Estimation device, estimation method, and estimation program

Abstract

To improve the accuracy of cell type estimation. [Solution] An estimation device having a processor that executes a program and a storage device that stores the program performs an acquisition process to obtain the first expression levels of a plurality of genes in a single cell of a first sample and the second expression levels of the plurality of genes in a cell tumor of a second sample; a calculation process to calculate the similarity between the first expression levels and the second expression levels; an identification process to identify the single cell cell tumor based on the calculation results from the calculation process; and an output process to output the identification results from the identification process.

Description

Technical Field

[0001] The present invention relates to an estimation device, an estimation method, and an estimation program.

Background Art

[0002] It has been found that it is difficult to eradicate cancer only with cancer cell-targeted therapeutic drugs (molecular target drugs), and new therapeutic drugs (immune checkpoint inhibitors: ICI) that suppress the growth of cancer cells by activating the immune mechanism have been developed. However, although ICI is expensive, the reactivity varies among individual patients, and the therapeutic effect cannot be confirmed without administration. Therefore, a technology that can select an appropriate ICI for each patient is required.

[0003] In addition, by analyzing the gene expression of single cells, the behavior of individual cells in the patient's tumor environment has become clear. If the types and states of immune cells responding to tumor cells can be grasped, application to ICI selection can be expected.

[0004] Patent Document 1 discloses a correlation analysis system that automates the identification of cell populations and the correlation analysis between cell profiling results and clinical information. This correlation analysis system is a cell population identification system, and the cell population identification system includes a cell population identification system that outputs the ratio of cells from measurement data of a plurality of cells of a first subject, a medical data receiving means for receiving medical data of the first subject, a generation means for generating composite data by matching the medical data and the ratio of the cells, and an analysis means for analyzing the composite data.

[0005] Patent Document 2 discloses a system and method for associating compounds with physiological states. In this system and method, a fingerprint of the compound's chemical structure is obtained and input into a model that outputs one or more calculated activation scores. Each activation score represents a cellular component module in a set of modules, each module containing a subset of cellular components, and the first module in the set of modules is associated with a physiological state. If the activation score for the first module meets a threshold criterion, the compound is identified as being associated with a physiological state. In some embodiments, each activation score represents a perturbation signature associated with a physiological state, and the compound is identified if the activation score for the first perturbation signature meets a threshold criterion. A system and method for training a model that associates compounds with physiological states is also provided.

[0006] Patent Document 3 discloses an apparatus for predicting the cellular composition of a tissue image based on spatial gene expression level information. The apparatus includes a communication module for receiving tissue images of a subject to be examined; a memory storing a program for predicting cellular composition information from tissue images; and a processor for executing the program. The program predicts cellular composition information by inputting the tissue image into a cellular composition prediction model that has been trained based on training data consisting of spatial transcription information and tissue images spatially aligned therewith. The spatial transcription information includes transcription data containing spatial information and tissue image data sharing spatial information, where the spatial information means positional information for a plurality of spots arranged on a two-dimensional plane of the tissue image data, and includes the coordinates of each spot.

[0007] Patent Document 4 discloses a cell processing system that improves cell culture efficiency and allows for obtaining desired cells without incurring costs. This cell processing system comprises a cell culture unit capable of culturing a group of cells, a measuring unit capable of measuring the group of cells, and an estimation unit. The estimation unit estimates, based on information about the cells of the group of cells before processing, information about a predetermined cell to be reached after processing the group of cells, and at least one of predetermined processing conditions for the group of cells before processing, processing conditions that can lead to the predetermined cell information, or information about the cells of the group of cells after processing that are derived by the predetermined processing conditions. [Prior art documents] [Patent Documents]

[0008] [Patent Document 1] International Publication No. 2023 / 033056 [Patent Document 2] International Publication No. 2022 / 266259 [Patent Document 3] International Publication No. 2022 / 220385 [Patent Document 4] International Publication No. 2022 / 181049 [Overview of the project] [Problems that the invention aims to solve]

[0009] However, single-cell gene expression analysis is not very explainable to medical professionals such as doctors because the cell type estimation depends on the characteristics of each individual cell.

[0010] The technology described in Patent Document 1 does not transfer a model trained on bulk RNA-seq data to single-cell data. The technology described in Patent Document 2 is limited to integrating single-cell data, CITE-seq data, scATAC-seq data, etc. The technology described in Patent Document 3 is limited to integrating spatial information such as smear specimens with gene transcription information. The technology described in Patent Document 4 does not estimate cell type and does not handle gene expression data.

[0011] The present invention aims to improve the accuracy of cell type estimation. [Means for solving the problem]

[0012] An estimation device representing one aspect of the invention disclosed in this application is an estimation device having a processor that executes a program and a storage device that stores the program, wherein the processor performs an acquisition process to obtain the first expression levels of a plurality of genes in a single cell of a first sample and the second expression levels of the plurality of genes in a cell tumor of a second sample; a calculation process to calculate the similarity between the first expression levels and the second expression levels; a identification process to identify the cell tumor of the single cell based on the calculation results of the calculation process; and an output process to output the identification results of the identification process. [Effects of the Invention]

[0013] According to a typical embodiment of the present invention, the accuracy of cell type estimation can be improved. Other issues, configurations, and effects not mentioned above will be clarified by the following description of the examples. [Brief explanation of the drawing]

[0014] [Figure 1] Figure 1 is a block diagram showing an example of the hardware configuration of the estimation device. [Figure 2] Figure 2 is a flowchart showing an example of the cell tumor estimation procedure using the estimation device. [Figure 3] Figure 3 is an explanatory diagram showing an example of gene expression level information for a sample group. [Figure 4] Figure 4 is an explanatory diagram showing an example of a training dataset. [Figure 5] Figure 5 is an explanatory diagram showing an example of how importance is calculated. [Figure 6] Figure 6 is an explanatory diagram showing an example of gene selection. [Figure 7] Figure 7 is an explanatory diagram showing an example of single-cell gene expression level information for the estimated target sample. [Figure 8]FIG. 8 is an explanatory diagram showing an example of calculating the similarity regarding gene expression between a single cell of a specimen to be estimated and each of the cell tumors of a group of specimens. [Figure 9] FIG. 9 is an explanatory diagram showing a specific example of a cell tumor of a single cell of a specimen to be estimated.

Mode for Carrying Out the Invention

[0015] <FIG. 1 Hardware Configuration Example of the Estimation Device> FIG. 1 is a block diagram showing a hardware configuration example of the estimation device. The estimation device 100 includes a processor 101, a storage device 102, an input device 103, an output device 104, and a communication interface (communication IF) 105. The processor 101, the storage device 102, the input device 103, the output device 104, and the communication IF 105 are connected by a bus 106. The processor 101 controls the estimation device 100. The storage device 102 serves as a working area for the processor​​​​​​​​​​​The estimation device 100 acquires gene expression level information for multiple cell tumors in each sample of the sample group. The sample group is a collection of samples. Each sample is taken from a different subject and contains cells from multiple cell tumors. A cell tumor is a type of cell. Specifically, for example, a cell tumor is a cluster of cells classified by unsupervised learning based on the gene expression of single cells, with cells having similar expression profiles, and is a diverse collection of cells. Gene expression level information is information about the level of gene expression.

[0018] [Figure 3: Gene expression level information for each sample group] Figure 3 is an explanatory diagram showing an example of gene expression level information for a sample group. Samples T1 to TZ (where Z is an integer greater than or equal to 1) are samples taken from subjects H1 to HZ. The gene expression level information 300(T1) for sample T1 to 303(TZ) for sample TZ each contains expression level data 303(T1) to 303(TZ) for each gene G1 to GN (where N is an integer greater than or equal to 1) in the cell tumor 301 identified by the combination of cell tumor 301 and gene 202.

[0019] The following explanation uses gene expression level information 300(Tx) from any sample Tx (where x is an integer satisfying 1 ≤ x ≤ Z) within the sample groups T1 to TZ (which may be referred to as sample group T) as an example. The (T1) to (TZ) appended to the end of the code are used to identify which sample the data originates from.

[0020] Gene 202 is a set of identification information G1-GN (where N is an integer greater than or equal to 2) that uniquely identifies each of multiple genes. Hereafter, these will be referred to as genes G1-GN.

[0021] Cell tumor 301 defines m cell tumors CT1(Tx) to CTm(Tx) in sample Tx (where m is an integer greater than or equal to 1). Cell tumor CTj(Tx) represents the j-th cell tumor 301 in sample Tx (where j is an integer satisfying 1 ≤ j ≤ m). Note that cell tumors CTj(T1), ..., CTj(Tx), ..., CTj(TZ) are the same cell tumor CTj, although they are from different samples. Cell tumor CTj(Tx) has expression levels Ea1j(Tx) to EaNj(Tx) for each gene G1 to GN. The set of expression levels Ea1j(Tx) to EaNj(Tx) for each gene G1 to GN for each cell tumor CTj(Tx) is the expression level data 303(Tx) for sample Tx.

[0022] (Step S202) Returning to Figure 2, the estimation device 100 determines whether or not an estimation model for the cell tumor exists. Specifically, for example, the estimation device 100 determines whether or not an estimation model for the cell tumor is stored in the memory device 102. An estimation model for the cell tumor is a model that estimates a cell tumor when the expression levels of multiple genes are input. If an estimation model for the cell tumor exists (step S202: Yes), the process proceeds to step S203; otherwise, it proceeds to step S204.

[0023] (Step S203) The estimation device 100 obtains an estimated model of the cell tumor and proceeds to step S205. Specifically, for example, the estimation device 100 reads the estimated model of the cell tumor from the memory device 102.

[0024] (Step S204) The estimation device 100 generates an estimation model for the cell tumor and proceeds to step S205. The estimation model for the cell tumor is a machine learning model trained with the gene expression levels of the cell tumor as explanatory variables (training data) and the cell tumor as the target variable (ground truth label). The estimation model for the cell tumor may be, for example, a random forest or a neural network.

[0025] [Figure 4 Training Dataset] Figure 4 is an explanatory diagram showing an example of a training dataset. The cell tumor estimation model 403 is generated, for example, by supervised learning. Its training dataset 400 is a combination of the training data 401 and the ground truth labels 402.

[0026] The training data 401 consists of gene expression level information 300(T1) for sample T1 to gene expression level information 300(TZ) for sample TZ. More specifically, the training data 401 consists of expression level data 303(T1) within the gene expression level information 300(T1) for sample T1 to expression level data 303(TZ) within the gene expression level information 300(TZ) for sample TZ. The correct label 402 is cell tumor 301, i.e., cell tumors CT1 to CTm.

[0027] (Step S205) The estimation device 100 performs gene filtering in multiple cell tumors of the sample group. Specifically, for example, the estimation device 100 calculates the importance of features using the cell tumor estimation model 403. In this example, the features are genes G1 to GN. The importance is an index value that indicates how much each of genes G1 to GN contributed to the estimation by the cell tumor estimation model 403, and a higher value indicates a greater contribution to the prediction.

[0028] The estimation device 100 may selectively execute step S205. Specifically, for example, based on instructions from the user, the estimation device 100 may choose not to execute step S205.

[0029] [Figure 5: Calculation of Importance] Figure 5 is an explanatory diagram illustrating an example of importance calculation. The estimation device 100 records how much the Gini impurity (or entropy) decreases for each feature (gene expression level) used at each node constituting each decision tree in the group of decision trees that make up the random forest, and sums up the amounts of this decrease within the random forest. This sum is the importance of that gene. Furthermore, if the cell tumor estimation model is explainable AI, the estimation device 100 calculates the importance of the feature gene.

[0030] For example, if the gene expression level information for the cell tumor in sample Tx is 300(Tx), the estimation device 100 calculates the importance W(Tx) as the importance of gene G1 w1(Tx) to the importance of gene GN wN(Tx), using the cell tumor estimation model 403.

[0031] [Figure 6: Gene filtering] Figure 6 is an explanatory diagram illustrating an example of gene filtering. The estimation device 100 filters genes G1 to GN to those whose importance W(Tx) is above the threshold. In other words, the estimation device 100 removes the expression levels of genes whose importance W(Tx) is below the threshold from the gene expression level information 300(Tx) of the cell tumor of the sample Tx.

[0032] The gene expression level information 600(Tx) for the cell tumor of sample Tx, obtained by deleting the expression levels of genes with importance W(Tx) below the threshold from the cell tumor gene expression level information 300(Tx) of sample Tx, has expression level data 603(Tx). Cell tumor CTj(Tx) has expression levels Ea1j(Tx) to Eanj(Tx) for each gene G1 to Gn (n is an integer less than or equal to N) after deletion. The set of expression levels Ea1j(Tx) to Eanj(Tx) for each gene G1 to Gn for each cell tumor CTj(Tx) is the expression level data 603(Tx) of sample Tx.

[0033] Note that gene G1 is also deleted if its importance is below the threshold, but here, for convenience, the deleted genes are designated G1 to Gn to show that the number of genes has been reduced from N to n. In this way, the gene expression level information 300(Tx) of the cell tumor of sample Tx having expression level data 303(Tx) is changed to the gene expression level information 600(Tx) of the cell tumor of sample Tx having expression level data 603(Tx).

[0034] (Step S206) The estimation device 100 acquires gene expression level information of single cells of the sample to be estimated. Specifically, for example, the estimation device 100 acquires gene expression level information of single cells of the sample to be estimated by input from the input device 103, reception from the communication ID 105, or reading from the storage device 102.

[0035] [Figure 7: Gene expression level information of single cells in the estimated target sample] Figure 7 is an explanatory diagram showing an example of single-cell gene expression level information for a suspected sample. The suspected sample T0 is a sample targeted for cytoma estimation and is a single-cell sample taken from a subject H0.

[0036] The gene expression level information 700 of the single cell of the target sample T0 for estimation includes gene expression level data 703 for each gene G1 to GN in the single cell 701, which is identified by the combination of the single cell 701 and gene 202. The estimation device 100 may also calculate the importance of each gene G1 to GN, Wb=wb1,...,wbN, using the cell tumor estimation model 403.

[0037] (Step S207) Returning to Figure 2, the estimation device 100 performs the filtering of genes G1 to GN in single cells of the target sample T0. Specifically, for example in Figure 7, the estimation device 100 removes the gene expression levels of genes whose importance, calculated in step S205, is below the threshold from the single cell gene expression level information 700 of the target sample T0.

[0038] This provides single-cell gene expression level information 710 of the estimated target sample T0 after deletion. The single-cell gene expression level information 710 of the estimated target sample T0 after deletion contains gene expression level data 713 for each gene G1 to Gn in the single cell 701, which is identified by the combination of the single cell 701 and gene 202. Furthermore, the importance levels Wb=wb1,...,wbN are narrowed down to Wb=wb1,...,wbi,...,wbn.

[0039] The estimation device 100 may selectively execute step S207. Specifically, for example, if the estimation device 100 selects not to execute step S205 based on instructions from the user, it will select not to execute step S207.

[0040] (Step S208) The estimation device 100 calculates the similarity in gene expression between each of the single cells 701 of the target sample T0 and the cell tumors 301 of sample group T. The similarity can be expressed as, for example, cosine similarity or Euclidean distance. Here, we will explain using cosine similarity as an example.

[0041] [Figure 8 Similarity Calculation] Figure 8 is an explanatory diagram illustrating an example of calculating the similarity of gene expression between 701 single cells of the estimated target sample T0 and 301 cell tumors of sample group T. The expression levels of genes G1 to Gn in cell tumor CTj(Tx), the j-th cell tumor (where j is an integer satisfying 1 ≤ j ≤ m) among m cell tumors CT1(Tx) to CTm(Tx) of a given sample Tx, can be expressed in vector form as [Ea1j(Tx),…,Eanj(Tx)] (see Figure 6).

[0042] In the h-th single cell SCh (where i is an integer satisfying 1 ≤ h ≤ k) among k single cells SC1 ≤ SCk, the expression levels of genes G1 ≤ Gn are expressed in vector form as [Eb1h ≤ Ebnh] (see Figure 7).

[0043] The estimation device 100 calculates the cosine similarity SMhj(Tx) between the expression vector of cell tumor CTj(Tx) [Ea1j(Tx),…,Eanj(Tx)] and the expression vector of single cell SCh [Eb1h,…,Ebnh] for each combination of cell tumor CTj(TX) and single cell SCh.

[0044] This yields the similarity data 800(Tx) for sample Tx. The estimation device 100 performs this calculation for each of the samples T1 to TZ, thereby obtaining the similarity data 800(T1) for sample T1 to the similarity data 800(TZ) for sample TZ.

[0045] The estimation device 100 may, prior to calculating the similarity, weight the expression vector of cell tumor CTj(Tx) [Ea1j(Tx),…,Eanj(Tx)] with an importance vector [Wa1(Tx),…,Wan(Tx)] indicating its importance, to obtain the weighted expression vector of cell tumor CTj(Tx) [Wa1(Tx)×Ea1j(Tx),…,Wan(Tx)×Eanj(Tx)].

[0046] Similarly, the estimation device 100 may, in calculating similarity, weight the single-cell SCh expression vector [Eb1h,…,Ebnh] by an importance vector [Wb1,…,Wbn] indicating its importance, to obtain the weighted single-cell SCh expression vector [Wb1×Eb1h,…,Wbn×Ebnh].

[0047] As a result, the estimation device 100 calculates the cosine similarity SMhj(Tx) between the weighted expression vector of cell tumor CTj(Tx) [Wa1(Tx)×Ea1j(Tx),…,Wan(Tx)×Eanj(Tx)] and the weighted expression vector of single cell SCh [Wb1×Eb1h,…,Wbn×Ebnh] for each combination of cell tumor CTj(TX) and single cell SCh, and outputs similarity data 800. This allows the similarity to be increased for genes that are important in generating the estimated cell tumor model 403.

[0048] (Step S209) Returning to Figure 2, the estimation device 100 identifies the cell tumor 301 of the single cell 701 in the target sample T0. Specifically, for example, the estimation device 100 sorts the similarity SMhj(Tx) in descending order of similarity SMhj(Tx) and identifies the cell tumor 301 in the top group. The cell tumor 301 in the top group may be, for example, a cell tumor whose similarity SMhj(Tx) is above a threshold, or a cell tumor 301 whose similarity SMhj(Tx) is above a predetermined percentile. The estimation device 100 identifies the cell tumor 301 of the single cell 701 in the target sample t0 by majority vote of the occurrence count of the cell tumor 301 in the top group.

[0049] [Figure 9: Identification of 301 single-cell tumors from 701 single cells in a suspected target sample T0] Figure 9 is an explanatory diagram showing an example of the identification of cell tumor 301 from single cell 701 in the estimated target sample T0. The similarity list 900 is information obtained by sorting the similarity SMhj(Tx) of the similarity data 800 in descending order. The similarity list 900 has similarity 901, single cell 701, and cell tumor 301. Step S902 identifies cell tumor 301 in the top group 902. Cell tumor 301 in the top group 902 is designated as specific cell tumor 903. In the example in Figure 9, specific cell tumors 903 are cell tumors CT1 to CT9.

[0050] The estimation device 100 counts the number of occurrences 904 of specific cell tumors 903 in the top group 902. In the counting result 910, the number of occurrences 904 for cell tumor CT3 is "5", which is the highest. Therefore, the estimation device 100 identifies cell tumor 301 of single cell 701 in the target sample T0 as CT3. In this way, the estimation device 100 estimates cell tumor 301 of single cell 701 by majority vote within the top group 902.

[0051] The estimation device 100 may also identify the cell tumor 301 with the highest similarity score of 901 as the single-cell cell tumor 301.

[0052] (Step S210) The estimation device 100 outputs the processing result of step S209. The processing result may include, for example, the counting result 910. The processing result may also include the top group 902. The processing result may also include the similarity list 900. The processing result may also include the similarity data 800.

[0053] The estimation device 100 may store the processing results in the storage device 102, display them on a display device which is an example of an output device 104, print them out on a printer which is an example of an output device 104, or transmit them via the communication IF 105 to another computer which is connected to the estimation device 100 via a network. This completes the series of processes.

[0054] Thus, this embodiment makes it possible to improve the accuracy of cell type estimation. Therefore, it is possible to improve explainability to medical professionals such as physicians, who are the main users. In addition, by utilizing multiple cell tumors (bulk), the characteristics of single cells can be eliminated, and the extraction of characteristics as a cell type is made easier. Furthermore, by narrowing down the importance of genes G1-GN to genes G1-Gn, it is not necessary to consider the scale difference of TPM when calculating similarity.

[0055] It should be noted that the present invention is not limited to the embodiments described above, but includes various modifications and equivalent configurations within the spirit of the attached claims. For example, the embodiments described above are described in detail to make the present invention easier to understand, and the present invention is not necessarily limited to having all of the described configurations. Furthermore, some of the configurations of one embodiment may be replaced with those of another embodiment. Furthermore, some of the configurations of one embodiment may be added to those of another embodiment. Furthermore, some of the configurations of each embodiment may be added, deleted, or replaced with other configurations.

[0056] Furthermore, each of the aforementioned configurations, functions, processing units, and processing means may be implemented in hardware, for example, by designing them as integrated circuits, or they may be implemented in software by having a processor interpret and execute programs that realize each function.

[0057] Information such as programs, tables, and files that implement each function can be stored in memory, hard disks, SSDs (Solid State Drives), or on recording media such as IC (Integrated Circuit) cards, SD cards, and DVDs (Digital Versatile Discs).

[0058] Furthermore, the control lines and information lines shown are those deemed necessary for explanation purposes and do not necessarily represent all control lines and information lines required for implementation. In reality, it can be assumed that almost all components are interconnected. [Explanation of Symbols]

[0059] 100 Estimator 101 Processors 102 Storage Devices 202 genes Gene expression level information for 300(Tx) cell tumors 301 Cell tumor Expression levels data for 303(Tx) cell tumors 400 training datasets 401 Training Data 402 Correct Label 403 Estimated Model Gene expression level information for cell tumors after filtering by 600(Tx). Expression levels of cell tumors after filtering for 603(Tx) 700 Single-cell gene expression level information 701 Single cell 703 Single-cell expression data 710 Single-cell gene expression level information after filtering 713 Single-cell expression data after filtering 800 Similarity Data 900 Similarity List 901 Similarity 902 Top Group 903 Specified cell tumor 904 appearances 910 Counting Results

Claims

1. An estimation device having a processor for executing a program and a storage device for storing the program, The aforementioned processor, An acquisition process to obtain the first expression levels of multiple genes in a single cell of the first sample and the second expression levels of the same multiple genes in a cell tumor of the second sample, A calculation process for calculating the similarity between the first expression level and the second expression level, Based on the calculation results from the calculation process described above, a specific process is performed to identify the single-cell cell tumor, Output processing to output the specific result obtained by the aforementioned specific processing, An estimation device characterized by performing the following actions.

2. An estimation device according to claim 1, In the acquisition process described above, the processor acquires the second expression level for each of the multiple cell tumors in the second sample. In the calculation process described above, the processor calculates the similarity for each cell tumor, In the aforementioned specific processing, the processor identifies the single-cell cell tumor based on the similarity calculated for each cell tumor. An estimation device characterized by the following features.

3. The estimation device according to claim 2, In the acquisition process described above, the processor acquires the second expression level for each of the multiple second samples, for each cell tumor. In the calculation process described above, the processor calculates the similarity for each of the plurality of second samples for each of the cell tumors, In the aforementioned specific processing, the processor identifies the single-cell cell tumor based on the similarity calculated for each of the plurality of second samples. An estimation device characterized by the following features.

4. The estimation device according to claim 3, In the aforementioned specific processing, the processor identifies the single-cell cell tumor within a specific group of cell tumors with high similarity, based on the number of cells that exhibit the aforementioned similarity. An estimation device characterized by the following features.

5. The estimation device according to claim 4, In the aforementioned specific processing, the processor identifies the single-cell cell tumor by majority vote of the number of similarities that exist. An estimation device characterized by the following features.

6. The estimation device according to claim 3, The aforementioned processor, A generation process that generates an estimation model which estimates the cell tumor and calculates the importance of each gene indicating which genes are important for the estimation result of the cell tumor by supervised learning using a learning dataset which uses the expression levels of the multiple genes for each cell tumor of the second sample as training data and the cell tumor as the ground truth label, Based on the importance of each gene, a filtering process is performed to narrow down the first and second expression levels. In the calculation process described above, the processor calculates the similarity based on the first expression level after filtering by the filtering process and the second expression level after filtering by the filtering process. An estimation device characterized by the following features.

7. An estimation device according to claim 6, In the calculation process described above, the processor calculates the similarity based on the importance and first expression level after the filtering process, and the importance and second expression level after the filtering process. An estimation device characterized by the following features.

8. An estimation method is performed by an estimation device having a processor that executes a program and a storage device that stores the program, The aforementioned processor, An acquisition process to obtain the first expression levels of multiple genes in a single cell of the first sample and the second expression levels of the same multiple genes in a cell tumor of the second sample, A calculation process for calculating the similarity between the first expression level and the second expression level, Based on the calculation results from the calculation process described above, a specific process is performed to identify the single-cell cell tumor, Output processing to output the specific result obtained by the aforementioned specific processing, An estimation method characterized by performing the following.

9. In the processor, An acquisition process to obtain the first expression levels of multiple genes in a single cell of the first sample and the second expression levels of the same multiple genes in a cell tumor of the second sample, A calculation process for calculating the similarity between the first expression level and the second expression level, Based on the calculation results from the calculation process described above, a specific process is performed to identify the single-cell cell tumor, Output processing to output the specific result obtained by the aforementioned specific processing, An estimation program characterized by causing the execution of a specific action.

Images