Embodiment 1.1 Configuration of Information Processing System 1.2 Operation of Information Processing Apparatus 1.3 Actions and Effects 2. Second Embodiment 2.1 Configuration of Information Processing System 2.2 Operation Examples of Information Processing Apparatus 2.2.1 Operation Example of Preprocessing / Fluorescence Separation Unit 2.2.2 Operation Example of Clustering Processing Unit 2.2.3 Regarding Initialization of Representative Node Vectors 2.2.4 Regarding Clustering by Batch Learning 2.2.5 Determining the number of clusters using consensus clustering 2.3 Action and Effects 3. Hardware configuration of the information processing device

[0010] 1. First Embodiment Recent flow cytometers have become capable of obtaining richer information by staining particles such as cells with multiple fluorescent dyes and measuring many fluorescence signals at once. On the other hand, the increasing number of colors in these flow cytometers has led to the limitation of analysis using conventional manual gating. For example, when plotting n types of fluorescent dyes in a two-dimensional coordinate system, the number of possible combinations of fluorescent dyes is nC2. In other words, there are 15 combinations with 6 colors, while there are 190 combinations with 20 colors.

[0011] One possible method for analyzing such large amounts of data is to replace traditional manual gating with automatic classification using clustering, where data is classified by clustering before analysis.

[0012] However, in general clustering, any data is classified into one of the clusters, meaning that even data that lies at the boundary between two or more clusters and is difficult to classify will be classified into one of the clusters. Therefore, when classifying data acquired by a flow cytometer using clustering, there is a possibility that some degree of misclassification may occur.

[0013] On the other hand, flow cytometers include a filter method that detects the intensity of one or more specific wavelengths in fluorescence, and a spectral method that can obtain more information by detecting the fluorescence emitted from each fluorescent dye as a wavelength spectrum. In the spectral method, since the fluorescence intensity for each wavelength is obtained, it is possible to obtain many features from a single particle (also called a sample). Therefore, by adopting the spectral method, more detailed analysis becomes possible.

[0014] However, spectral data obtained using the spectral method exhibits an exponential rather than linear change in fluorescence intensity for each wavelength. Therefore, scaling the spectral data is necessary to display it in a way that is easily recognizable to the user. On the other hand, clustering classifies data based on the distance between them. Therefore, if spectral data with exponentially changing fluorescence intensity for each wavelength is clustered directly, the difference in high fluorescence intensity strongly affects the overall distance between data, potentially leading to misclassification. As a result, the clustering results may not accurately correspond to the scaled spectral data displayed.

[0015] Therefore, in this embodiment, an information processing device, information processing method, program, and information processing system that enable a more accurate correspondence between the display results of spectral data and the clustering results will be described with examples.

[0016] Specifically, a logice transformation is performed as a preprocessing step on the spectral data acquired by the flow cytometer. Then, clustering is performed using the logice-transformed spectral data, and the resulting clustering results are displayed to the user. In this embodiment, it is possible to decide whether or not to perform clustering on the portion of the spectral data with the largest values, thereby enabling clustering in a way that more accurately corresponds the displayed spectral data and the clustering results. The information processing apparatus, information processing method, program, and information processing system according to this embodiment will be described in detail below with reference to the drawings.

[0017] However, the technology relating to this disclosure does not need to solve all of the above-mentioned problems simultaneously. Therefore, any technology that solves some or all of the above-mentioned problems by incorporating some or all of the configuration described later is considered to fall within the technical scope of this disclosure.

[0018] 1.1 Configuration of the Information Processing System Figure 1 is a block diagram showing the configuration of the information processing system according to this embodiment. As shown in Figure 1, the information processing system 4 comprises an information processing device 1 and a measuring device 3.

[0019] The measuring device 3 is a measuring device capable of detecting fluorescence spectra of various colors from cells, etc., that are to be measured. The measuring device 3 is, for example, a flow cytometer. The sample measured by the flow cytometer may be living organism-derived particles such as cells, microorganisms, and living organism-related particles. For example, cells may be animal cells (e.g., hematopoietic cells), plant cells, etc. For example, microorganisms may be bacteria such as Escherichia coli, viruses such as tobacco mosaic virus, fungi such as yeast, etc. Living organism-related particles may be particles that make up cells, such as chromosomes, liposomes, mitochondria, and various organelles (cellular organelles). Living organism-related particles may also include living organism-related macromolecules such as nucleic acids, proteins, lipids, glycans, and complexes thereof. These living organism-derived particles may be spherical or non-spherical in shape, and their size and mass are not particularly limited.

[0020] Furthermore, the measurement sample may be industrially synthesized particles such as latex particles, gel particles, or industrial particles. For example, industrially synthesized particles may be organic resin materials such as polystyrene and polymethyl methacrylate, inorganic materials such as glass, silica, and magnetic materials, or particles synthesized from metals such as gold colloid and aluminum. Similarly, these industrially synthesized particles may be spherical or non-spherical in shape, and there are no particular limitations on their size and mass.

[0021] Prior to measuring the fluorescence spectrum, the sample may be labeled (stained) with one or more fluorescent dyes. Labeling of the sample with fluorescent dyes may be carried out by known methods. Specifically, if the sample is a cell, the target cell can be labeled with a fluorescent dye by mixing it with a fluorescently labeled antibody that selectively binds to an antigen present on the cell surface, thereby binding the fluorescently labeled antibody to the antigen on the cell surface. Alternatively, the target cell can be labeled with a fluorescent dye by mixing it with a fluorescent dye that is selectively taken up by a specific cell.

[0022] Fluorescently labeled antibodies are antibodies to which a fluorescent dye has been conjugated as a label. Fluorescently labeled antibodies may be antibodies to which the fluorescent dye has been directly conjugated. Alternatively, fluorescently labeled antibodies may be biotin-labeled antibodies to which a fluorescent dye conjugated with avidin has been conjugated via the avidin-biodin reaction. Both polyclonal and monoclonal antibodies can be used.

[0023] The fluorescent dyes used to label cells are not particularly limited, and at least one known dye used for staining cells can be used. For example, as fluorescent dyes, phycoerythrin (PE), fluorescein isothiocyanate (FITC), PE-Cy5, PE-Cy7, PE-Texas Red (registered trademark), allophycocyanin (APC), APC-Cy7, ethidium bromide, propidium iodide, Hoechst (registered trademark) 33258, Hoechst (registered trademark) 33342, DAPI (4',6-diamidino-2-phenylindole), acridine orange, chromomycin, mithramycin, olivomycin, pyronin Y, and thiazole orange can be used. Ingredients such as orange, rhodamine 101, isothiocyanate, BCECF, BCECF-AM, C.SNARF-1, C.SNARF-1-AMA, aequorin, Indo-1, Indo-1-AM, Fluo-3, Fluo-3-AM, Fura-2, Fura-2-AM, oxonol, Texas Red (registered trademark), rhodamine 123, 10-N-nony-acridin orange, fluorescein, fluorescein diacetate, carboxyfluorescein, carboxyfluorescein diacetate, carboxydichlorofluorescein, and carboxydichlorofluorescein diacetate can be used. Furthermore, derivatives of the fluorescent dyes mentioned above can also be used.

[0024] A flow cytometer comprises a laser light source that emits laser light having a wavelength capable of exciting a fluorescent dye labeled to a sample S, a flow cell that allows the sample S to flow in one direction, and a photodetector that receives one or more of the fluorescence, phosphorescence, and scattered light from the sample S irradiated with laser light.

[0025] The laser light source is, for example, a semiconductor laser light source that emits laser light of a predetermined wavelength. Multiple laser light sources may be provided. When multiple laser light sources are provided, the positions in the flow cell where the laser light from the laser light sources irradiates may be the same or different. However, when the laser light from multiple laser light sources irradiates different positions, it becomes possible to detect the light from the measurement sample S with different photodetectors, and in such cases, even when dyes that emit fluorescence at close wavelengths are used, it becomes possible to measure the fluorescence spectra of each without mixing. The laser light emitted from the laser light source may be either pulsed light or continuous light. For example, the laser light source may be multiple semiconductor laser light sources that emit laser light with wavelengths of 480 nm and 640 nm, respectively.

[0026] A flow cell is a channel that aligns and passes multiple measurement samples S in one direction. Specifically, a flow cell aligns and passes multiple measurement samples S in one direction by flowing a sheath liquid containing the measurement samples S at high speed as a laminar flow. Flow cells can be formed within a microchip or cuvette.

[0027] The photodetector detects light from a sample S irradiated with laser light by photoelectric conversion. The light from the sample S may include at least one of fluorescence, phosphorescence, and scattered light.

[0028] For example, the photodetector may include a detector that detects scattered light LS, including forward and side scattered light, from the measurement sample S, and a photodetector array that detects fluorescence from the measurement sample S.

[0029] The detector may be a known photoelectric conversion element such as a CCD (Charge Coupled Device), CMOS (Complementary Metal Oxide Semiconductor), or photodiode. The photodetector array may be composed of multiple independent detection channels with different wavelength ranges for detecting light. Specifically, the photodetector array may be a photodetector array in which multiple PMTs (Photo Multiplier Tubes) or photodiodes with different wavelength ranges for detecting are arranged in one dimension, or an image sensor in which pixels are arranged in a two-dimensional grid. The photodetector array converts fluorescence from the measurement sample S, which has been spectrally separated for each wavelength by a spectroscopic element such as a prism or grating, into photoelectric energy.

[0030] In a flow cytometer with the above configuration, first, laser light emitted from a laser light source is irradiated onto the sample S being measured as it flows through the flow cell. Upon irradiation with laser light, the sample S emits scattered light and fluorescence (or phosphorescence). The scattered light emitted from the sample S is detected by a detector. Meanwhile, the fluorescence emitted from the sample S is spectrally separated into light of each wavelength by a spectroscopic element and then received by a photodetector array. This allows the fluorescence spectrum emitted from the sample S to be detected.

[0031] In the following explanation, the object of measurement is assumed to be cells. Measurement device 3 is a spectral flow cytometer that detects the fluorescence spectrum of each fluorescent dye emitted from cells by, for example, flowing fluorescently stained cells at high speed through a flow cell and irradiating the flowing cells with light. Figure 2 shows an example of a fluorescence spectrum. As shown in Figure 2, the fluorescence spectrum is expressed by the fluorescence intensity of each channel corresponding to the wavelength.

[0032] The measuring device 3 outputs the detected fluorescence spectrum as measurement data 2. Measurement data 2 includes fluorescence spectrum data for each cell. The measuring device 3 transfers the measurement data 2 to, for example, the information processing device 1.

[0033] The information processing device 1 acquires measurement data 2 measured by the measuring device 3, performs analysis, and displays the analysis results. The information processing device 1 and the measuring device 3 may be connected via a network, and the information processing device 1 may acquire the measurement data 2 via the network.

[0034] The information processing device 1 includes a pre-processing unit 11, a pre-processing parameter table 12, a spectrum output unit 13, a clustering processing unit 14, a clustering result presentation unit 15, a fluorescence separation unit 16, and a normal analysis presentation unit 17. Note that all or part of these functional units may be performed in the cloud. For example, the pre-processing parameter table 12, the clustering processing unit 14, and the fluorescence separation unit 16 may be performed in the cloud. In this case, the measurement data 2 is also transferred to the cloud.

[0035] The preprocessing unit 11 performs preprocessing of the spectral data according to the selected parameters. Here, preprocessing refers to a coordinate transformation that displays the actual observed values. The transformation can be, for example, a simple log10 transformation, or a logice transformation that takes into account the characteristics of the observation instrument. The parameters are W, T, M, and A.

[0036] W is a value that linearly displays values ​​near zero. T is the maximum fluorescence intensity, for example, 10 4 M is the maximum value of the displayed coordinates after the transformation. A is the smallest negative value to transform.

[0037] The pre-processing parameter table 12 is a table that stores pre-processing parameters. Figure 3 shows an example of the pre-processing parameter table 12. As shown in Figure 3, the pre-processing parameter table 12 stores multiple combinations (also called parameter sets) of values ​​for W, T, M, and A. The parameter ID is an identifier that identifies the combination of parameters.

[0038] The preprocessing unit 11 performs a logice transformation using the combination of W, T, M, and A selected by the user from the preprocessing parameter table 12. Alternatively, the preprocessing unit 11 may use default values ​​instead of selecting parameters from the preprocessing parameter table 12. The user can also specify values ​​other than those stored in the preprocessing parameter table 12. The preprocessing unit 11 performs preprocessing each time the user changes the preprocessing parameters. The preprocessing unit 11 may also perform nonlinear transformations such as log transformations or bi-exponential transformations instead of logice transformations.

[0039] The spectral output unit 13 generates a spectral plot image using the spectral data pre-processed by the pre-processing unit 11 and displays the generated image. Figure 4 shows an example of a spectral plot. The spectral plot shows the detection wavelength (wavelength) on the horizontal axis and the fluorescence intensity on the vertical axis, and represents information about the number of particles (number of events or density) (population information) using the intensity of color or hue.

[0040] In Figure 4, "LD488" on the vertical axis indicates fluorescence when irradiated with laser light with a wavelength of 488 nm (nanometers), and "_A" indicates that the measured value is the cumulative intensity. Also, in Figure 4, the number of particles is represented by shading, but on the actual screen, the number of particles is represented by color.

[0041] As shown on the vertical axis of Figure 4, in the spectral plot, the vertical axis corresponding to fluorescence intensity is displayed after logice transformation. Therefore, although arrows 91 and 92 are displayed as having the same length in Figure 4, the ranges they represent are quite different. In other words, when the vertical axis is linear, their lengths are completely different, with arrow 91 being much longer than arrow 92.

[0042] The clustering processing unit 14 performs cell clustering using spectral data that has been pre-processed by the pre-processing unit 11. The clustering processing unit 14 classifies the spectral data into K clusters, for example, by specifying K, as in K-means. Alternatively, the clustering processing unit 14 may automatically determine the number of divisions, as in FlowSOM (Self-Organizing Map).

[0043] Alternatively, the clustering processing unit 14 may perform clustering by performing dimensionality reduction, such as T-SNE, and then performing gating on the result of dimensionality reduction. Alternatively, the clustering processing unit 14 may perform two-stage clustering, such as meta-clustering, and use two definitions of clusters, such as a cluster ID and a meta-cluster ID. Here, a meta-cluster is a collection of clusters.

[0044] The clustering processing unit 14 performs clustering using the logicle-converted spectral data, making it easier to associate the display results of the spectral data with the clustering results.

[0045] The clustering result display unit 15 displays the clustering results from the clustering processing unit 14 on a display device. The clustering result display unit 15 visualizes the number of classifications, or which cell groups belong to which classification. Figure 5 shows an example of clustering results. Figure 5 shows the case when clustering is performed using FlowSOM. In Figure 5, circles represent clusters, and clusters are classified into metaclusters M#1 to M#5 with different shading. In the actual display, metaclusters M#1 to M#5 are displayed in different colors. By displaying the clustering results, the user can learn about the distribution of clusters and metaclusters.

[0046] The fluorescence separation unit 16 acquires the measurement data 2 and performs fluorescence separation processing (also called unmixing). Figure 6 is a diagram illustrating the fluorescence separation processing. Figure 6(a) shows the measured fluorescence spectrum. The measured fluorescence spectrum is a superposition of, for example, the fluorescence spectra of three fluorescence sources, as shown in Figure 6(b).

[0047] Therefore, the fluorescence separation unit 16 separates the fluorescence into three spectra, #1 to #3, using the reference spectrum shown in Figure 6(c). Here, the reference spectrum is the fluorescence spectrum for each fluorescence. The separated spectra of fluorescence #1 to #3 are shown in Figure 6(d). The fluorescence separation unit 16 calculates the fluorescence intensity using the spectra separated for each fluorescence, for example, by taking a weighted average. Figure 6(e) shows the fluorescence intensities of fluorescence #1 to #3 calculated by the fluorescence separation unit 16.

[0048] Normally, the analysis display unit 17 performs analysis using the fluorescence intensity separated by the fluorescence separation unit 16 and displays the analysis results on the display device. Figure 7 shows an example of the display of analysis results. Figure 7 shows a two-dimensional plot with APC-Cy7::CD24 and PE-Dazzle594::CD38 as the two axes. Here, APC-Cy7::CD24 and PE-Dazzle594::CD38 are fluorescent dye-labeled antibodies used to measure fluorescence intensity. "APC-Cy7" and "PE-Dazzle594" are fluorescent dyes, and "CD24" and "CD38" are antibodies. The user can see the distribution of cells related to the two fluorescent dyes from the two-dimensional plot.

[0049] Furthermore, the information processing device 1 may perform pre-processing and clustering on the cell populations contained in a region selected by the user in a two-dimensional plot, and display the clustering results. The information processing device 1 may also perform pre-processing and clustering on the cell populations contained in a region selected by the user in a spectral plot, and display the clustering results. In particular, characteristic areas within the spectral plot, such as densely populated regions, are often selected by the user.

[0050] 1.2 Operation of the Information Processing Device Next, the operation of the information processing device 1 will be described with reference to Figures 8 to 10. Figure 8 is a flowchart showing the pre-processing flow by the pre-processing unit 11. As shown in Figure 8, the pre-processing unit 11 selects pre-processing parameters based on user instructions (step S1).

[0051] Then, the preprocessing unit 11 performs preprocessing of the spectral data using the preprocessing parameters (step S2). The preprocessing unit 11 then determines whether the user has changed the preprocessing parameters (step S3), and if so, changes the preprocessing parameters and performs preprocessing of the spectral data (step S4), and returns to step S3.

[0052] On the other hand, if the user does not change the pre-processing parameters, the pre-processing unit 11 determines whether all target cells have been processed (step S5). If there are unprocessed target cells, it returns to step S2 and processes other cells. If all target cells have been processed, the pre-processing unit 11 instructs the spectral output unit 13 to present the pre-processed spectral data to the user (step S6). The pre-processing unit 11 also passes the pre-processed spectral data to the clustering unit 14.

[0053] In this way, the pre-processing unit 11 performs pre-processing on the spectral data, allowing the clustering unit 14 to correlate the clustering results with the display results of the spectral data.

[0054] Figure 9 is a flowchart showing the processing flow by the clustering processing unit 14. As shown in Figure 9, the clustering processing unit 14 acquires pre-processed spectral data from the pre-processing unit 11 (step S11) and performs clustering processing (step S12). Then, the clustering processing unit 14 instructs the clustering result presentation unit 15 to present the clustering results (step S13).

[0055] In this way, the clustering processing unit 14 performs clustering using pre-processed spectral data, so the clustering results can be correlated with the display results of the spectral data.

[0056] Figure 10 is a flowchart showing the flow of the fluorescence separation process performed by the fluorescence separation unit 16. As shown in Figure 10, the fluorescence separation unit 16 acquires a reference spectrum (step S21). Then, the fluorescence separation unit 16 acquires spectral data for one cell (step S22) and performs fluorescence separation processing (step S23). Then, the fluorescence separation unit 16 determines whether all cells have been processed (step S24). If there are cells that have not been processed, it returns to step S22; if all cells have been processed, it terminates the fluorescence separation process.

[0057] As described above, the fluorescence separation unit 16 performs the fluorescence separation process, so the normal analysis and display unit 17 can analyze the fluorescence data and display the analysis results.

[0058] 1.3 Action and Effects As described above, according to one embodiment, the pre-processing unit 11 acquires spectral data and performs logice conversion as pre-processing. Then, the clustering processing unit 14 performs clustering using the spectral data that has undergone logice conversion. Finally, the clustering result presentation unit 15 displays the clustering results on the display device. Therefore, the information processing device 1 can prevent clustering from being performed on the portion of the spectral data with large values. For this reason, the information processing device 1 can perform clustering so that the displayed spectral data results correspond to the clustering results.

[0059] 2. Second Embodiment As mentioned above, recent advancements in flow cytometry have led to the development of multispectral imaging, which stains particles such as cells with multiple fluorescent dyes.

[0060] As flow cytometers become more multispectral, the number of fluorescent substances that can be measured at once increases, resulting in a combinatorial explosion that makes manual analysis difficult. For example, when measuring N (where N is a natural number) colors at once, if we try to handle each color in two dimensions, the number of combinations becomes n(n-1) / 2, which increases roughly in proportion to half the square of the number of colors N.

[0061] As mentioned above, one possible method for analyzing data that has grown due to such a combinatorial explosion is to classify the data using clustering methods such as FlowSOM before analyzing it.

[0062] Figure 11 is a schematic diagram illustrating the sequence of events from initialization to training in FlowSOM. In the example shown in Figure 11, 100 vectors (coordinate values) contained in the data set 900 to be analyzed are classified into nine representative nodes N11 to N33 arranged in a 3x3 grid in a 2D coordinate system in the initial state.

[0063] In the example shown in Figure 11, first, a predetermined number of data points (9 in this example) are selected from the data set 900 by random sampling, and these selected data points are set as the initial node vector N_ij for representative nodes N11 to N33 (step S1).

[0064] Next, training is performed to update the representative node vector N_ij for each of the representative nodes N11 to N33. Specifically, first, one data point is obtained from the data set 900 by random sampling (step S2).

[0065] Next, the representative node vector N_ij of the representative node is updated by performing the process shown in equation (1) below on the representative node closest to the acquired data (vector) (step S3). In equation (1), α is the learning rate. This learning rate α may be set based on empirical rules, etc., and may be a value such as 0.1 or 0.01. New representative node vector N_ij = (1-α) × old representative node vector N_ij + α × vector of acquired data (1)

[0066] Next, for the representative nodes located around the representative node whose representative node vector N_ij was updated in step S3 (hereinafter referred to as "nearby representative nodes"), the process shown in equation (2) below is executed by multiplying the change by equation (1) above by β, thereby updating the representative node vector (hereinafter referred to as "nearby representative node vector") N_ij of the surrounding representative nodes (step S4). In equation (2), β(r) may be an algorithm parameter determined based on empirical rules, and may be a value such as 0.1 or 0.01. Also, β(r) may be a fixed value or may be changed according to the number of training iterations. For example, the value of β(r) may be changed between the first training and the second training. The same may apply to α. New peripheral representative node vector N_ij = (1 - β(r)) × old peripheral representative node vector + β(r) × vector of acquired data (2)

[0067] In equation (2), r may be the distance from the node to be updated, and may be a variable used to weight β with a function related to distance. For example, β(r) = 0.9 for adjacent nodes and β(r) = 0.5 for nodes two steps away. As mentioned above, β(r) may be changed according to the number of training iterations; for example, the value of β may be decreased as the number of training iterations increases.

[0068] From this point onward, steps S2 to S4 are repeatedly executed until processing is completed for all data included in data set 900.

[0069] However, because FlowSOM uses the SOM algorithm, the following challenges exist.

[0070] Firstly, in the initialization of the SOM representative node (hereinafter also referred to as the SOM node), as described above, the initial node vector is determined by random sampling of data (step S1), which presents the challenge that the clustering results will differ each time it is executed.

[0071] Secondly, the SOM algorithm performs clustering while learning with random sampling (steps S2-S4), which means that the learning results differ depending on the sampling order, and consequently, the clustering results differ from one execution to the next.

[0072] Thirdly, as a consequence of the second issue, sequential learning is performed using random sampling (steps S2-S4), which means that updating the SOM node vectors is an indivisible process that cannot be parallelized (divided). This results in a problem where processing efficiency and processing speed cannot be improved.

[0073] Furthermore, while multispectral analysis of flow cytometers enables more detailed analysis, it also increases the amount of data handled, leading to challenges such as increased processing time and the curse of dimensionality when classifying data through clustering.

[0074] Furthermore, in recent years, improvements in the measurement speed of flow cytometers have led to an increase in the amount of data to be analyzed. However, as the amount of data increases, not only does the plotting time in data analysis increase, but when additional clusters generated by the user are analyzed, in addition to this increase in plotting time, the analysis time also increases simply because the number of target clusters increases.

[0075] Therefore, in this embodiment, an information processing device, information processing method, program, and information processing system that enable efficient execution of clustering capable of suppressing variations in results from one execution to the next will be described with examples.

[0076] However, the technology relating to this disclosure does not need to solve all of the above-mentioned problems simultaneously. Therefore, any technology that solves some or all of the above-mentioned problems by incorporating some or all of the configuration described later is considered to fall within the technical scope of this disclosure.

[0077] Furthermore, in the following description, components and operations similar to those of the first embodiment described above will be denoted by the same reference numerals, and their detailed descriptions will be omitted.

[0078] 2.1 Configuration of the Information Processing System Figure 12 is a block diagram showing the configuration of the information processing system according to this embodiment. As shown in Figure 12, the information processing system 200 comprises an information processing device 201 and a measuring device 3. In this embodiment, the measuring device 3 and the measurement sample may be the same as those described in the first embodiment, so a detailed explanation is omitted here.

[0079] The information processing device 201 includes a pre-processing / fluorescence separation unit 211, a pre-processing parameter table 12, a spectrum output unit 13, a clustering processing unit 214, a clustering result presentation unit 15, and a normal analysis presentation unit 17. In this configuration, the pre-processing parameter table 12, the clustering result presentation unit 15, and the normal analysis presentation unit 17 may be the same as those in the first embodiment. Furthermore, the pre-processing / fluorescence separation unit 211 may have the functions of both the pre-processing unit 11 and the fluorescence separation unit 16 in the first embodiment. Although the spectrum output unit 13 is omitted in Figure 12, a spectrum output unit 13 similar to that in the first embodiment may be added.

[0080] The pre-processing / fluorescence separation unit 211, similar to the fluorescence separation unit 16 in the first embodiment, uses a reference spectrum to separate the measurement data 2, which is spectral data, into fluorescence spectra for each fluorescent dye. The pre-processing / fluorescence separation unit 211 then performs pre-processing, such as Logicle conversion, on the fluorescence spectra after unmixing. The pre-processed fluorescence spectra for each fluorescent dye are normally presented to the user by the analysis presentation unit 17.

[0081] Furthermore, the parameter set of conversion parameters used in pre-processing may be selected by the user from the parameter sets managed in the pre-processing parameter table 12 (see Figure 3), similar to the first embodiment. In addition, the parameter sets managed in the pre-processing parameter table 12 may be fine-tuned by the user.

[0082] In this embodiment, the clustering processing unit 214 performs clustering on the pre-processed fluorescence spectrum output from the pre-processing / fluorescence separation unit 211. The clustering results generated by the clustering processing unit 214 are presented to the user by the clustering result presentation unit 15.

[0083] 2.2 Examples of Information Processing Device Operation Next, the operation of the information processing device 201 according to this embodiment will be described.

[0084] 2.2.1 Example of operation of the pre-processing / fluorescence separation unit First, an example of the operation of the pre-processing / fluorescence separation unit 211 according to this embodiment will be described. Figure 13 is a flowchart showing an example of the operation of the pre-processing / fluorescence separation unit according to this embodiment. As shown in Figure 13, in this embodiment, the pre-processing / fluorescence separation unit 211 first selects pre-processing parameters based on user instructions, similar to step S1 in Figure 8 (step S201).

[0085] Next, the pre-processing / fluorescence separation unit 211 performs fluorescence separation using a reference spectrum on the spectral data of all cells included in the measurement data 2, similar to the fluorescence separation process described with reference to Figure 10 in the first embodiment (step S202).

[0086] Next, the pre-processing / fluorescence separation unit 211 performs pre-processing on the spectral data of all cells included in the measurement data 2 using pre-processing parameters selected or modified by the user, similar to steps S2 to S5 in Figure 8 (steps S203 to S206).

[0087] Subsequently, once pre-processing of the spectral data of all cells is complete (YES in step S206), the pre-processing / fluorescence separation unit 211 inputs the pre-processed spectral data to the clustering processing unit 214 (step S208), and terminates this operation. The pre-processing / fluorescence separation unit 211 may also instruct the spectral output unit 13 to present the pre-processed spectral data to the user, for example, in step S6 of Figure 8.

[0088] In this operation, the spectral data of all cells included in measurement data 2 is targeted, but it is not limited to this; any spectral data of more than the number of nodes to be divided in the clustering process described later (hereinafter also referred to as the required number) will suffice. The same applies to the operation of the clustering processing unit described later.

[0089] 2.2.2 Example of clustering processing unit operation Next, an example of the operation of the clustering processing unit 214 according to this embodiment will be described. The basic flow of operations performed by the clustering processing unit 214 according to this embodiment may be the same as the operation described with reference to Figure 9 in the first embodiment. However, in this embodiment, the clustering process performed in step S12 of Figure 9 is replaced with the clustering process described later with reference to Figure 14.

[0090] Figure 14 is a flowchart showing an example of clustering processing performed by the clustering processing unit according to this embodiment. As can be seen from the operation of the clustering processing unit 214 shown in Figure 14, the clustering processing unit 214 according to this embodiment can also function as one or more of the dimensionality reduction unit, initial value determination unit, clustering unit, average value calculation unit, one or more allocation units, update unit, node number change unit, meta-clustering unit, division unit, and node setting unit in the claims.

[0091] As shown in Figure 14, in this embodiment, the clustering processing unit 214 first performs principal component analysis on the spectral data of all (or more than the required number) cells included in the measurement data 2 (step S221).

[0092] Furthermore, the spectral data targeted in step S221 may be spectral data that has undergone pre-processing after fluorescence separation. Also, although this explanation assumes two principal components, it is not limited to this and may be three or more. Moreover, in step S221, various dimensionality reduction methods using statistical data analysis techniques such as T-SNE may be performed, not limited to principal component analysis.

[0093] Next, the clustering processing unit 214 determines the initial value of the representative node vector of the SOM node based on the first and second principal components determined in step S221 and the values ​​(vectors) of all cells (or more than the required number of cells) included in the measurement data 2 (step S222). The details of step S222 will be explained later with reference to Figure 15.

[0094] Next, the clustering processing unit 214 performs SOM clustering by performing batch learning (step S223). The details of step S223 will be explained later with reference to Figure 16.

[0095] Next, the clustering processing unit 214 performs a process to determine the number of clusters (hereinafter referred to as meta-clustering), such as consensus clustering, on the result of the SOM clustering process performed in step S223 (step S224). This determines the number of clusters in the clustering process. Details of step S223 when using consensus clustering will be explained later with reference to Figure 17.

[0096] Subsequently, the clustering processing unit 214 instructs the clustering result presentation unit 15 to present the clustering results to the user (step S225). For visualization of the clustering results to be presented to the user, methods such as the Minimum Spanning Tree (MST) method can be used.

[0097] 2.2.3 Initialization of Representative Node Vectors Figure 15 is a diagram illustrating the operation when initializing the representative node vectors of each representative node according to this embodiment, as described in step S222 of Figure 14. In the example shown in Figure 15, similar to the example shown in Figure 11, 100 processed spectral data (vectors (coordinate values)) that have undergone fluorescence separation and preprocessing by the preprocessing / fluorescence separation unit 211 are classified into nine representative nodes N11 to N33 arranged in a 3x3 grid in a 2D coordinate system in their initial state.

[0098] As shown in Figure 15, in the initialization of the representative node vector according to this embodiment, the clustering processing unit 214 first calculates the average value of each dimension for a data group 302 (hereinafter simply referred to as the data group) that includes spectral data after fluorescence separation and pre-processing (hereinafter referred to as processed spectral data) (step S301). For example, if the number of dimensions of the data is 10, the values ​​of each dimension in all processed spectral data included in the data group 302 are added together for each of the 1st to 10th dimensions, and the average is calculated. The number of dimensions is a value corresponding to the number of data types, and for example, in the case of spectral data, it may correspond to the number of channels. Therefore, for example, if the photodetector array has 32 channels, that is, if the photodetector array divides the entire detection range into 32 wavelength bands to detect fluorescence, the number of dimensions of the spectral data acquired by this photodetector array will be 32.

[0099] Next, the clustering processing unit 214 performs dimensionality reduction on all processed spectral data included in the data group 302 to determine the first eigenvalue and the second eigenvalue, as well as the variance of the first eigenvalue and the variance of the second eigenvalue (step S302). For dimensionality reduction of the data, statistical data analysis methods such as principal component analysis can be used. The first and second eigenvalues ​​may be selected from the dimensions after dimensionality reduction according to a predetermined rule or arbitrarily, and their number is not limited to just the first and second eigenvalues, but may be one or three or more, for example.

[0100] Next, the clustering processing unit 214 calculates the initial node vectors for each representative node N11 to N33 using the following equation (3) (step S303). In equation (3), N_ij represents the initial value (initial node vector) of the representative node Nij at coordinates (i,j). In equation (3), the row division number is the number of representative nodes arranged in the row direction, which is 3 in this example. Similarly, the column division number is the number of representative nodes arranged in the column direction, which is 3 in this example. Initial node vector N_ij = mean value of each dimension + (first eigenvalue / number of row partitions) × i + (second eigenvalue / number of column partitions) × j (3)

[0101] By determining the initial node vector (initial value) for each representative node in the manner described above, it becomes possible to always set the same initial node vector for data sets 302 that belong to the same population. This makes it possible to avoid the clustering results differing from one execution to the next (corresponding to the first problem mentioned above).

[0102] 2.2.4 Clustering using batch learning Figure 16 is a flowchart showing an example of clustering by batch learning according to this embodiment, as described in step S223 of Figure 14. In the example shown in Figure 16, similar to the example shown in Figure 15, 100 processed spectral data (vectors (coordinate values)) that have undergone fluorescence separation and preprocessing by the preprocessing / fluorescence separation unit 211 are initially classified into nine representative nodes N11 to N33 arranged in a 3x3 grid in a 2D coordinate system.

[0103] As shown in Figure 16, in the batch learning clustering according to this embodiment, the clustering processing unit 214 first selects processed spectral data one by one from the data group 302 and assigns the selected processed spectral data to the representative node closest to it (step S304). This assignment process is repeated until all (or more than the required number) processed spectral data in the data group 302 have been assigned. In the example shown in Figure 16, processed spectral data with cell IDs '1' and '3' are assigned to representative node N11, processed spectral data with cell ID '2' is assigned to representative node N12, and processed spectral data with cell ID '100' is assigned to representative node N33 (explanation for cell IDs '4' to '99' is omitted).

[0104] The cell ID may be an identifier that uniquely identifies the cell corresponding to the processed spectral data registered in data group 302. Furthermore, the order in which processed spectral data is selected from data group 302 is not particularly limited and may be in various orders, such as in descending order of the cell IDs that uniquely identify the cells of the processed spectral data registered in data group 302, or in the order of registration to data group 302.

[0105] Next, the clustering processing unit 214 performs a process to update the representative node vectors for each representative node N11 to N33 based on the processed spectral data assigned to each representative node N11 to N33 (step S305). In updating the representative node vectors, for example, the updated new representative node vectors are calculated based on the following equation (4). New representative node vector N_ij = (1-α) × old representative node vector N_ij + α × average value of assigned processed spectral data (4)

[0106] Next, the clustering processing unit 214 updates the surrounding representative node vector N_ij of the surrounding representative nodes located around the representative node whose representative node vector N_ij was updated in step S305 by executing the process shown in equation (5) below, which is obtained by further multiplying the change by equation (4) above by β (step S306). In equation (5), β(r) may be the same as β(r) in equation (2) above. New peripheral representative node vector N_ij = (1 - β(r)) × Old peripheral representative node vector N_ij + β(r) × Average value of processed spectral data assigned to the representative node (5)

[0107] In this embodiment, the representative node vectors of each representative node N11 to N33 are updated by repeating the process described in steps S304 to S306 a predetermined number of times (for example, once or two or more times). Note that the learning rate α and / or β may change (for example, decrease or increase) with each repetition of the process in steps S304 to S306.

[0108] As described above, by assigning all of the processed spectral data to be analyzed to one of the representative nodes N11 to N33, and then updating the representative node vector of each representative node N11 to N33 using the processed spectral data assigned to each representative node N11 to N33, it becomes possible to eliminate random sampling during training, thus avoiding the clustering results differing from one execution to the next (corresponding to the second problem).

[0109] Furthermore, since the representative node vector is updated only after all processed spectral data has been allocated, it becomes possible to subdivide the process from allocating processed spectral data to updating the representative node vector and have it executed by different information processing devices.

[0110] For example, the processed spectral data included in the data group 302 can be divided into multiple groups, and the allocation of spectral data belonging to each group to a representative node can be performed by different information processing devices. For example, the clustering processing unit 214 according to this embodiment includes an allocation unit that allocates multiple processed spectral data to any of the representative nodes N11 to N33, and the number of allocation units in the clustering processing unit 214 may be equal to or less than the number of groups, and each allocation unit may be executed by a different information processing device.

[0111] This allows for the parallel execution of the allocation process for processed spectral data, thereby improving processing efficiency and speed.

[0112] Furthermore, by subdividing the processing, it becomes possible to reduce the amount of data handled by each information processing device, thereby suppressing the increase in processing time during clustering and the occurrence of problems such as the curse of dimensionality.

[0113] 2.2.5 Determining the number of clusters using consensus clustering Figure 17 is a diagram illustrating the operation of determining the number of clusters using consensus clustering, as described in step S224 of Figure 14. In the example shown in Figure 17, similar to the examples shown in Figures 15 and 16, 100 processed spectral data (vectors (coordinate values)) that have undergone fluorescence separation and preprocessing by the preprocessing / fluorescence separation unit 211 are initially classified into nine representative nodes N11 to N33 arranged in a 3x3 grid in a 2D coordinate system.

[0114] In the explanation above, we provided an example where the number of representative nodes for the processed spectral data included in data group 302 is initially set to nine, as shown in Figure 17(a). However, depending on the processed spectral data included in data group 302, the initially set number of representative nodes (nine) may not necessarily be optimal.

[0115] Therefore, in this embodiment, in step S224 of Figure 14, meta-clustering such as consensus clustering is performed to determine a more suitable number of divisions and a combination of representative nodes that constitute the representative nodes after meta-clustering (hereinafter referred to as meta-representative nodes), as shown in Figure 17(b).

[0116] The example shown in Figure 17 illustrates a case where, as a result of meta-clustering, one meta-representative node NN1 is formed from representative nodes N11, N12, and N21; one meta-representative node NN2 is formed from representative nodes N22, N31, and N32; and one meta-representative node NN3 is formed from representative nodes N13, N23, and N33.

[0117] In this embodiment, the number of node divisions set when clustering is performed (9 in this example) may be a default number of divisions, or it may be a number of divisions set by the user. If the user is to set the initial number of node divisions, the information processing device 201 may further include an operation input unit that functions as a node setting unit to allow the user to set the number of node divisions.

[0118] 2.3 Action and Effects As described above, according to this embodiment, the initial node vector (initial value) of each representative node is determined based on the mean value for each dimension of the processed spectral data and the eigenvalues ​​obtained by dimensionality reduction of the processed spectral data. Therefore, it is possible to always set the same initial node vector for data sets 302 with the same population. This makes it possible to avoid different clustering results each time it is run.

[0119] Furthermore, according to this embodiment, since all of the processed spectral data to be analyzed is assigned to one of the representative nodes N11 to N33, and then the representative node vectors of each representative node N11 to N33 are updated using the processed spectral data assigned to each representative node N11 to N33, it is possible to avoid different clustering results each time the process is executed.

[0120] Furthermore, in this embodiment, since the representative node vector is updated after all processed spectral data has been allocated, it is possible to subdivide the process from allocating processed spectral data to updating the representative node vector and have it executed by different information processing devices.

[0121] Furthermore, by subdividing the processing, it becomes possible to reduce the amount of data handled by each information processing device, thereby suppressing the increase in processing time during clustering and the occurrence of problems such as the curse of dimensionality.

[0122] In this embodiment, the example given is that spectral data after fluorescence separation is used as the target for clustering by the clustering processing unit 214. However, the embodiment is not limited to this, and for example, measurement data 2 acquired by the measuring device 3 can also be used as the target for clustering by the clustering processing unit 214. In that case, the pre-processing / fluorescence separation unit 211 performs pre-processing on the measurement data 2 and inputs the spectral data obtained before fluorescence separation to the clustering processing unit 214. The clustering processing unit 214 then performs the clustering process described above on the input spectral data before fluorescence separation.

[0123] Furthermore, in this embodiment, we have illustrated the case in which the pre-processing / fluorescence separation unit 211 performs pre-processing on the fluorescence spectra of each fluorescent dye obtained by performing fluorescence separation processing on the spectral data included in the measurement data 2. However, we are not limited to this, and for example, the pre-processing / fluorescence separation unit 211 may first perform pre-processing on the spectral data included in the measurement data 2 before performing fluorescence separation processing.

[0124] Other configurations, operations, and effects may be the same as those of the first embodiment described above, so a detailed explanation is omitted here.

[0125] 3. Hardware configuration of the information processing device Next, with reference to Figure 18, the hardware configuration of the information processing devices 1 and 201 according to the above-described embodiment will be explained. Figure 18 is a block diagram showing an example of the hardware configuration of the information processing device according to the embodiment of this disclosure. In the following explanation, the information processing device 1 is used as an example, but the same can be applied to the information processing device 201.

[0126] As shown in Figure 18, the information processing device 1 includes a CPU (Central Processing Unit) 901, a ROM (Read Only Memory) 903, and a RAM (Random Access Memory) 905. The information processing device 1 also includes a host bus 907, a bridge 909, an external bus 911, an interface 913, an input device 915, an output device 917, a storage device 919, a drive 921, a connection port 925, and a communication device 929. The information processing device 1 may have a processing circuit such as a DSP (Digital Signal Processor) or an ASIC (Application Specific Integrated Circuit) instead of, or together with, the CPU 901.

[0127] The CPU 901 functions as an arithmetic processing unit and control unit, and controls the overall operation or a part thereof within the information processing device 1 according to various programs recorded in the ROM 903, RAM 905, storage device 919, or removable recording medium 923. For example, the CPU 901 controls the overall operation of each functional unit included in the information processing device 1 in the above embodiment. The ROM 903 stores programs and arithmetic parameters used by the CPU 901. The RAM 905 temporarily stores programs used in the execution of the CPU 901 and parameters that change as appropriate during its execution. The CPU 901, ROM 903, and RAM 905 are interconnected by a host bus 907, which is composed of internal buses such as the CPU bus. Furthermore, the host bus 907 is connected to an external bus 911, such as a PCI (Peripheral Component Interconnect / Interface) bus, via a bridge 909.

[0128] The input device 915 is a device operated by the user, such as a mouse, keyboard, touch panel, buttons, switches, and levers. The input device 915 may be, for example, a remote control device using infrared or other radio waves, or an external connection device 927 such as a mobile phone that is compatible with the operation of the information processing device 1. The input device 915 includes an input control circuit that generates an input signal based on information entered by the user and outputs it to the CPU 901. The user inputs various data to the information processing device 1 or instructs it to perform processing operations by operating this input device 915.

[0129] The output device 917 is comprised of a device capable of notifying the user of acquired information visually or audibly. The output device 917 may be, for example, a display device such as an LCD, PDP, or OLED, an audio output device such as a speaker or headphones, or a printer. The output device 917 outputs the results obtained from the processing of the information processing device 1 as text or images, or as sound.

[0130] The storage device 919 is a data storage device configured as an example of the storage unit of the information processing device 1. The storage device 919 is composed of, for example, a magnetic storage device such as an HDD (Hard Disk Drive), a semiconductor storage device, an optical storage device, or a magneto-optical storage device. This storage device 919 stores programs executed by the CPU 901, various data, and various data acquired from external sources.

[0131] The drive 921 is a reader / writer for removable recording media 923 such as magnetic disks, optical disks, magneto-optical disks, or semiconductor memory, and is either built into or external to the information processing device 1. The drive 921 reads information recorded on the installed removable recording media 923 and outputs it to the RAM 905. The drive 921 also writes data to the installed removable recording media 923.

[0132] The connection port 925 is a port for directly connecting equipment to the information processing device 1. The connection port 925 may be, for example, a USB (Universal Serial Bus) port, an IEEE 1394 port, or a SCSI (Small Computer System Interface) port. Alternatively, the connection port 925 may be an RS-232C port, an optical audio terminal, or an HDMI (High-Definition Multimedia Interface) port. By connecting an external device 927 to the connection port 925, various types of data can be exchanged between the information processing device 1 and the external device 927.

[0133] The communication device 929 is a communication interface composed of, for example, a communication device for connecting to a communication network NW. The communication device 929 may be, for example, a communication card for wired or wireless LAN (Local Area Network), Bluetooth (registered trademark), or WUSB (Wireless USB). Alternatively, the communication device 929 may be a router for optical communication, a router for ADSL (Asymmetric Digital Subscriber Line), or a modem for various types of communication. The communication device 929 sends and receives signals, etc., to and from the Internet or other communication devices using a predetermined protocol such as TCP / IP. The communication network NW connected to the communication device 929 is a network connected by wire or wireless, and may be, for example, the Internet, a home LAN, infrared communication, radio wave communication, or satellite communication.

[0134] The technical scope of this disclosure is not limited to the embodiments described above, and various modifications are possible without departing from the spirit of this disclosure. Furthermore, components from different embodiments and modifications may be combined as appropriate.

[0135] For example, in the above embodiment, the information processing system 4 is configured to include an information processing device 1 or 201 and a measuring device 3, but this technology is not limited to such examples. For example, the information processing device 1 or 201 may have the functions (measurement functions) of the measuring device 3. In this case, the information processing system 4 is realized by the information processing device 1 or 201. Alternatively, the measuring device 3 may have the functions of the information processing device 1 or 201. In this case, the information processing system 4 is realized by the measuring device 3. Furthermore, the measuring device 3 may have some of the functions of the information processing device 1 or 201, and the information processing device 1 or 201 may have some of the functions of the measuring device 3.

[0136] Furthermore, this technology can also be configured as follows. (1) A dimensionality reduction unit performs dimensionality reduction on each of the multiple spectral data containing the fluorescent components emitted from each of the multiple particles, each labeled with one or more fluorescent dyes. An initial value determination unit that determines the initial value of each of the multiple nodes based on the result of the dimensionality reduction, A clustering unit that performs clustering on the plurality of spectral data using the aforementioned initial values, An information processing device equipped with the following features. (2) The system further includes an average value calculation unit that calculates the average value for each dimension of the plurality of spectral data, The initial value determination unit determines the initial value for each of the multiple nodes based on the dimensionality reduction result and the average value for each dimension of the multiple spectral data. The information processing device described in (1) above. (3) The clustering unit, An allocation unit that assigns each of the plurality of spectral data to one of the plurality of nodes, An update unit updates the node vector of each of the plurality of nodes based on the spectral data assigned to each of the plurality of nodes, The information processing apparatus according to (1) or (2) above, including the following: (4) The clustering unit, A node number changing unit changes the number of nodes by performing consensus clustering on the node vectors of each of the multiple nodes updated by the update unit, A meta-clustering unit performs meta-clustering on the nodes after the change by the node number changing unit, based on the node vectors of each of the multiple nodes before the change. The information processing apparatus described in (3) above, further comprising the above. (5) The clustering unit further comprises a division unit that divides the plurality of spectral data into two or more groups, The allocation unit assigns each of the multiple spectral data to one of the multiple nodes for each of the two or more groups. The information processing device described in (4) above. (6) The clustering unit comprises the same number of allocation units as the number of groups or less, Each of the aforementioned allocation units is arranged in a different information processing device. The information processing device described in (5) above. (7) The clustering unit performs clustering using the SOM (Self-Organizing Map) algorithm. An information processing device as described in any one of the above items (1) to (6). (8) The system further includes a node setting unit that allows the user to set the number of nodes. An information processing device as described in any one of the above items (1) to (7). (9) The dimensionality reduction unit performs principal component analysis of the plurality of spectral data as the dimensionality reduction. An information processing device as described in any one of the above items (1) to (8). (10) The spectral data mentioned above is spectral data measured by a spectral-type flow cytometer. An information processing device as described in any one of the above items (1) to (9). (11) The system further includes a preprocessing unit that performs a scale transformation on each of the aforementioned plurality of spectral data, The dimensionality reduction unit performs dimensionality reduction on each of the plurality of spectral data on which the scale transformation has been performed. An information processing device as described in any one of items (1) to (10) above. (12) The system further includes a fluorescence separation unit that separates each of the plurality of spectral data into fluorescence spectra for each of the fluorescent dyes, The preprocessing unit performs the scale transformation for each of the multiple fluorescence spectra. The information processing device described in (11) above. (13) The preprocessing unit performs a nonlinear transformation as the scale transformation. The information processing apparatus described in (11) or (12) above. (14) The pre-processing unit performs logice transformation, log transformation, or bi-exponential transformation as the scale transformation. The information processing device described in (13) above. (15) The clustering unit further comprises a display control unit that displays the results of the clustering performed by the clustering unit. An information processing device as described in any one of the above items (1) to (14). (16) The system further includes a display control unit that displays the results of the clustering performed by the clustering unit, The display control unit displays the data that has been scaled by the preprocessing unit. An information processing device as described in any one of items (11) to (14) above. (17) Dimensionality reduction is performed on each of the multiple spectral data containing the fluorescence components emitted from each of the multiple particles, each labeled with one or more fluorescent dyes. Based on the results of the dimensionality reduction, the initial values ​​for each of the multiple nodes are determined, Using the aforementioned initial values, clustering is performed on the plurality of spectral data. Information processing methods that include the following. (18) On the computer, A step of performing dimensionality reduction on each of the multiple spectral data containing the fluorescent components emitted from each of the multiple particles, each labeled with one or more fluorescent dyes, A step of determining the initial values ​​of each of the multiple nodes based on the results of the dimensionality reduction, A step of performing clustering on the plurality of spectral data using the initial values, A program to execute. (19) A measuring device for detecting multiple spectral data containing fluorescent components emitted from multiple particles, each labeled with one or more fluorescent dyes, An information processing device that clusters the plurality of spectral data detected by the measuring device, Equipped with, The aforementioned information processing device is A dimension reduction unit that performs dimension reduction on each of the plurality of spectral data detected by the measuring device, An initial value determination unit that determines the initial value of each of the multiple nodes based on the result of the dimensionality reduction, A clustering unit that performs clustering on the plurality of spectral data using the aforementioned initial values, An information processing system equipped with the following features. [Explanation of symbols]

[0137] 1,201 Information Processing Device 2. Measurement data 3. Measuring device 4,200 Information Processing Systems 11 Pre-processing 12 Pre-processing parameter table 13. Spectrum Output Section 14,214 Clustering Processing Unit 15. Clustering Result Presentation Section 16 Fluorescence separation section 17 Normal analysis presentation section 211 Pre-treatment / Fluorescence separation section 302 Datasets

Claims

1. A scale transformation is performed on each of the multiple fluorescence intensity data, which include the fluorescence components emitted from each of the multiple particles, each labeled with one or more fluorescent dyes. Dimensionality reduction is performed on each of the plurality of fluorescence intensity data from which the scale transformation has been performed. The average value for each dimension of the aforementioned plurality of fluorescence intensity data is calculated, Based on the eigenvalues ​​obtained as a result of the dimensionality reduction and the average value, the initial values ​​of each of the multiple nodes are determined. This includes performing clustering on the plurality of fluorescence intensity data using the aforementioned initial values. Information processing methods.

2. The aforementioned scaling transformation performs a nonlinear processing transformation. The information processing method according to claim 1.

3. The aforementioned scaling transformation involves performing a logicle transformation, a log transformation, or a bi-exponential transformation. The information processing method according to claim 2.

4. The plurality of fluorescence intensity data are plurality of spectral data containing the fluorescence component emitted from each of the plurality of particles. The information processing method according to claim 1.

5. The plurality of fluorescence intensity data are data separated into fluorescence spectra for each of the fluorescence components from each of the plurality of spectral data emitted from each of the plurality of particles. The information processing method according to claim 1.

6. The aforementioned clustering is, Each of the aforementioned plurality of fluorescence intensity data is assigned to one of the plurality of nodes. This is performed by updating the node vector of each of the multiple nodes based on the fluorescence intensity data assigned to each of the multiple nodes. The information processing method according to claim 1.

7. The aforementioned clustering is, By performing consensus clustering on the node vector of each of the updated nodes, the number of nodes is changed. This is performed by performing meta-clustering on the modified node based on the node vectors of each of the multiple nodes before the modification. The information processing method according to claim 6.

8. The aforementioned clustering is, The plurality of fluorescence intensity data are divided into two or more groups, For each of the two or more groups, the process is performed by assigning each of the multiple fluorescence intensity data to one of the multiple nodes. The information processing method according to claim 7.

9. The clustering is performed by an allocation unit having the same number as or less than the number of groups, where each allocation unit is located on a different information processing device. The information processing method according to claim 8.

10. The aforementioned clustering is performed using the SOM (Self-Organizing Map) algorithm. The information processing method according to claim 1.

11. The number of nodes is set by the user. The information processing method according to claim 1.

12. The aforementioned dimensionality reduction is the principal component analysis of the plurality of fluorescence intensity data. The information processing method according to claim 1.

13. The aforementioned spectral data is spectral data measured by a spectral-type flow cytometer. The information processing method according to claim 4.

14. The results of the clustering are shown below. The information processing method according to claim 1.

15. On the computer, A step of performing a scale transformation on each of the multiple fluorescence intensity data, each containing the fluorescence component emitted from each of the multiple particles, each labeled with one or more fluorescent dyes, A step of performing dimensionality reduction on each of the plurality of fluorescence intensity data on which the scale conversion has been performed, A step of calculating the average value for each dimension of the plurality of fluorescence intensity data, A step of determining the initial value of each of the multiple nodes based on the eigenvalues ​​obtained as a result of the dimensionality reduction and the average value, A step of performing clustering on the plurality of fluorescence intensity data using the aforementioned initial values, A program to execute.

16. A measuring device for detecting multiple fluorescence intensity data, each containing fluorescence components emitted from multiple particles, each labeled with one or more fluorescent dyes, An information processing device that clusters the plurality of fluorescence intensity data detected by the measuring device, Equipped with, The aforementioned information processing device is A preprocessing unit that performs scaling on each of the plurality of fluorescence intensity data detected by the measuring device, A dimensionality reduction unit that performs dimensionality reduction on each of the plurality of fluorescence intensity data from which the scale conversion has been performed, An average value calculation unit that calculates the average value for each dimension of the plurality of fluorescence intensity data, An initial value determination unit that determines the initial value of each of a plurality of nodes based on the eigenvalues ​​obtained as a result of the dimensionality reduction and the average value, A clustering unit that performs clustering on the plurality of fluorescence intensity data using the aforementioned initial values, An information processing system equipped with the following features.

Images