Drug efficacy estimation method, drug efficacy estimation model generation method, drug efficacy estimation system, control program, recording medium, and trained model

Abstract

To precisely estimate the treatment effect of a drug by using a medical image and information of genetic expression.SOLUTION: The method for estimating a drug efficiency includes: an acquisition step of acquiring a first medical image of a lesion area of a first patient and first expression information regarding a genetic expression in the lesion area of the first patient; and an estimation step of inputting the first medical image and the first expression information into a trained model trained by using second drug information of a second drug input into a second patient and second effect information showing the effect of the treatment with the second drug as purpose variables, by using a second medical image and second expression information of the second patient as explanatory variables, and outputting a first drug which can be input into the first patient and the estimated effect of the treatment.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a drug efficacy estimation method, a drug efficacy estimation model generation method, a drug efficacy estimation system, a control program, a recording medium, and a trained model. [Background technology]

[0002] In the medical field, radiomics technology has attracted attention, which uses a large number of feature quantities extracted from medical images such as CT (Computed Tomography) images to estimate the presence or absence of disease, treatment prognosis, etc. For example, Non-Patent Document 1 discloses that radiomics technology is used to estimate the genetic nature of cancer from medical images and predict the effectiveness of radiation therapy for that cancer. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Ryoichi Uchiyama, Molecular Classification and Treatment Strategy by Radiomics, Medical Physics, Vol. 40, No. 1, pp. 19-22, 2022 Summary of the Invention [Problem to be solved by the invention]

[0004] As shown in Non-Patent Document 1, the relationship between features contained in medical images and gene mutations has been discussed. However, no research has been conducted to date on the correlation between gene expression patterns, which are thought to be more directly related to the biological phenotype of the lesion site than gene mutations, and features contained in medical images. Therefore, no research has been conducted on estimating the therapeutic effect of drugs by correlating the features of medical images with both the gene expression patterns of the lesion site and data on the therapeutic effect of drugs.

[0005] An object of one aspect of the present invention is to provide a drug efficacy estimation method and the like that can estimate the therapeutic effect of a drug with high accuracy using medical images and gene expression information. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems, a drug efficacy estimation method according to one embodiment of the present invention includes an acquisition step of acquiring a first medical image of a lesion site related to a target disease of a first patient and first expression information regarding gene expression at the lesion site corresponding to the first medical image; and an estimation step of inputting the first medical image and the first expression information into a trained model trained using training data including, as explanatory variables, a second medical image of the lesion site of a second patient and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and including, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating the therapeutic effect of the second drug on the target disease, and causing the trained model to output first drug information indicating a first drug that can be administered to the first patient and first effect information indicating the estimated therapeutic effect of the first drug on the target disease.

[0007] In order to solve the above-mentioned problems, a method for generating a drug efficacy estimation model according to one embodiment of the present invention generates a trained model by machine learning using training data that includes, as explanatory variables, a second medical image of a lesion site related to a target disease in a second patient and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and includes, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating the therapeutic effect of the second drug on the target disease.

[0008] In order to solve the above-mentioned problems, a drug efficacy estimation system according to one embodiment of the present invention includes: an acquisition unit that acquires a first medical image of a lesion site related to a target disease of a first patient, and first expression information regarding gene expression at the lesion site corresponding to the first medical image; and an estimation unit that inputs the first medical image and the first expression information into a trained model trained using training data that includes, as explanatory variables, a second medical image of the lesion site of a second patient, and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and includes, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating the therapeutic effect of the second drug on the target disease; and causes the trained model to output first drug information indicating a first drug that can be administered to the first patient and first effect information indicating the estimated therapeutic effect of the first drug on the target disease.

[0009] In order to solve the above-mentioned problem, one embodiment of the present invention provides a trained model for estimating a first drug that can be administered to a first patient and an estimated therapeutic effect of the first drug on the target disease, based on a first medical image of a lesion site related to a target disease in the first patient and first expression information regarding gene expression at the lesion site corresponding to the first medical image, the trained model including, as explanatory variables, a second medical image of the lesion site in a second patient and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and obtained by machine learning using training data including, as objective variables, second drug information indicating the second drug administered to the second patient and second effect information indicating the therapeutic effect of the second drug on the target disease, and causing a computer to function to output, from the first medical image and the first expression information, first drug information indicating the first drug and first effect information indicating the estimated therapeutic effect of the first drug.

[0010] The drug efficacy estimation system according to each aspect of the present invention may be realized by a computer. In this case, the scope of the present invention also includes a computer control program that causes the computer to operate as each part (software element) of the drug efficacy estimation system, and a computer-readable recording medium on which the program is recorded. [Effects of the Invention]

[0011] According to one aspect of the present invention, it is possible to realize a drug efficacy estimation method and the like that can estimate the therapeutic effect of a drug with high accuracy using medical images and gene expression information. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a block diagram showing an example of a schematic configuration of a drug efficacy estimation system according to a first embodiment. [Figure 2] 1 is a flowchart showing an example of the processing flow of a model generation step executed by the drug efficacy estimation device according to the first embodiment. [Figure 3] 1 is a flowchart showing an example of the processing flow of an estimation step executed by a drug efficacy estimation device according to the first embodiment. [Figure 4] FIG. 1 is a diagram showing an example of a schematic configuration of a drug efficacy estimation system according to a second embodiment. [Figure 5] FIG. 10 is a block diagram showing an example of a schematic configuration of a drug efficacy estimation device according to a second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0013] [Embodiment 1] <Overview of the drug efficacy estimation system> The schematic configuration of the drug efficacy estimation system 100 will be described below with reference to Fig. 1. Fig. 1 is a block diagram showing an example of the schematic configuration of the drug efficacy estimation system 100.

[0014] As shown in Fig. 1, a drug efficacy estimation system 100 may include a drug efficacy estimation device 1 and a display device 4. Fig. 1 shows a drug efficacy estimation system 100 including one drug efficacy estimation device 1 and one display device 4. However, the configuration of the drug efficacy estimation system 100 is not limited to this. For example, the drug efficacy estimation system 100 may not include the display device 4, or may include multiple display devices 4.

[0015] In the drug efficacy estimation system 100, the drug efficacy estimation device 1 and the display device 4 are connected to each other so that they can communicate with each other. The drug efficacy estimation device 1 and the display device 4 may be connected directly, by wire or wirelessly, or via a communication network. The type of the communication network is not limited, and may be a local area network (LAN) or the Internet.

[0016] The drug efficacy estimation device 1 is a device that estimates the therapeutic effect of a drug on a target disease using medical images of a lesion site and expression information on gene expression at the lesion site in a first patient who is a patient to be estimated. The estimation result by the drug efficacy estimation device 1 may be transmitted from the drug efficacy estimation device 1 to a display device 4.

[0017] The display device 4 may be a computer, a smartphone, a tablet terminal, or the like used by a user of the drug efficacy estimation system 100. Note that FIG. 1 shows the drug efficacy estimation system 100 in which the display device 4 is separate from the drug efficacy estimation device 1. However, the configuration of the drug efficacy estimation system 100 is not limited to this. For example, the display device 4 may be a device integrated with the drug efficacy estimation device 1, and in this case, the display device 4 may be a display unit (such as a display) included in the drug efficacy estimation device 1.

[0018] <Configuration of drug efficacy estimation device 1> Next, we will explain the configuration of the drug efficacy estimation device 1. The drug efficacy estimation device 1 includes a control unit 10, a storage unit 20, and an input unit 30.

[0019] The storage unit 20 may be, for example, a hard disk drive (HDD) or a solid state drive (SSD). The storage unit 20 stores information transmitted from the control unit 10, and the stored information may be read by the control unit 10.

[0020] The input unit 30 is configured to allow a user of the drug efficacy estimation device 1 to input information. The input unit 30 may be, for example, at least one of a keyboard, a mouse, or a touchpad. The input unit 30 may also be an interface that accepts input of information from an external storage device such as a USB memory.

[0021] The control unit 10 is a control device that controls all the components of the drug efficacy estimation device 1. The control unit 10 may be, for example, a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The control unit 10 includes an acquisition unit 11, a model generation unit 12, and an estimation unit 13.

[0022] The control unit 10 may read a control program, which is software stored in the storage unit 20, and load it into a memory such as a RAM (Random Access Memory), and execute the functions of each unit. For the sake of simplicity, the storage unit 20 shown in Fig. 1 does not show stored data such as the control program.

[0023] (Acquisition part 11) The acquisition unit 11 acquires a first medical image of a lesion site related to a target disease of a first patient, and first expression information regarding gene expression at the lesion site corresponding to the first medical image. In this specification, the first patient refers to a patient suffering from the target disease and a patient who is a subject of drug efficacy estimation. Furthermore, the second patient, which will be described later, refers to a patient suffering from the target disease and for whom information regarding the course of treatment is known.

[0024] In this specification, various information such as medical images related to a first patient will be labeled "first," and various information related to a second patient will be labeled "second." In addition, when describing general matters that do not specify the type of patient, the names of these various information may not be labeled "first" or "second."

[0025] The target disease is not particularly limited as long as a lesion site appears due to the onset or progression of the pathology and medical images of the lesion site can be obtained. The lesion site does not need to be visually identifiable, and may be one in which, for example, a feature quantity in the medical image that can be extracted by computer image analysis or the like changes compared to the normal state.

[0026] Examples of target diseases include cancer and stroke. The cancer may be a solid cancer, and examples thereof include brain cancer, head and neck cancer, lung cancer, kidney cancer, liver cancer, digestive cancer, skin cancer, bladder cancer, breast cancer, uterine cancer, ovarian cancer, prostate cancer, and sarcoma.

[0027] The lesion site is appropriately set depending on the type of target disease. For example, if the target disease is lung cancer, it may be the entire lung or a specific part of the lung. Examples of medical images of the lesion site include computed tomography (CT) images, magnetic resonance imaging (MRI) images, X-ray images, endoscopic images, and tissue staining images.

[0028] The first expression information is information about gene expression in the lesion site corresponding to the first medical image. The information about gene expression may be, for example, information indicating RNA expression levels. Such expression information may be obtained by a comprehensive RNA expression analysis method such as RNA sequencing (RNA-seq).

[0029] The first medical image and the first expression information of the first patient acquired by the acquisition unit 11 are used to estimate the therapeutic effect of a first drug that can be administered to the first patient on the first patient.

[0030] The acquiring unit 11 may acquire the various types of information from the storage unit 20. The acquiring unit 11 may also acquire at least a part of the various types of information as input information from the input unit 30.

[0031] (Model generation unit 12) The model generation unit 12 generates a trained model used to estimate the therapeutic effect of the first drug on the first patient. Specifically, the model generation unit 12 generates the trained model by machine learning using training data that includes the second medical image and the second expression information as explanatory variables and the second drug information and the second effect information as objective variables.

[0032] The second medical images and second expression information are medical images and expression information of the second patient, and are the same as those of the first patient. The second medical images and second expression information are preferably information obtained before the start of drug treatment for the second patient. Note that the model generation unit 12 preferably generates a trained model using various information about multiple second patients, and the more second patients for which information exists, the more preferable it is.

[0033] Each of the first and second medical images may include information indicating the type of prognosis-related feature, which is a feature related to the prognosis of the target disease. Furthermore, each of the first and second expression information may include function information indicating the function of a gene whose expression level changes with a change in the prognosis-related feature.

[0034] The feature amount may be an index that quantitatively indicates the state of each feature included in a medical image. The feature amount may be represented as a one-dimensional scalar amount or a two- or more-dimensional vector amount. From the viewpoint of reducing the processing load of the drug efficacy estimation device 1, a dimension reduction process may be performed on the feature amount represented as a two- or more-dimensional vector amount.

[0035] The types of features include the size, shape, density, and texture of the lesion. The number of types of features may be one or more, ranging from a few to tens of thousands, or even more. The features contained in the medical image may be extracted by a known method, for example, by performing image processing such as wavelet transform on the medical image.

[0036] The prognosis-related feature is a feature contained in a medical image that is related to the prognosis of the target disease. Information indicating the association with the prognosis is, for example, information including the overall survival time of a patient with the target disease after treatment. The correlation between each feature of the medical image and the prognosis of a patient with the target disease can be estimated using a known regression model or the like. For example, an example of such a regression model is the Cox regression model when the information on the prognosis is information on survival time such as overall survival time.

[0037] The "prognosis-related feature" refers to a feature related to the progress or predicted progress of the target disease after treatment, and may be, for example, a feature that changes in correlation with the overall survival time before and after treatment. The model generation unit 12 may extract and use only the prognosis-related feature from among the feature included in the second medical image based on information indicating the type of such prognosis-related feature.

[0038] Regarding the functional information indicating the function of a gene whose expression level changes with a change in a prognosis-related feature, the gene function may be, for example, a function classified by GO (Gene Ontology) terms, or a function classified by other methods. When multiple genes whose expression level changes with a change in a prognosis-related feature are found, common functions or representative functions of the gene group can be extracted, for example, by enrichment analysis.

[0039] With regard to the above-mentioned information indicating the type of prognosis-related feature and functional information regarding the second patient, the information may be generated by the model generation unit 12, and the information regarding the first patient may be generated by the estimation unit 13. An example of processing when the model generation unit 12 generates the information indicating the type of prognosis-related feature and functional information regarding the second patient is shown below. The estimation unit 13 may also generate this information regarding the first patient using a similar method.

[0040] The model generation unit 12 generates information indicating the type of prognosis-related feature of the second patient using information on the second medical image and second effect information (described later) including information on the overall survival time after treatment of the second patient, etc. In addition to the information indicating the type of prognosis-related feature of the second patient, the model generation unit 12 also acquires the magnitude of the prognosis-related feature, etc.

[0041] The model generation unit 12 may, for example, estimate the correlation between each feature included in the second medical image and the prognosis such as the overall survival time of the second patient using a known regression model, etc. Then, the model generation unit 12 may extract, from among the feature amounts included in the second medical image, a feature amount that is estimated to be statistically significantly correlated with the prognosis of the second patient, as a prognosis-related feature amount.

[0042] Furthermore, the model generation unit 12 generates functional information indicating the functions of genes whose expression levels change with changes in the prognosis-related features using the extracted information on the prognosis-related features and the second expression information. The model generation unit 12 may, for example, estimate the correlation between comprehensive gene expression information obtained by RNA sequencing or the like included in the second expression information and the prognosis-related features using a known regression model or the like. Furthermore, the functions of genes estimated to be correlated with the prognosis-related features may be estimated by enrichment analysis.

[0043] Furthermore, the model generation unit 12 may use an ensemble learning method that combines multiple learning algorithms to estimate genes correlated with prognosis-related features and generate functional information about them. For example, the model generation unit 12 may use an algorithm that combines processing using a learning model suitable for natural language processing with correlation analysis using other regression models, etc. An example of a learning model suitable for natural language processing is the Transformer model.

[0044] When the model generation unit 12 uses a Transformer model, the Transformer model may be adjusted to generate a connection between information on features included in the second medical image and gene expression information included in the second expression information. For example, the Transformer model can read associations such as the order and relationship of each prognosis-related feature, including position information of the prognosis-related feature in the second medical image, and include such associations in the learning. Furthermore, the association information used for learning preferably includes associations between the prognosis-related feature included in the second medical image and the gene expression information. The Transformer model may be trained using such association information to estimate combinations of prognosis-related feature and gene expression information that are highly associated.

[0045] Furthermore, the model generation unit 12 may use a reinforcement learning algorithm as a learning model used to generate functional information. By using such a reinforcement learning algorithm, the model generation unit 12 can weight learning by setting rewards and accurately estimate a group of genes correlated with prognosis-related features and the functions of the group of genes.

[0046] The information used for setting the reward may be, for example, a correlation value obtained by correlation analysis between the prognosis-related feature extracted from the second medical image and the second expression information. Alternatively, the information used for setting the reward may be an enrichment score obtained by enrichment analysis of a gene group estimated to be correlated with the prognosis-related feature. Alternatively, the information may be a combination of the correlation value and the enrichment score.

[0047] The model generating unit 12 may perform learning on the correlation between the prognosis-related feature and the second expression information by weighting the information (value) used for setting such a reward using the enrichment score and the correlation value.

[0048] The second drug information is information about a second drug administered to a second patient. The second drug information may be, for example, identification information of the second drug, such as the generic name or compound name of the second drug. Furthermore, if the second drug is a molecular targeted therapeutic drug, the second drug information preferably further includes information about the target molecule.

[0049] The second efficacy information is information indicating the therapeutic efficacy of the second drug for the target disease. The second efficacy information may be information such as overall survival time (OS, the number of days from the start of treatment to death or discontinuation) and survival or death of the second patient to whom the drug was administered.

[0050] Furthermore, the second effect information may further include information on a drug response value indicating the effect on the cells of the second patient as information indicating a more specific therapeutic effect. The second drug may selectively damage cells (abnormal cells) in the lesion site as a therapeutic effect to shrink the lesion site. The second drug may also cause some damage to cells (normal cells) outside the lesion site.

[0051] The information on the drug response value of the second patient may thus be information that quantifies the positive or negative reaction that the second drug has on the cells of the second patient from the perspective of treating the target disease. In other words, the second effect information may include, for example, information indicating the degree to which the second drug causes cell damage to normal cells and abnormal cells of the second patient.

[0052] The second medical image, second expression information, second drug information, and second effect information used by the model generation unit 12 may be acquired by the acquisition unit 11. When the acquisition unit 11 acquires this information from the storage unit 20, the storage unit 20 may store a database in which various information indicating the treatment progress of the target disease, such as the second medical image, second expression information, second drug information, and second effect information of the second patient, is stored.

[0053] The model generation unit 12 generates a trained model by machine learning using training data that includes the second medical image and the second expression information as explanatory variables and the second drug information and the second effect information as objective variables. Specifically, the model generation unit 12 performs machine learning on the correlation between the second medical image and the second expression information and the second effect information by linking them to the second drug information for each second drug, and generates a trained model.

[0054] Such a trained model can output first effect information, which is an estimated therapeutic effect, for each first drug corresponding to each second drug used in training by inputting the first medical image and the first expression information. The trained model may also extract first drugs whose estimated therapeutic effect is equal to or greater than a predetermined threshold, and output only the corresponding first drug information and first effect information.

[0055] The model generation unit 12 may perform machine learning using, as explanatory variables, not only the second medical image and the second expression information themselves, but also prognosis-related features including functional information of related genes that can be generated from the second medical image and the second expression information. A trained model trained in this manner can output first effect information that estimates with high accuracy the effect on the prognosis of a first patient who has been administered the first drug based on the first medical image and the first expression information.

[0056] In this way, the model generation unit 12 may generate a trained model using information that can be generated from the second medical image and the second expression information, rather than the second medical image and the second expression information themselves, as explanatory variables included in the training data. This method is included as one of the aspects in which the model generation unit 12 trains a trained model using the second medical image and the second expression information as explanatory variables.

[0057] The machine learning algorithm used by the model generation unit 12 to train the trained model is not particularly limited, and may be, for example, a reinforcement learning algorithm or a supervised learning algorithm. These algorithms are preferably deep learning algorithms that use a neural network at least in part.

[0058] The model generation unit 12 preferably uses a reinforcement learning algorithm. However, reinforcement learning algorithms are not as widely used as supervised learning algorithms, particularly in the medical field. This is thought to be due to the fact that the absolute number of AI (Artificial Intelligence) developers among medical professionals, such as doctors, is small, and the absolute number of AI developers involved in clinical settings is also small. Therefore, while supervised learning algorithms, for which general-purpose development tools are becoming more widely used in the medical field, other algorithms, such as reinforcement learning algorithms, are not being widely used.

[0059] In reinforcement learning algorithms, it is important to appropriately set the agent's state, policy, and reward in order to obtain accurate estimation results. The inventors have conducted extensive research, particularly into reward setting, and have discovered a method for accurately estimating the therapeutic effects of drugs, especially those resulting from the combined use of multiple drugs.

[0060] When the model generation unit 12 generates a trained model using a reinforcement learning algorithm, the trained model may be trained using a reinforcement learning algorithm in which the agent's state, policy, and reward are set as follows: Status: Second medical image and second manifestation information; Strategy: Selection of one or more second drugs; Reward: When multiple second drugs are selected, integrated information of the corresponding multiple second effect information.

[0061] The model generation unit 12 causes the agent to sequentially select one or more selectable second drugs in each state set by the second medical image and the second expression information, and learns the correlation between each state and the second effect information for each selected second drug or combination thereof. Furthermore, the model generation unit 12 performs learning by weighting the correlation between each state and the second effect information based on the reward setting.

[0062] The model generation unit 12 may cause the agent to execute a strategy for each type of prognosis-related feature in the second medical image. The information indicating the type of prognosis-related feature may include functional information of related genes. This allows the model generation unit 12 to generate a trained model that estimates the therapeutic effect of a drug for each type of prognosis-related feature.

[0063] In this way, the state of the agent is not limited to the second medical image and the second expression information themselves, but may be set based on information that can be generated based on the second medical image and the second expression information. Such a method is included as one of the modes in which the model generation unit 12 sets the second medical image and the second expression information as the state of the agent and trains the trained model.

[0064] The model generation unit 12 may cause the agent to select one or more second drugs. When the second effect information of a certain second patient is information indicating an effect obtained by the combined use of multiple second drugs, the combination of the second drugs used in combination is preferably a combination of the second drugs. Note that the model generation unit 12 may also select a combination for which there is no second effect information due to the combined use as a combination of multiple second drugs.

[0065] When the agent selects multiple second drugs, the model generation unit 12 learns by weighting the correlation between information indicating a biological phenotype related to prognosis-related features and the second effect information based on the integrated information of the corresponding multiple second effect information.

[0066] For example, if the second effect information includes information on drug response values, the integrated information of the second effect information may be the integrated value, average value, and / or correlation value of the drug response values ​​for each second drug. Also, if the second effect information includes only information on overall survival time, the integrated information of the second effect information may be the average value and / or correlation value of overall survival time after administration of each second drug.

[0067] A correlation value between multiple drug response values ​​or overall survival times for multiple second drugs can be calculated, for example, by Pearson's correlation analysis, Spearman's rank correlation analysis, or mutual information.

[0068] The model generation unit 12 may perform learning so that, for example, the larger any one of the integrated value, average value, and correlation value of the drug response values ​​is, the larger (smaller) the gene expression level associated with the prognosis-related feature amount becomes. Note that, when the agent selects only one second drug, the model generation unit 12 may not perform weighting and may simply learn the correlation between information indicating a biological phenotype associated with the prognosis-related feature amount and the second effect information.

[0069] The trained model generated in this manner is a drug efficacy estimation model that, when input with a first medical image and first expression information of a first patient, can output first drug information indicating a first drug that can be administered to the first patient and the estimated therapeutic effect of the first drug on the target disease.

[0070] The control unit 10 does not necessarily have to include the model generation unit 12. In other words, the drug efficacy estimation device 1 may be a device that performs processing to estimate the therapeutic effect of a drug that can be administered to the first patient using a trained model that has been generated in advance. In this case, the drug efficacy estimation device 1 may store the trained model in the storage unit 20 in advance.

[0071] (Estimation part 13) The estimation unit 13 inputs the first medical image and the first expression information to the trained model, and causes the trained model to output the first drug information and the first effect information.

[0072] The first drug information is information indicating a first drug that can be administered to a first patient. The first drug information may be identification information of the first drug, similar to the second drug information.

[0073] The information on the first medical image may include information indicating the type of prognosis-related feature, similar to the information on the second medical image. Similarly to the second expression information, the first expression information may include functional information indicating the function of a gene whose expression level changes in accordance with a change in the prognosis-related feature. Similarly to the model generation unit 12, the estimation unit 13 may generate the information indicating the type of prognosis-related feature and the functional information based on the first medical image and the first expression information.

[0074] The trained model may output an estimated therapeutic effect for each first drug. For example, the trained model may treat each second drug included in the training data as a first drug that can be administered to the first patient, and output an estimated therapeutic effect for the first patient linked to first drug information for each first drug. That is, the estimation unit 13 may cause the trained model to output information of a list including combinations of multiple first drugs and estimated therapeutic effects.

[0075] The estimation unit 13 may include correspondence information between combinations of multiple first drugs and estimated therapeutic effects of the combinations in the list and output the list to the trained model. A user of the drug efficacy estimation device 1 can use such a list to refer to the estimated therapeutic effects of each of the multiple first drugs and determine one or more types of drugs to administer to the first patient.

[0076] The estimation unit 13 may extract, from the output of the trained model, a first drug or a combination thereof whose estimated therapeutic effect is equal to or greater than a predetermined threshold, and output only the corresponding first drug information and first effect information.

[0077] Furthermore, when the trained model is trained using the above-described reinforcement learning algorithm, the estimation unit 13 can output an estimated therapeutic effect of the first drug for each type of prognosis-related feature. For example, when a favorable estimated therapeutic effect is output for the same first drug or a combination thereof for multiple types of prognosis-related feature, the user can estimate that the first drug or the combination thereof is likely to have a favorable therapeutic effect.

[0078] The first patient and the second patient may each be a patient with a poor prognosis or a poor outcome. In other words, the estimation unit 13 may cause a trained model trained using second medical images and second expression information of a second patient diagnosed with a poor prognosis or a poor outcome to output an estimated therapeutic effect of a first drug that can be administered to the first patient diagnosed with a poor prognosis or a poor outcome.

[0079] With this configuration, the estimation unit 13 can cause the trained model to output an estimated therapeutic effect of a first drug for a first patient who has been diagnosed with a poor prognosis or poor progress and is considered to be relatively difficult to treat. Therefore, a user of the drug efficacy estimation device 1 can use the output result to easily select a first drug that is effective for a first patient who is considered to be relatively difficult to treat.

[0080] In this way, the drug efficacy estimation device 1 can estimate an effective first drug with high accuracy based on the first medical image and the first expression information. The drug efficacy estimation device 1 can also be applied to estimating an effective drug for a first patient with a poor prognosis or poor progress, and can also be applied to promoting treatment for patients who have traditionally been difficult to treat. Such effects contribute to achieving, for example, Goal 3 of the Sustainable Development Goals (SDGs) advocated by the United Nations, "Ensure good health and promote well-being for all."

[0081] <Drug efficacy estimation method> A drug efficacy estimation method according to one embodiment of the present invention will be described with reference to Figs. 2 and 3, taking as an example the flow of processing executed by a drug efficacy estimation device 1. Fig. 2 is a flowchart showing an example of the flow of processing of a model generation step executed by the drug efficacy estimation device 1. Fig. 3 is a flowchart showing an example of the flow of processing of an estimation step executed by the drug efficacy estimation device 1. Figs. 2 and 3 also show the flow of processing executed by a drug efficacy estimation system 100 including the drug efficacy estimation device 1. Note that the contents already explained in the section on the drug efficacy estimation system 100 above will not be explained here.

[0082] First, the model generation step will be described. As shown in Fig. 2, the acquisition unit 11 acquires second medical images, second expression information, second drug information, and second effect information for each second patient (S1). The various pieces of information acquired here are used as learning data by the model generation unit 12.

[0083] Next, the model generation unit 12 generates information indicating the type of prognosis-related feature based on the second medical image and the second effect information (S2). Note that the information indicating the type of prognosis-related feature may be included in the information of the second medical image acquired by the acquisition unit 11. In this case, the model generation unit 12 omits the process of S2.

[0084] Furthermore, the model generation unit 12 estimates a group of genes whose expression levels change with changes in the prognosis-related feature, based on information indicating the type of the prognosis-related feature and the second expression information. Then, it generates functional information indicating the biological functions of the estimated group of genes (S3). Note that the group of genes whose expression levels change with changes in the prognosis-related feature and the functional information indicating their biological functions may be included in the second expression information acquired by the acquisition unit 11. In this case, the model generation unit 12 omits the process of S3.

[0085] Next, the model generation unit 12 generates training data including the second medical image information including information indicating the type of prognosis-related feature and the function information as explanatory variables, and the second drug information and the second effect information as objective variables (S4). Note that the model generation unit 12 may generate training data using the second medical image and the second expression information themselves as explanatory variables.

[0086] Next, the model generation unit 12 generates a trained model by machine learning using the generated training data (S5). The model generation unit 12 preferably uses a reinforcement learning algorithm as a machine learning algorithm for generating the trained model. Note that the drug efficacy estimation device 1 may acquire training data generated by a computer other than the drug efficacy estimation device 1 performing the processes from S1 to S4, and execute the process of generating the trained model in S5.

[0087] Next, the estimation step will be described. As shown in Fig. 3, the acquisition unit 11 acquires a first medical image and first expression information (S6, acquisition step). The information of the first medical image may include information indicating the type of prognosis-related feature, similar to the information of the second medical image, and this information may be generated by the estimation unit 13. Furthermore, the first expression information may include function information indicating the function of a gene associated with the prognosis-related feature, similar to the second expression information, and this information may be generated by the estimation unit 13.

[0088] Next, the estimation unit 13 inputs the first medical image and the first expression information to the trained model generated by the model generation unit 12 (S7). Then, the estimation unit 13 causes the trained model to output the first drug information and the first effect information (S8). The output of the trained model includes the first drug information and the first effect information for each first drug that can be administered to the first patient.

[0089] <Pre-trained model> The present invention also encompasses the trained model described above. Specifically, a trained model according to one aspect of the present invention is a trained model for estimating a first drug administrable to a first patient and the estimated therapeutic effect of the first drug on a target disease based on a first medical image and first expression information of the first patient. The trained model is obtained by machine learning using training data that includes a second medical image and second expression information of a second patient as explanatory variables and includes second drug information and second effect information of the second drug administered to the second patient as objective variables. The trained model thus obtained causes a computer to function to output first drug information indicating the first drug and first effect information indicating the estimated therapeutic effect of the first drug from the first medical image and the first expression information.

[0090] [Embodiment 2] Other embodiments of the present invention are described below.

[0091] In the drug efficacy estimation system 100 shown in Fig. 1, the drug efficacy estimation device 1 includes an input unit 30 that accepts input of a first medical image, first expression information, etc. by a user, and outputs a drug efficacy estimation result to a display device 4, but the configuration is not limited to this. For example, as shown in Fig. 4, the drug efficacy estimation system 100a may include a drug efficacy estimation device 1a that is communicably connected to communication terminals 5a and 5b used by each user via a communication network 9.

[0092] In a drug efficacy estimation system 100a shown in Fig. 4, a drug efficacy estimation device 1a receives information such as a first medical image and first expression information from each of communication terminals 5a and 5b. Then, the drug efficacy estimation device 1a transmits an estimation result corresponding to the information received from the communication terminal 5a to the communication terminal 5a, and transmits an estimation result corresponding to the information received from the communication terminal 5b to the communication terminal 5b. The estimation results are the first drug information and first effect information that the estimation unit 13 causes the trained model to output.

[0093] 4 shows a drug efficacy estimation system 100a including the communication terminals 5a and 5b and the drug efficacy estimation device 1a, but is not limited thereto. In the drug efficacy estimation system 100a, the drug efficacy estimation device 1a may be capable of communicating with, for example, three or more communication terminals.

[0094] (Configuration of drug efficacy estimation device 1a) The configuration of a drug efficacy estimation device 1a will be described with reference to Fig. 5. Fig. 5 is a functional block diagram showing a configuration example of a drug efficacy estimation system 100a according to one embodiment of the present invention. For convenience of explanation, components having the same functions as those described in the above embodiment are denoted by the same reference numerals, and their description will not be repeated.

[0095] 5, the drug efficacy estimation device 1a includes a communication unit 60 that functions as a communication interface with the communication terminals 5a and 5b. The acquisition unit 11 receives information such as the first medical image and the first expression information via the communication unit 60.

[0096] The estimation unit 13 transmits the estimation result to each of the communication terminals 5a and 5b via the communication unit 60. The drug efficacy estimation device 1a may generate a web page showing the estimation result regarding the received information and provide information for accessing the web page to the user who is the sender of the information.

[0097] [Software implementation example] The functions of the drug efficacy estimation device 1 (hereinafter referred to as the "device") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device (particularly each part included in the control unit 10).

[0098] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program, thereby realizing the functions described in each of the above embodiments.

[0099] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0100] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0101] 〔summary〕 A drug efficacy estimation method according to aspect 1 of the present invention includes an acquisition step of acquiring a first medical image of a lesion site related to a target disease of a first patient, and first expression information regarding gene expression at the lesion site corresponding to the first medical image; and an estimation step of inputting the first medical image and the first expression information into a trained model trained using training data including, as explanatory variables, a second medical image of the lesion site of a second patient, and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and including, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating the therapeutic effect of the second drug on the target disease, and causing the trained model to output first drug information indicating a first drug that can be administered to the first patient and first effect information indicating the estimated therapeutic effect of the first drug on the target disease.

[0102] A drug efficacy estimation method according to aspect 2 of the present invention may be such that, in aspect 1, each of the information on the first medical image and the second medical image includes information indicating a type of prognosis-related feature, which is a feature related to the prognosis of the target disease, the first expression information and the second expression information each include functional information indicating a function of a gene whose expression level changes with a change in the prognosis-related feature, and the first patient and the second patient each may be a patient with a poor prognosis or a poor course of disease.

[0103] A drug efficacy estimation method according to aspect 3 of the present invention may be such that, in aspect 1 or 2, the trained model is trained by a reinforcement learning algorithm in which the second medical image and the second expression information are set as the state of an agent, the selection of the second drug is set as the strategy of the agent, and, when multiple second drugs are selected, integrated information of the corresponding multiple pieces of second effect information is set as a reward for the strategy.

[0104] A method for generating a drug efficacy estimation model according to a fourth aspect of the present invention generates a trained model by machine learning using training data including, as explanatory variables, a second medical image of a lesion site related to a target disease in a second patient and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating the therapeutic effect of the second drug on the target disease.

[0105] A drug efficacy estimation system according to a fifth aspect of the present invention includes an acquisition unit that acquires a first medical image of a lesion site related to a target disease of a first patient and first expression information regarding gene expression at the lesion site corresponding to the first medical image; and an estimation unit that inputs the first medical image and the first expression information into a trained model that is trained using training data that includes, as explanatory variables, a second medical image of the lesion site of a second patient and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and includes, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating the therapeutic effect of the second drug on the target disease, and outputs, to the trained model, first drug information indicating a first drug that can be administered to the first patient and first effect information indicating the estimated therapeutic effect of the first drug on the target disease.

[0106] A control program according to a sixth aspect of the present invention is a control program for controlling a computer, and causes the computer to execute an acquisition step of acquiring a first medical image of a lesion site related to a target disease of a first patient, and first expression information regarding gene expression at the lesion site corresponding to the first medical image; and an estimation step of inputting the first medical image and the first expression information into a trained model trained using training data including, as explanatory variables, a second medical image of the lesion site of a second patient, and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and including, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating the therapeutic effect of the second drug on the target disease, and causing the trained model to output first drug information indicating a first drug that can be administered to the first patient and first effect information indicating the estimated therapeutic effect of the first drug on the target disease.

[0107] A recording medium according to a seventh aspect of the present invention is a computer-readable recording medium on which the control program according to the sixth aspect is recorded.

[0108] A trained model according to aspect 8 of the present invention is a trained model for estimating a first drug that can be administered to a first patient and an estimated therapeutic effect of the first drug on the target disease, based on a first medical image of a lesion site related to a target disease in the first patient and first expression information regarding gene expression at the lesion site corresponding to the first medical image, the trained model including, as explanatory variables, a second medical image of the lesion site in a second patient and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and obtained by machine learning using training data including, as objective variables, second drug information indicating the second drug administered to the second patient and second effect information indicating the therapeutic effect of the second drug on the target disease, and causing a computer to function to output, from the first medical image and the first expression information, first drug information indicating the first drug and first effect information indicating the estimated therapeutic effect of the first drug.

[0109] [Additional Notes] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Example]

[0110] An embodiment of the present invention will be described below, but the present invention is not limited to the scope of each example shown below.

[0111] Example 1: Prediction of prognosis using medical images and estimation of prognosis-related features We attempted to predict prognosis using medical images of the lungs of non-small cell lung cancer patients and evaluated its accuracy. As a control group, we also estimated the accuracy of prognosis prediction using gene expression analysis data of non-small cell lung cancer patients and compared it with the results using medical images.

[0112] Non-small cell lung cancer patients who met the following conditions (1) to (3) were selected, and CT image data and postoperative follow-up data were obtained. (1) Histopathologically diagnosed non-small cell lung cancer (2) Preoperative pulmonary CT image data is available; (3) Postoperative follow-up data includes information on overall survival (OS, the number of days from the start of irradiation to death or discontinuation) and survival.

[0113] These data were obtained from The Cancer Imaging Archive portal and The Cancer Genome Atlas database. Data from 150 cases were obtained, 100 cases were used as the training dataset, and 50 cases were used as the validation dataset.

[0114] To reduce variations between institutions, we used plain CT images of the lungs of patients with non-small cell lung cancer without contrast enhancement. The plain CT images were visually classified into tumor and non-tumor areas. Feature values ​​for each of the tumor and non-tumor areas were extracted using the wavelet imaging filter function in PyRadiomics software. The extracted feature values ​​were then subjected to dimensionality reduction using Lasso regression.

[0115] Next, prognosis-related features associated with prognosis were estimated from the extracted features. Using a Lasso-Cox regression model, the values ​​of tumor and / or non-tumor areas were set as covariates for each feature in the medical image, and the impact on overall patient survival was evaluated. Features evaluated by the model to have a significant impact on overall patient survival (p-value < 0.05) were estimated as prognosis-related features.

[0116] The types of estimated prognosis-related features are shown in Table 1 below. The names of the types of prognosis-related features are based on the classification of the PyRadiomics software.

[0117] [Table 1]

[0118] A Radiomics-score (Rad-score) was calculated by extracting the importance of each prognosis-related feature obtained by Lasso regression using data correlating prognosis-related features with progression data in the training dataset. A trained model was created using an ensemble learning model that combined a one-dimensional machine learning model using the Rad-Score as input data with a multidimensional machine learning model that input all the features in the medical images obtained by Lasso regression. Medical images included in the validation dataset were input into the resulting trained model, and prognosis prediction was performed.

[0119] As a control, we also created a model trained on gene expression data rather than medical images. RNA sequencing expression analysis data from lung tissues obtained from The Cancer Genome Atlas database for the above patients was associated with progression data and used as input data. This data was then input into the ensemble learning model to create a trained model. The RNA sequencing expression analysis data included in the validation dataset was input into the resulting trained model, and prognosis prediction was performed.

[0120] The predictive accuracy of each trained model was evaluated using the AUC (Area Under Curve) after ROC (Receiver Operating Characteristic) analysis. The AUC of the trained model trained using medical images was 0.90, while the AUC of the trained model trained using gene expression data was 0.71. In other words, the trained model trained using medical images showed better prognostic prediction accuracy than the trained model trained using gene expression data. These results demonstrate that prognostic information can be obtained with high accuracy by using medical images.

[0121] Example 2: Association of prognosis-related features in medical images with biological phenotypes We attempted to correlate prognosis-related features with biological phenotypes based on gene expression data.

[0122] The prognosis-related features estimated in Example 1 were analyzed for their association with biological phenotypes. A Transformer model, which has been utilized in language models, was used for a group of genes whose expression levels change in conjunction with changes in the prognosis-related features. The Transformer model is a model improved for medical use so as to generate connections between features in medical images and gene expression data. Furthermore, enrichment analysis was combined with the estimation to estimate the functional classification of the estimated gene group.

[0123] The gene expression data used was the RNA sequencing expression analysis data described above. The functional classification data for genes were obtained from the MSigDB C2 and C5 collections registered in the Molecular Signature Database (MSigDB, https: / / www.gsea-msigdb.org / gsea / msigdb / ). The C2 collection mainly contains gene annotation information from literature, while the C5 collection mainly contains classification information based on GO (Gene Ontology) terms.

[0124] Below are biological phenotypes that are estimated to be associated with prognosis-related features in medical images of non-small cell lung cancer patients: RNA binding: A function that plays an important role in intracellular signal transduction and protein synthesis. It may be associated with cancer cell proliferation and may affect prognosis. Nucleolus: A specific region within the cell nucleus that is involved in rRNA synthesis and ribosome assembly. It may be associated with cancer cell proliferation and may affect prognosis. Macroautophagy (macrophagy), Regulation of autophagy: Autophagy is a function that breaks down waste products within cells, providing energy and maintaining survival. Cancer cells may utilize autophagy to maintain their survival, which may affect prognosis.

[0125] All of the estimated phenotypes were suggested to be related to the treatment prognosis of non-small cell lung cancer. In other words, it was suggested that prognosis-related features extracted from medical images may be effective indicators of biological phenotypes related to prognosis.

[0126] Example 3: Drug Discovery Method We attempted to use medical images to identify drugs suitable for patients with poor prognosis. A reinforcement learning model was used for drug discovery. The training data for the reinforcement learning model consisted of medical images of patients with non-small cell lung cancer, containing information indicating the relationship between prognosis-related features and biological phenotypes based on gene expression data.

[0127] In addition, drug treatment response data registered in the LINCS1000 database, which indicates the effect on each gene in cells after drug administration (drug response value), was obtained and input into the reinforcement learning model. The effect on cells indicates cell repair (proliferation) or damage (reduction), etc. When evaluating these effects on tumor cells (abnormal cells) as treatment effects, cell repair can be said to be a negative treatment effect, and cell damage can be said to be a positive treatment effect.

[0128] In training the reinforcement learning model, the various conditions were set as follows: Environment, Agent: A simulation model using a reinforcement learning algorithm. It learns the correlation between information indicating biological phenotypes related to prognosis-related features and drug response values. Status: Prognostic features, including information related to biological phenotypes based on gene expression data, as described above. Strategy: Selection of one or more drugs Reward: Accumulation, average and correlation of drug response values ​​when multiple drugs are selected The trained model was input with medical images of the patient to be estimated, including information indicating the relationship between prognosis-related features and biological phenotypes based on gene expression data.The trained model was then made to output estimated results of the therapeutic response of each drug for each prognosis-related feature, and from these, the drug estimated to show the optimal therapeutic response for the patient to be estimated was selected.

[0129] The drugs that are presumed to be suitable for patients with poor prognosis or poor course of disease are listed below. GSK650394: SGK1 inhibitor SB.366791: TRPV1 ligand VE-922 (Berzosertib): Atr inhibitor These drugs are all used for some lung cancer patients. Because these drugs were estimated using the reinforcement learning model, it is expected that they will improve therapeutic effects by selectively applying them to patients with poor prognosis or poor course of disease, rather than as conventional comprehensive options for small cell lung cancer patients.

[0130] By using the above-mentioned reinforcement learning model, it is possible to select one drug that is estimated to have the optimal therapeutic effect from one type of prognosis-related feature. Therefore, by using multiple prognosis-related feature values, it is possible to select multiple types of drugs, and it is also possible to find a drug that is estimated to be optimal for two or more prognosis-related feature values.

[0131] According to a drug discovery method according to one aspect of the present invention, drug discovery can be easily performed taking into account multiple biological phenotypes using multiple types of prognosis-related features, compared to conventional methods such as correlation analysis, etc. In other words, it is possible to discover drugs with superior therapeutic effects more effectively than conventional methods. [Explanation of symbols]

[0132] 11 Acquisition Department 12 Model Generation Unit 13 Estimation part 100 Drug Efficacy Estimation System

Claims

1. an acquiring step of acquiring a first medical image of a lesion site related to a target disease of a first patient, and first expression information regarding gene expression in the lesion site corresponding to the first medical image; an estimation step of inputting the first medical image and the first expression information into a trained model trained using training data including, as explanatory variables, a second medical image of the lesion site of a second patient and second expression information related to gene expression at the lesion site corresponding to the second medical image, and including, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating a therapeutic effect of the second drug on the target disease, and causing the trained model to output first drug information indicating a first drug that can be administered to the first patient and first effect information indicating an estimated therapeutic effect of the first drug on the target disease; each piece of information about the first medical image and the second medical image includes information indicating a type of prognosis-related feature, which is a feature related to the prognosis of the target disease; each of the first expression information and the second expression information includes function information indicating a function of a gene whose expression level changes in accordance with a change in the prognosis-related feature amount; A method for estimating drug efficacy, wherein the first patient and the second patient are each patients with poor prognosis or poor course of disease.

2. an acquiring step of acquiring a first medical image of a lesion site related to a target disease of a first patient, and first expression information regarding gene expression in the lesion site corresponding to the first medical image; an estimation step of inputting the first medical image and the first expression information into a trained model trained using training data including, as explanatory variables, a second medical image of the lesion site of a second patient and second expression information related to gene expression at the lesion site corresponding to the second medical image, and including, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating a therapeutic effect of the second drug on the target disease, and causing the trained model to output first drug information indicating a first drug that can be administered to the first patient and first effect information indicating an estimated therapeutic effect of the first drug on the target disease; a drug efficacy estimation method, wherein the trained model is trained using a reinforcement learning algorithm in which the second medical image and the second expression information are set as the state of an agent, the selection of one or more second drugs is set as the agent's strategy, and when multiple second drugs are selected, integrated information of the corresponding multiple pieces of second effect information is set as a reward for the strategy.

3. each piece of information about the first medical image and the second medical image includes information indicating a type of prognosis-related feature, which is a feature related to the prognosis of the target disease; each of the first expression information and the second expression information includes function information indicating a function of a gene whose expression level changes in accordance with a change in the prognosis-related feature amount; The method for estimating a drug efficacy according to claim 2 , wherein the first patient and the second patient are each a patient with a poor prognosis or a poor course of disease.

4. generating a trained model by machine learning using training data including, as explanatory variables, a second medical image of a lesion site related to a target disease in a second patient and second expression information related to gene expression at the lesion site corresponding to the second medical image, and including, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating a therapeutic effect of the second drug on the target disease; the information on the second medical image includes information indicating a type of prognosis-related feature, which is a feature related to the prognosis of the target disease; the second expression information includes function information indicating a function of a gene whose expression level changes in accordance with a change in the prognosis-related feature amount, The method for generating a drug efficacy prediction model, wherein the second patient is a patient with poor prognosis or poor course of disease.

5. an acquisition unit that acquires a first medical image of a lesion site related to a target disease of a first patient and first expression information regarding gene expression in the lesion site corresponding to the first medical image; an estimation unit that inputs the first medical image and the first expression information into a trained model trained using training data including, as explanatory variables, a second medical image of the lesion site of a second patient and second expression information related to gene expression at the lesion site corresponding to the second medical image, and including, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating a therapeutic effect of the second drug on the target disease, and causes the trained model to output first drug information indicating a first drug that can be administered to the first patient and first effect information indicating an estimated therapeutic effect of the first drug on the target disease; each piece of information about the first medical image and the second medical image includes information indicating a type of prognosis-related feature, which is a feature related to the prognosis of the target disease; each of the first expression information and the second expression information includes function information indicating a function of a gene whose expression level changes in accordance with a change in the prognosis-related feature amount; A drug efficacy estimation system, wherein the first patient and the second patient are each patients with poor prognosis or poor course of disease.

6. A control program for controlling a computer, The computer, an acquiring step of acquiring a first medical image of a lesion site related to a target disease of a first patient, and first expression information regarding gene expression in the lesion site corresponding to the first medical image; an estimation step of inputting the first medical image and the first expression information into a trained model trained using training data including, as explanatory variables, a second medical image of the lesion site of a second patient and second expression information related to gene expression at the lesion site corresponding to the second medical image, and including, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating a therapeutic effect of the second drug on the target disease, and outputting, from the trained model, first drug information indicating a first drug that can be administered to the first patient and first effect information indicating an estimated therapeutic effect of the first drug on the target disease; each piece of information about the first medical image and the second medical image includes information indicating a type of prognosis-related feature, which is a feature related to the prognosis of the target disease; each of the first expression information and the second expression information includes function information indicating a function of a gene whose expression level changes in accordance with a change in the prognosis-related feature amount; A control program, wherein the first patient and the second patient are each patients with poor prognosis or poor course of disease.

7. A computer-readable recording medium on which the control program according to claim 6 is recorded.

8. A trained model for estimating a first drug that can be administered to a first patient and an estimated therapeutic effect of the first drug on the target disease, based on a first medical image of a lesion site related to a target disease of the first patient and first expression information regarding gene expression at the lesion site corresponding to the first medical image, the data is obtained by machine learning using learning data including, as explanatory variables, a second medical image of the lesion site of a second patient and second expression information regarding gene expression at the lesion site corresponding to the second medical image, and including, as objective variables, second drug information indicating a second drug administered to the second patient and second effect information indicating a therapeutic effect of the second drug on the target disease; causing a computer to function to output, from the first medical image and the first expression information, first drug information indicating the first drug and first effect information indicating the estimated therapeutic effect of the first drug; each piece of information about the first medical image and the second medical image includes information indicating a type of prognosis-related feature, which is a feature related to the prognosis of the target disease; each of the first expression information and the second expression information includes function information indicating a function of a gene whose expression level changes in accordance with a change in the prognosis-related feature amount; A trained model, wherein the first patient and the second patient are each patients with poor prognosis or poor course of illness.

Images