JP2021043600A - Method for generating learning model, evaluation estimation device, learning model and computer program - Google Patents

Abstract

To provide a method for generating a learning model capable of estimating the physical properties of cells induced to differentiate from artificial pluripotent stem cells into predetermined cells, an evaluation estimation device, a learning model and a computer program.SOLUTION: A method for generating a learning model comprises: inducing differentiation from artificial pluripotent stem cells into predetermined cells; acquiring cell information of the cells induced to differentiate into the predetermined cells; acquiring teacher data including the cell information and physical property information of the cells based on measurements using an analyzer; and when the cell information of the cells induced to differentiate into predetermined cells is input, generating a learning model that outputs the physical property information of the cells, based on the teacher data.SELECTED DRAWING: Figure 13

Description

The present invention relates to a learning model generation method, an evaluation estimation device, a learning model, and a computer program.

Drug discovery research is costly and time consuming. In particular, drug discovery research for intractable diseases is difficult from the viewpoint that it is difficult to obtain disease model cells or disease model animals.
In addition, drug discovery screening by animal tests has problems in terms of animal protection and inefficiency of animal tests (non-correlation with tests in the human body).
Induced differentiation of patient-derived induced pluripotent stem cells to produce cells corresponding to the target disease (hereinafter referred to as differentiated cells), the target drug is added to the cells, and the drug is based on the characteristics of the cells. Research is underway to evaluate the toxicity and efficacy of (for example, Patent Document 1). Since patient-derived induced pluripotent stem cells are used, disease symptoms can be reproduced, the onset mechanism can be elucidated, and the development of therapeutic agents for intractable diseases can be expected.
Differentiated cells express the same functions as cells of human living tissue, and can be expected to reduce lot differences and provide a stable supply.

Re-table 2016/076435

Even in the case of drug discovery screening of Patent Document 1, there is a problem that a great cost is required to measure the physical properties of cells before and after the addition of the drug, and the evaluation protocol is complicated.

The present invention has been made in view of such circumstances, and is a method for generating a learning model capable of estimating the physical characteristics of a predetermined cell induced to differentiate from an induced pluripotent stem cell, an evaluation estimation device, a learning model, and a learning model. The purpose is to provide computer programs.

The method for generating a learning model according to one aspect of the present invention is to induce differentiation from an artificial pluripotent stem cell into a predetermined cell, acquire cell information of the cell induced to differentiate into a predetermined cell, and obtain the cell information and the cell information. When teacher data including physical property information of the cell based on measurement using an analyzer is acquired and the cell information of the cell whose differentiation is induced into a predetermined cell is input based on the teacher data, the cell Generate a learning model that outputs the physical property information of.

According to the above configuration, the cell information of the predetermined cell is input based on the cell information of the cell induced to differentiate into the predetermined cell and the teacher data including the physical property information of the cell based on the measurement using the analyzer. If so, it is possible to generate a learning model that satisfactorily estimates the physical property information of the cell. Therefore, the compound evaluation step can be omitted. It is also possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells based on the cell information before the addition of the compound.
In addition, cell information as a cell group including cells corresponding to the target disease induced to differentiate from artificial pluripotent stem cells of the same patient and cells induced to differentiate into cells of different cell types, and the above-mentioned cells. Based on the teacher data including the physical property information of the above, it is possible to generate a learning model that suitably estimates the possibility of the cell dynamics of the cell group and the mechanism by the signal regarding the aggravation or the cause of the symptom of the disease.
That is, the disease classification can be classified in more detail or the disease can be reclassified based on the cell information of the cell group induced to differentiate in the same disease patient or the disease reserve patient. With the detailed and accurate classification of the above-mentioned diseases, it is also possible to determine which compound is added to facilitate the evaluation result.

In the method for generating the learning model described above, the cell information of the cell induced to differentiate into a predetermined cell and the physical property information based on the actual measurement of the cell are acquired, and the learning model is obtained based on the cell information and the physical property information. You may relearn.

According to the above configuration, re-learning is performed based on the physical property information based on the actual measurement of the cells, so that the physical property information can be output more satisfactorily.

In the above-mentioned learning model generation method, compound information was acquired, the teacher data included compound information, and cell information and compound information of cells induced to differentiate into predetermined cells were input based on the teacher data. In some cases, a learning model that outputs physical property information may be generated.

According to the above configuration, since the teacher data includes the compound information, the physical property information can be output satisfactorily in consideration of the structure of the compound and the like.

In the method for generating the learning model described above, the cell information includes first information including cell type, clinical information, genetic information, and culture conditions, and second information including cell image and cell number after culture. It may be information.

According to the above configuration, cells can be specified under a plurality of conditions and physical property information can be output with high accuracy.

In the method for generating the learning model described above, the physical property information is at least one kind of information selected from the number of cells, light absorption characteristics of a predetermined wavelength, chromosomal information, image information, action potential, medium information, and marker value. There may be.

According to the above configuration, physical property information corresponding to a desired evaluation item can be output and evaluated.

In the method for generating the learning model described above, the physical property information may indicate the amount of change over time.

According to the above configuration, the effect of the compound over time can be obtained based on the change over time in the physical properties.

The evaluation estimation device according to one aspect of the present invention obtains cell information of cells induced to differentiate into predetermined cells from artificial pluripotent stem cells, and cell information of cells induced to differentiate into predetermined cells. When input, the cell information acquired by the acquisition unit is input to the learning model that outputs the physical property information of the cell based on the measurement using the analyzer, and the estimation unit estimates the physical property information. It is provided with an output unit that outputs physical property information.

According to the above configuration, since the cell information of the cell induced to differentiate into a predetermined cell is input to the learning model and the physical property information is acquired, the costly and complicated evaluation step of the compound can be omitted. It is also possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells based on the cell information before the addition of the compound.

In the evaluation estimation device described above, the acquisition unit acquires compound information, the learning model outputs physical property information when the compound information is input, and the estimation unit outputs the physical property information to the learning model by the acquisition unit. The acquired compound information may be input.
According to the above configuration, since the teacher data includes the compound information, the physical property information can be output satisfactorily in consideration of the structure of the compound and the like.

In the evaluation estimation device described above, the acquisition unit acquires the cell information of the cell before the compound is added, and the learning model inputs the cell information before the compound is added, and the first physical property information is obtained. Is a model that outputs the above, and the estimation unit may estimate the first physical property information.
According to the above configuration, it is possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells based on the cell information before the addition of the compound.

In the evaluation estimation device described above, the acquisition unit acquires the cell information of the cell after the compound is added, and the learning model has the second physical characteristics when the cell information after the compound is added is input. It is a model that outputs information, and the estimation unit may estimate the second physical property information.
The evaluation estimation device may include a calculation unit that calculates the physical property information of the compound based on the first physical property information and the second physical property information estimated by the estimation unit.
According to the above configuration, the effect / efficacy or toxicity / safety of the compound can be evaluated based on the difference between the second physical property information and the first physical property information.

In the above-mentioned evaluation estimation device, the cell information is information including first information including cell type, clinical information, genetic information, and culture conditions, and second information including cell image and cell number after culturing. There may be.
According to the above configuration, cells can be specified in detail and physical property information can be output with high accuracy.

In the above-mentioned evaluation estimation device, the physical property information includes the number of cells, absorption characteristics of light of a predetermined wavelength, chromosomal information, image information, action potential, medium information (metabolism (autologous secretion, amount of amino acid change)), and cells. At least one, preferably 2 to 5, more preferably 6 to 10 or more, selected from marker values indicating the presence or absence of expression of surface antigens and intracellular transcription factors, cell fluidity, or cell motility. Information is good.
According to the above configuration, it is possible to output and evaluate the physical property information corresponding to the desired evaluation item, and it is also possible to find a new surrogate marker by using the evaluation result.

The evaluation estimation device described above includes a determination unit for determining whether or not to add the compound based on a biomarker value, a surrogate marker value, or an image of the cell before adding the compound to the cell. You may.
According to the above configuration, it is possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells, and the evaluation process can be stopped.

In the above-mentioned evaluation estimation device, a display unit for displaying each physical property information estimated by the estimation unit is provided corresponding to each well of the differentiation-inducing culture device that induces differentiation from induced pluripotent stem cells into predetermined cells. May be good.
According to the above configuration, the effect / efficacy or toxicity / safety of the compound can be satisfactorily visually recognized by comparing the amount of change in physical properties before and after the addition of the compound for each well with that of the control.

The learning model according to one aspect of the present invention includes an input layer into which cell information of cells induced to differentiate from artificial pluripotent stem cells into predetermined cells is input, and physical property information of the cells based on measurement using an analyzer. An artificial pluripotent stem cell having an output layer for outputting the above, an intermediate layer in which parameters are learned based on cell information of cells induced to differentiate from artificial pluripotent stem cells into predetermined cells and physical property information of the cells. When the cell information of one cell whose differentiation is induced into a predetermined cell is input from the cell, the computer is made to function so as to output the physical property information of the cell from the output layer through the calculation by the intermediate layer.

According to the above configuration, when the cell information is input based on the teacher data including the cell information of the cell induced to differentiate into a predetermined cell and the physical property information of the cell based on the measurement using an analyzer. , Estimate good cell property information. Therefore, the compound evaluation step can be omitted.

The computer program according to one aspect of the present invention acquires cell information of a cell that has been induced to differentiate into a predetermined cell from an artificial pluripotent stem cell, and inputs the cell information of the cell that has been induced to differentiate into a predetermined cell. , The computer is made to execute the process of inputting the acquired cell information into a learning model that outputs the physical property information of the cell based on the measurement using the analyzer and outputting the physical property information.

According to the above configuration, when the cell information is input based on the teacher data including the cell information of the cell induced to differentiate into a predetermined cell and the physical property information of the cell based on the measurement using an analyzer. , Estimate good cell property information. Therefore, the compound evaluation step can be omitted.

According to the present invention, it is possible to estimate the physical characteristics of cells induced to differentiate into predetermined cells from induced pluripotent stem cells.

It is a schematic diagram which shows an example of the structure of the information processing system which concerns on Embodiment 1. FIG. It is a block diagram which shows an example of the structure of the evaluation estimation apparatus. It is explanatory drawing which shows an example of the record layout of the manufacturing protocol DB. It is explanatory drawing which shows an example of the record layout of the manufacturing program DB. It is explanatory drawing which shows an example of the record layout of the evaluation protocol DB. It is explanatory drawing which shows an example of the record layout of the compound information DB. It is explanatory drawing which shows an example of the record layout of a teacher data DB. It is explanatory drawing which shows an example of the record layout of the physical characteristic information output model DB. It is explanatory drawing which shows an example of the record layout of the use history DB. It is a block diagram which shows an example of the structure of a culture apparatus. It is a block diagram which shows an example of the configuration of a terminal. It is a flowchart which shows the procedure for producing a nerve cell by inducing differentiation from an induced pluripotent stem cell. It is a schematic diagram which shows the structure of the physical characteristic information output model. It is a flowchart which shows an example of the procedure of the generation processing of the physical characteristic information output model by a control unit. It is a flowchart which shows an example of the procedure of the process of acquisition and transmission of physical property information by a control unit. It is a flowchart which shows the procedure of the relearning process of the physical characteristic information output model by a control unit. It is a block diagram which shows an example of the structure of the evaluation estimation apparatus of the information processing system which concerns on Embodiment 2. FIG. It is a block diagram which shows an example of the structure of the terminal which concerns on Embodiment 2. FIG. It is explanatory drawing which shows an example of the record layout of a teacher data DB. It is explanatory drawing which shows an example of the record layout of the use history DB. It is a schematic diagram which shows the structure of the 1st physical characteristic information output model and the 2nd physical characteristic information output model. It is a flowchart which shows an example of the procedure of the generation processing of the 1st physical characteristic information output model by a control unit. It is a flowchart which shows an example of the procedure of the process of acquisition and transmission of physical property information by a control unit. It is a graph which shows the amount of amyloid β (hereinafter referred to as Aβ) before and after the addition of a compound. It is a display screen which shows Aβ38 / Aβ40 and Aβ42 / Aβ40 before the addition of a compound. It is a display screen which shows Aβ38 / Aβ40 and Aβ42 / Aβ40 after addition of a compound.

Hereinafter, the present invention will be described in detail with reference to the drawings showing the embodiments thereof.
(Outline of Embodiment)
The method for generating a learning model according to the embodiment is to induce differentiation from artificial pluripotent stem cells into predetermined cells, acquire cell information of the cells induced to differentiate into predetermined cells, and obtain the cell information and an analyzer. When teacher data including the physical property information of the cell based on the measurement using the above is acquired and the cell information of the cell whose differentiation is induced into a predetermined cell is input based on the teacher data, the physical property of the cell is input. Generate a learning model that outputs information.

In the present specification, the cells induced to differentiate into predetermined cells are neurons, astrosites, microglia, astrocytomas, oligodendrogliomas, glioblastomas, neuroperitoneal cells, corinagic, if they are of a nervous system. For nerve cells and musculoskeletal lines, osteoblasts, osteoclasts, cartilage cells, fibroblasts, myoblasts, polynuclear giant cells, and blood circulatory lines, red cells, B cells, T cells, NK In the case of cells, vascular endothelial cells, atrial muscle cells, ventricular muscle cells, base maker cells, digestive system, intestinal epithelial cells, exocrine cells, smooth muscle cells, connective tissue cells, layered columnar epithelial cells, superficial mucus cells, etc. Can be mentioned. By fusing or organizing the cells, the above-mentioned cell groups such as muscle fiber cells, beating cardiomyocytes, intestinal tissues, cerebral surface cortical cells, and nerve support cells can be included in predetermined cells.
Early differentiated cells undergo specific cell surface antigen / gene expression such as nerve stem cells, hematopoietic stem cells, intestinal stem cells, and mesenchymal stem cells through stages of inducing differentiation from pluripotent stem cells to endoderm / mesodermal / ectoderm. Examples of living stem cells exhibiting. Early differentiated cells also include cells that can be subcultured / proliferated under a specific environment (transformed cells) even though they do not have the specificity of living stem cells.

In addition to having an OCT-3 / 4 or NANOG of 10% or less in flow cytometry, the early differentiated cells have specific cell surface antigens (Nestin, PGDFR, FLK-1, SOX17, CXCR4) known as the living stem cells. , AFP, GFAP, CD34, CD45, CD117, CD73, CD90, CD105). For transformed cells, OCT-3 / 4 or NANOG is 10% or less, and Lin-28 or SSEA-4, CXCR4, CD90, ITG is 1% or more.

According to the above configuration, when the cell information is input based on the cell information of the cell induced to differentiate into a predetermined cell and the teacher data including the physical property information of the cell based on the measurement using an analyzer. Can generate a learning model that satisfactorily estimates cell physical property information. Therefore, the compound evaluation step can be omitted. It is also possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells based on the cell information before the addition of the compound. Such prediction results can be used not only for screening drugs but also for screening materials for preventive drugs and supplements.

The cell information to be input to the learning model includes, for example, pre-culture input information (first information) including basic cell information such as cell type, clinical information, and genetic information, and culture information (first information) including culture conditions. , Cell characteristic information including the cell image after culturing and the number of cells (input information after culturing, second information).

Examples of the physical property information include at least one type of information selected from the number of cells, absorption characteristics of light having a predetermined wavelength, chromosomal information, image information, action potential, medium information, and marker value. Specifically, the following items can be mentioned.

Cell measurement (total number of cells, number of living cells, number of dead cells, number of cell adhesion, cell migration rate)
Genetic information Information obtained by PCR (including rSSO and SSP), SNP, epigenome analysis Chromosome information Information obtained by chromosome analysis (G-Band, mFISH), CGH Surface antigen information, immunostaining Microscopic information (phase difference, fluorescence, confocal) Focus, electron, interatomic force)
Image information (cell morphology, colonization, cell migration, extensibility, motility, multilayer, pulsatile)
Information obtained by analysis of image data and time lapse Cell surface unevenness Pericellular viscoelasticity (measurement of extracellular matrix content such as collagen, elastin, laminin, fibronectin, and water content in the matrix)
Cell migration measurement Ultraviolet light-visible light-infrared-far infrared, excitation light, X-ray and other light measurements Magnetic measurement Active potential Medium information Information obtained by analysis of the concentration of compounds (markers, etc.) in the medium Protein Amount of (marker, etc.), amount of cytokine (marker, etc.), microRNA amino acid amount, measurement of heavy metal ELISA, LC-MS / MS, amino acid analysis, IPC-MS, Nano LC-TOF-MS, Western Brod, electrophoresis ( 2D, 3D), HLPC, microRNA analysis

The physical property information may be the amount of change over time. For example, artificial pluripotent stem cells derived from patients with Alzheimer's disease are induced to differentiate into nerve cells, and the amount of Aβ obtained from the cells changes with time.

The teacher data before learning is data in which physical property information is labeled with respect to cell information including the following pre-culture input information, culture information, and post-culture input information, for example.
1. 1. Input information before culturing (first information)
・ Basic cell information: Cell name (cell type): ○○, Number of passages: △ Passage, Gender: ♂ / ♀ (XaXi / XaXa), Cell donation age: ◎ years □ months, Human leukocyte antigen type (HLA-A) / HLA-B / HLA-C / HLA-DR / HLA-DQ / HLA-DP), for example, (A2: A24 / B38: B54 / Cw1: Cw6 / DR4DR53: DR8 / DQ8: DQ6 / DPw5: DPw5), etc. Is.
Examples of cell types include myocardial cells, pacemaker cells, ventricular cells, vascular endothelial cells, mesenchymal stem cells, fat cells, interventricular cells, fibroblasts, etc. in the case of heart lineage, and in the case of neuronal lineage, Examples include nerve cells, dopaminergic neurons, glutamate-operated neurons, Schwan cells, glial cells, astrocytes, purkinye cells, and the like.
-Clinical information: Different from healthy people and diseases (however, if it is a healthy person / disease, the reserve army and the potential reserve army can be classified.
In the case of disease origin, classification of disease symptoms (body parts): circulatory organs (myocardium, cardiovascular, etc.), bone / cartilage, muscles, nerves, blood, digestive organs (stomach, small intestine, large intestine), and disease classification: In addition to classifying by cancer, dementia (familial, hereditary, sporadic, etc.), it may be classified according to the degree of disease progression. On the other hand, there are cases in which the process itself, which is clinically asymptomatic but gradually becomes MCI (mild cognitive impairment) and eventually dementia, is regarded as dementia. In this case, the clinical information will be constantly updated, and the teacher data as an effort for ultra-early intervention / preventive response will be the behavior evaluation data of a specific healthy person and the possibility that this healthy person will develop the disease. By using induced pluripotent stem cells related to a certain disease, it is possible to develop medicines, supplement materials, medical devices, robot care, etc. for ultra-early intervention / prevention measures. For patients who have already developed the disease, the onset time, therapeutic drug, drug administration history, surgical history, overseas travel history, infection history, transplant history, and other clinical findings (severity or stage (MCI), independent walking, light-headedness, dizziness, etc.) Clinical information such as.
Examples of the disease name include ischemic heart disease and dilated cardiomyopathy when the tissue is the heart, and Alzheimer-type and Lewy body-type dementia when the tissue is the brain.
-Genetic information: Memorizes abnormalities of the XX gene located on chromosome XX △△. For example, the expression is increased or decreased due to a gene abnormality related to PRC1, DYRK1, RCAN1, OLIG1 / 2, APP, etc. at the full length of chromosome 21 or at a specific position (disease name: Down's syndrome).
Diseases related to this genetic abnormality are extracted from databases such as PubMed, DeCIPHER, GDBS, and OGDD (arranged and quantified in descending order of the number of patients among related diseases and in descending order of gene mutation and probability of disease onset).

2. Culture information (first information)
-Culture conditions: medium model number, volume, addition factor, addition amount, manufacturing company, sales company, transportation environment, reagent storage / control temperature, medium adjustment date and time, medium consumption (planned) time, medium creator, culture worker, work Date and time (work start time to work end time), (culture) work protocol, all work process numbers, operation program name of differentiation induction culture device or No. , ID, program operation date and time, implementer, exposure time in culture environment, medium exchange date and time, operation availability, etc.

3. 3. Post-culture input information (cell characteristic information: second information)
-Cell characteristics: Cell morphology image for each motion program, total number of cells by cell measuring instrument, number of viable cells, cell viability, cell adhesion rate, cell size (cytoplasm / cell nucleus: C / N ratio), input from the sensor Information for indirectly acquiring cell characteristics by non-destructive testing (for example, amino acid fluctuation information in culture solution, cell metabolite fluctuation information, pH fluctuation information, lactic acid level fluctuation information, glucose fluctuation information), cell surface antigen (FACS) ), Cell aging measurement, presence / absence of (immunostaining) antigen expression, aseptic test, mycoplasma denial test, endotoxin measurement, oxidative stress measurement, cell extension measurement (including neuroprotrusion extension measurement), cell migration test, etc.

4. Physical property information / marker value (biomarker / surrogate marker): Aβ amount (Alzheimer-type nerve cell) measured by the ELISA method from the medium supernatant with an evaluation device, Aβ amount inside the nerve cell, and other tissue cells with amyloid-like deposits. (Amyloidosis disease) Amount of amyloid protein, etc.
Measure proteins, cytokines, microRNA, amino acid amount, heavy metals by ELISA, HLPC, TOF-MS, LC-MS / MS, IPC-MS, amino acid analysis, western broth, electrophoresis (2D, 3D), etc., Gene Amount of change before and after culture of Chip, SNP, CGH, etc.
・ Potential measurement: QT prolongation, action potential (Ca ²⁺ , Na ⁺ , K ⁺ ) information, extracellular potential information ・ Cell measurement (total number of cells, number of living cells, number of dead cells, number of cell adhesion, cell migration rate)
-Genetic information Information obtained by PCR (including rSSO and SSP), SNP, epigenome analysis-Chromosome information Chromosome analysis (G-Band, mFISH), information obtained by CGH-Surface antigen information, immunostaining-Microscopic information (phase difference) , Fluorescence, confocal, electron, interatomic force)
-Image information (cell morphology, colonization, cell migration, extensibility, motility, stratification, pulsation)
・ Cell surface unevenness ・ Pericellular viscoelasticity (measurement of extracellular matrix content such as collagen, elastin, laminin, fibronectin, and water content in the matrix)
・ Measurement of cell migration ・ Measurement of ultraviolet light-visible light-infrared-far infrared light, excitation light, X-ray, etc. (evaluation of changes inside and outside cells)
-Magnetic measurement-Activity potential-Medium information (evaluation of intracellular primary and secondary metabolism, amino acid consumption in medium, pH, lactic acid concentration, microRNA, secretory protein (eg, Aβ, protease, NGF, etc.))
-Information on the device used: (IC chip model number, baseline and blank measurement results)
-Evaluation equipment information: manufacturing date, model number, validation record, administrator, maintenance history

The contents of pre-culture input information, culture information, post-culture input information and physical property information are not limited to the above cases. In addition, the item of the physical property information can be the item of the input information after culturing. For example, when the amount of Aβ in nerve cells obtained by inducing differentiation of induced pluripotent stem cells derived from patients with Alzheimer's disease is physical property information, basic cell information, clinical information, genetic information, and images before and after addition of the compound. The amount of Aβ in the culture supernatant (medium data), which is one of the items of the above physical property information, before and after the addition of the data and the compound is input to the learning model.

In the method for generating the learning model described above, the cell information of the cell induced to differentiate into a predetermined cell and the physical property information based on the actual measurement of the cell are acquired, and the learning model is obtained based on the cell information and the physical property information. You may relearn.
According to the above configuration, re-learning is performed based on the physical property information based on the actual measurement of the cells, so that the physical property information can be output more satisfactorily.

In the above-mentioned learning model generation method, compound information was acquired, the teacher data included compound information, and cell information and compound information of cells induced to differentiate into predetermined cells were input based on the teacher data. In some cases, a learning model that outputs physical property information may be generated.

Examples of compound information include low-molecular-weight systems, nucleic acid-based systems, skeletons, compound names, chemical formulas, low-molecular-weight systems such as drug codes, and nucleic acid systems. The compound is not limited to a synthetic product, and may be a natural product.
According to the above configuration, since the teacher data includes the compound information, the physical property information can be output satisfactorily in consideration of the structure of the compound and the like. In addition, by comparing with the positive control, it is possible to grasp the physical properties, that is, the accuracy of estimation of the evaluation of the compound for each production lot of differentiated cells.

In the method for generating the learning model described above, the physical property information may indicate the amount of change over time.
According to the above configuration, the effect of the compound over time can be obtained based on the change over time in the physical properties.

The evaluation estimation device according to one aspect of the present invention obtains cell information of cells induced to differentiate into predetermined cells from artificial pluripotent stem cells, and cell information of cells induced to differentiate into predetermined cells. When input, the cell information acquired by the acquisition unit is input to the learning model that outputs the physical property information of the cell based on the measurement using the analyzer, and the estimation unit estimates the physical property information. It is provided with an output unit that outputs physical property information.

According to the above configuration, since the cell information of the cell induced to differentiate into a predetermined cell is input to the learning model and the physical property information is acquired, the costly and complicated evaluation step of the compound can be omitted. It is also possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells based on the cell information before the addition of the compound.

In the evaluation estimation device described above, the acquisition unit acquires compound information, the learning model outputs physical property information when the compound information is input, and the estimation unit outputs the physical property information to the learning model by the acquisition unit. The acquired compound information may be input.

According to the above configuration, since the teacher data includes the compound information, the physical property information can be output satisfactorily in consideration of the structure of the compound and the like.

In the evaluation estimation device described above, the acquisition unit acquires the cell information of the cell before the compound is added, and the learning model inputs the cell information before the compound is added, and the first physical property information is obtained. Is a model that outputs the above, and the estimation unit may estimate the first physical property information.
According to the above configuration, it is possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells based on the cell information before the addition of the compound.

In the evaluation estimation device described above, the acquisition unit acquires the cell information of the cell after the compound is added, and the learning model has the second physical characteristics when the cell information after the compound is added is input. It is a model that outputs information, and the estimation unit may estimate the second physical property information.

The evaluation estimation device may include a calculation unit that calculates the physical property information of the compound based on the first physical property information and the second physical property information estimated by the estimation unit.

According to the above configuration, the effect / efficacy or toxicity / safety of the compound can be evaluated based on the difference between the second and first physical property information.

In the above-mentioned evaluation estimation device, the cell information is information including first information including cell type, clinical information, genetic information, and culture conditions, and second information including cell image and cell number after culturing. There may be.
According to the above configuration, cells can be specified in detail and physical property information can be output with high accuracy.

In the above-mentioned evaluation estimation device, the physical property information is at least one kind of information selected from the number of cells, light absorption characteristics of a predetermined wavelength, chromosomal information, image information, activity potential, medium information, and marker value. May be good.
According to the above configuration, it is possible to output and evaluate the physical property information corresponding to the desired evaluation item, and it is also possible to find a new surrogate marker by using the evaluation result.

The evaluation estimation device described above includes a determination unit for determining whether or not to add the compound based on a biomarker value, a surrogate marker value, or an image of the cell before adding the compound to the cell. You may.

According to the above configuration, it is possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells, and it is possible to stop the execution of unnecessary evaluation.

In the above-mentioned evaluation estimation device, the physical property information output by the output unit is displayed corresponding to each well of a plurality of microplates of the differentiation-inducing culture device that induces differentiation from induced pluripotent stem cells into predetermined cells. A display unit may be provided.

According to the above configuration, the effect / efficacy or toxicity / safety of the compound can be visually recognized by comparing the amount of change in physical properties before and after the addition of the compound for each well with that of the control.

The learning model according to one aspect of the present invention includes an input layer into which cell information of cells induced to differentiate from artificial pluripotent stem cells into predetermined cells is input, and physical property information of the cells based on measurement using an analyzer. An artificial pluripotent stem cell having an output layer for outputting the above, an intermediate layer in which parameters are learned based on cell information of cells induced to differentiate from artificial pluripotent stem cells into predetermined cells and physical property information of the cells. When the cell information of one cell whose differentiation is induced into a predetermined cell is input from the cell, the computer is made to function so as to output the physical property information of the cell from the output layer through the calculation by the intermediate layer.

According to the above configuration, when the cell information is input based on the teacher data including the cell information of the cell induced to differentiate into a predetermined cell and the physical property information of the cell based on the measurement using an analyzer. , A learning model that satisfactorily estimates cell physical property information is generated. Therefore, the compound evaluation step can be omitted. It is also possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells based on the cell information before the addition of the compound.

The computer program according to one aspect of the present invention acquires cell information of a cell that has been induced to differentiate into a predetermined cell from an artificial pluripotent stem cell, and inputs the cell information of the cell that has been induced to differentiate into a predetermined cell. , The computer is made to execute the process of inputting the acquired cell information into a learning model that outputs the physical property information of the cell based on the measurement using the analyzer and outputting the physical property information.

According to the above configuration, when the cell information is input based on the teacher data including the cell information of the cell induced to differentiate into a predetermined cell and the physical property information of the cell based on the measurement using an analyzer. , A learning model that satisfactorily estimates cell physical property information is generated. Therefore, the compound evaluation step can be omitted. It is also possible to determine whether or not the evaluation result can be obtained by adding the compound to the cells based on the cell information before the addition of the compound.

(Embodiment 1)
FIG. 1 is a schematic diagram showing an example of the configuration of the information processing system 10 according to the first embodiment. In the information processing system 10, the evaluation estimation device 1 of the information management company, the control device 21 of the plurality of differentiation-inducing culture devices (hereinafter referred to as culture devices) 2, the terminals 3 of each user, and the terminals 4 of the plurality of pharmaceutical companies. Is connected via a network N such as the Internet. The control device 21 and the terminal 3 are connected by LAN.

The culture device 2 includes a culture container such as a multi-well plate for culturing cells or a cell population, a mounting table on which the culture container is placed, a dispenser for dispensing liquid, a transfer arm for transporting the culture container, and a transfer arm. A culture device 28 (see FIG. 10) having an incubator for culturing cells, a cell preparation device 29, a cell analysis device 30, and the like are housed in a case 20. Under the control of the control device 21, the culture device 28 is configured to produce the desired differentiated cells from the induced pluripotent stem cells.

The evaluation estimation device 1 provides the culture device 2 with a production protocol for inducing differentiation of induced pluripotent stem cells to produce the desired differentiated cells and a production program for producing the differentiated cells based on the production protocol. Remember. The evaluation estimation device 1 updates or newly installs the manufacturing protocol, updates or newly sets the manufacturing program when a predetermined condition is satisfied, outputs these to the culture device 2, and stores them in the culture device 2 by the control unit 22 of the culture device 2. Let me. The culture device 2 produces differentiated cells based on the production program.

FIG. 2 is a block diagram showing an example of the configuration of the evaluation estimation device 1. The evaluation estimation device 1 includes a control unit 11, a main storage unit 12, a communication unit 13, and an auxiliary storage unit 14 that control the entire device. The evaluation estimation device 1 can be composed of one or a plurality of servers. A plurality of evaluation estimation devices 1 may be used for distributed processing.
The control unit 11 can be composed of a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like. The control unit 11 may include a GPU (Graphics Processing Unit). Moreover, you may use a quantum computer.
The control unit 11 has an acquisition unit 111, a selection unit 112, an estimation unit 113, and an output unit 114 as functional units. Each functional unit of the control unit 11 is a program module, and functions as follows by executing an information processing program (hereinafter, referred to as a program) 141 described later.
The acquisition unit 111 acquires cell information of differentiated cells from the control device 21 or the user's terminal 3.
The selection unit 112 reads the physical characteristic information output model from the physical characteristic information output model DB 147, and the estimation unit 113 selects the physical characteristic information output model.
The estimation unit 113 inputs cell information into the physical characteristic information output model, acquires the physical characteristic information output by the physical characteristic information output model, and estimates the physical characteristic information.
The output unit 114 outputs the estimated physical property information to the terminal 3 or the like by the communication unit 13.

The main storage unit 12 is a temporary storage area for SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), flash memory, etc., and temporarily stores data necessary for the control unit 11 to execute arithmetic processing. Remember.

The communication unit 13 has a function of communicating with the culture device 2, the terminals 3 and 4, via the network N, and can transmit and receive necessary information. Specifically, the communication unit 13 receives the cell information transmitted by the user's terminal 3. The communication unit 13 transmits the physical property information to the terminal 3 or the terminal 4.

The auxiliary storage unit 14 is a large-capacity memory, a hard disk, or the like, and includes a program required for the control unit 11 to execute processing, a program 141 for performing information processing described later, a manufacturing protocol DB 142, a manufacturing program DB 143, and an evaluation protocol DB 144. The compound information DB 145, the teacher data DB 146, the physical property information output model DB 147, and the usage history DB 148 are stored.

The program 141 stored in the auxiliary storage unit 14 may be provided by the recording medium 150 in which the program 141 is readablely recorded. The recording medium 150 is, for example, a portable memory such as a USB memory, an SD card, a micro SD card, or a compact flash (registered trademark). The program 141 recorded on the recording medium 150 is read from the recording medium 150 using a reading device (not shown) and installed in the auxiliary storage unit 14. Further, the program 141 stored in the auxiliary storage unit 14 may be provided by communication via the communication unit 13.

FIG. 3 is an explanatory diagram showing an example of the record layout of the manufacturing protocol DB 142. The production protocol DB 142 has an ID No. 1 for each incubator 2. It stores columns, cell type sequences, differentiation-inducing factor sequences, and data file sequences. In FIG. 3, the ID No. The manufacturing protocol of the culture apparatus 2 of 1 is shown. The ID column is an ID No. for identifying each manufacturing protocol. I remember. The cell type sequence remembers the cell type. Examples of the cell type include the above-mentioned cells of the heart lineage, cells of the nerve lineage, and cells of other tissues. The differentiation-inducing factor sequence stores the differentiation-inducing factors used for the production of differentiated cells. Examples of the differentiation-inducing factor include Activin, BPM4, Dkk-1, CHIR99021 in the case of cardiomyocytes, BMP inhibitor, Wnt inhibitor and the like in the case of nerve cells. The data file column stores the data file of the manufacturing protocol.

The manufacturing program DB 143 stores a manufacturing program corresponding to the manufacturing protocol.
FIG. 4 is an explanatory diagram showing an example of the record layout of the manufacturing program DB 143. The production program DB 143 has an ID No. 1 for each incubator 2. Column, manufacturing protocol No. Stores columns and data file columns. In FIG. 4, the ID No. The production program of the culture apparatus 2 of 1 is shown. ID No. The column is the ID No. for identifying each manufacturing program. I remember. Manufacturing protocol No. The columns indicate the corresponding manufacturing protocol No. I remember. The data file column stores the data file of the manufacturing program.

FIG. 5 is an explanatory diagram showing an example of the record layout of the evaluation protocol DB 144. The evaluation protocol DB 144 has an ID No. 1 for each incubator 2. Stores a column, a disease name column, a cell type column, a disease model column, an evaluation item column, and a data file column. The control unit 11 newly stores or updates the evaluation protocol in the evaluation protocol DB 144 when it is acquired by the article search server, when it is acquired from terminals 3, 4, or the like. In FIG. 5, the ID No. The evaluation protocol of the culture apparatus 2 of 1 is shown. ID No. The column is the ID No. for identifying each protocol. I remember. The disease name column remembers the disease name. Disease names include the above-mentioned heart-related diseases, brain-related diseases, and diseases of other organs. The cell type sequence remembers the cell type. Examples of the cell type include the above-mentioned cells of the heart lineage, nerve cells of the nervous system, and cell types of other lineages. The disease model column remembers the disease model. Examples of the disease model include humans, humans (derived from patients), monkeys, and the like. The evaluation item column stores the evaluation items. Evaluation items include heart-related calcium concentration, extracellular potential, activity potential, electrocardiogram, etc .; brain-related Aβ38 amount, Aβ40 amount, Aβ42 amount, Aβ38 / Aβ40 and Aβ42 / Aβ40, Aβ25, Aβ38, Aβ40, per cell number. The ratio of (Aβ38 / Aβ40) to the total amount of Aβ42 and the like (Aβ38 / Aβ40) / total Aβ amount, and the ratio of (Aβ42 / Aβ40) to the total amount of Aβ (Aβ42 / Aβ40) / total Aβ amount. , Neuroprotrusion length, brain wave, tau amount, phosphorylated tau amount, θ wave amount, aggregated protein, or sugar, lipid, etc .; Evaluation items of other organs can be mentioned. The data file stores the contents of each evaluation protocol.

FIG. 6 is an explanatory diagram showing an example of the record layout of the compound information DB 145.
The compound information DB 145 is described in ID No. It stores columns, sequences, skeleton sequences, drug code sequences, and compound name sequences. The control unit 11 newly stores or updates the compound information in the compound information DB 145 when it is acquired by the article search server, when it is acquired from terminals 3, 4, or the like. ID No. The column is the ID No. for identifying each compound. I remember. The series stores systems such as small molecule systems and nucleic acid systems. The skeleton sequence remembers the skeleton. The skeleton is a partial structure common to multiple molecules, for example, carotenoids (β-carotene, violaxanthin, astaxanthin), alkaloids, isoprenoids, saccharides and derivatives or polymers thereof, vitamins, amino acids and their polymerization. Substances, nucleic acids and their polymers, quinones / quinols (coenzyme Q10 (ubiquinone, ubiquinol)), terpenes / terpenoids, lipids (monoglycerides, diglycerides, triglycerides, phospholipids, sphingolipids), steroids, etc. Be done. The drug code string stores a code such as a drug price standard listed drug code or an individual drug code. The compound name sequence stores the compound name or structure (chemical formula). As the compound, a composition containing a plurality of components or a composition containing a component having an unknown structure, such as an extract derived from a natural product, a fungus, or a secretion produced by a cell line, may be used.

FIG. 7 is an explanatory diagram showing an example of the record layout of the teacher data DB 146. The teacher data is acquired from a plurality of terminals 3 and stored.
The teacher data DB 146 has an ID No. A column, a cell basic information column, a clinical information column, a genetic information column, a culture condition column, a compound information column, an image data column, a marker value column, a cell characteristic information column, and a physical property information column. I remember. ID No. The column is the ID No. for identifying each teacher data. I remember. The cell basal information sequence stores the above-mentioned cell basal information, and ID No. Line 1 remembers "nerve cells" as the cell type. Other basic cell information such as the passage number, sex, cell donation age, and human leukocyte antigen type described above may be further stored. The clinical information column stores the above-mentioned clinical information, and ID No. Line 1 remembers "Alzheimer's disease" as the name of the disease. Other clinical information described above may be stored. The genetic information sequence stores the above-mentioned genetic information such as chromosome number and position, and ID No. In line 1, "abnormality of chromosome 21 a gene" is memorized. Other genetic information described above may be stored. The culture condition column stores the above-mentioned culture conditions, and ID No. The line 1 remembers that it is the manufacturing protocol of "ID No. 5". The other culture conditions described above may be stored. The compound information string stores the above-mentioned compound information. The image data string stores the image data before and after the addition of the compound. The marker value sequence stores the above-mentioned biomarker value or surrogate marker value. ID No. In line 1, the amount of Aβ measured by the ELISA method or the like using the supernatant of the medium before and after the addition of the compound is stored. The cell characteristic information column stores the above-mentioned cell characteristic information other than the image data and the marker value. Hereinafter, the cell characteristic information indicates the above-mentioned cell characteristic information other than the image data and the marker value. The physical characteristic information string stores the above-mentioned physical characteristic information actually measured. ID No. 1, No. In row 2, the Aβ38 amount, Aβ40 amount, and Aβ42 amount of the differentiated cells measured by the ELISA method or the like are stored corresponding to the cases where the evaluation items are Aβ38 amount, Aβ40 amount, and Aβ42 amount.

FIG. 8 is an explanatory diagram showing an example of the record layout of the physical characteristic information output model DB 147. The physical property information output model DB147 has an ID No. The column, the evaluation item column, and the physical characteristic information output model column are stored. ID No. The column is the ID No. for identifying each physical characteristic information output model. I remember. The evaluation item column stores the above-mentioned evaluation items. The physical characteristic information output model column stores the physical characteristic information output model corresponding to the evaluation item. ID No. In the case of 2, the evaluation items are Aβ38 amount, Aβ40 amount, and Aβ42 amount, and the physical characteristic information output model that outputs Aβ38 amount, Aβ40 amount, and Aβ42 amount as evaluation items is stored. In the present embodiment, the case of generating the physical characteristic information output model corresponding to the evaluation item has been described, but the present invention is not limited to this. For example, a physical characteristic information output model may be generated for each cell such as a nerve cell or a cardiomyocyte.

FIG. 9 is an explanatory diagram showing an example of the record layout of the usage history DB 148. The usage history is stored for each incubator 2. The usage history DB 148 has an ID No. Columns, cell basic information columns, clinical information columns, genetic information columns, culture condition columns, compound information columns, image data columns, marker value columns, cell characteristic information columns, date and time columns, and physical properties. The information sequence and the measured value sequence are stored. ID No. The columns, cell basic information sequence, clinical information sequence, genetic information sequence, culture condition column, compound information sequence, image data sequence, marker value sequence, and cell characteristic information column are the ID No. of the teacher data DB146. It has the same structure as the column, cell basic information sequence, clinical information sequence, genetic information sequence, culture condition sequence, compound information sequence, image data sequence, marker value sequence, and cell characteristic information sequence. The date and time column stores the date and time of cell production. The physical characteristic information string stores the physical characteristic information obtained by the physical characteristic information output model. The measured value sequence stores the measured values obtained by using the analyzer. The physical characteristic information output model can be relearned as described later by using the physical characteristic information output by the physical characteristic information output model and the measured values.

FIG. 10 is a block diagram showing an example of the configuration of the culture apparatus 2. The culture device 2 includes a control device 21, a culture device 28, a cell preparation device 29, and a cell analysis device 30.
The control device 21 includes a control unit 22, a main storage unit 23, a communication unit 24, an operation unit 25, a display panel 26, and an auxiliary storage unit 27 that control the entire culture device 2.

The control device 21 can be composed of, for example, a desktop computer, a notebook personal computer, a tablet, a smartphone, a clock-type device, or the like.
The control unit 22 can be composed of a CPU, a ROM, a RAM, and the like. The control unit 22 may be configured to include a GPU.

The main storage unit 23 is a temporary storage area for SRAM, DRAM, flash memory, etc., and temporarily stores data necessary for the control unit 22 to execute arithmetic processing.
The communication unit 24 has a function of communicating with the evaluation estimation device 1 via the network N, and can transmit and receive required information.

The operation unit 25 is composed of, for example, a hardware keyboard, a mouse, a touch panel, etc., and can operate icons and the like displayed on the display panel 26, input characters and the like, and the like.

The display panel 26 can be composed of a liquid crystal panel, an organic EL (Electro Luminescence) display panel, or the like. The control unit 22 controls to display the required information on the display panel 26.

The auxiliary storage unit 27 is a large-capacity memory, a hard disk, or the like, and stores a program 271 other than the manufacturing program, which is necessary for the control unit 22 to execute the process. Further, the auxiliary storage unit 27 stores the manufacturing protocol DB 272, the manufacturing program DB 273, and the evaluation protocol DB 274 that store the manufacturing protocol and the manufacturing program acquired from the evaluation estimation device 1, respectively. The manufacturing protocol DB 272, the manufacturing program DB 273, and the evaluation protocol DB 274 have the same configurations as the manufacturing protocol DB 142, the manufacturing program DB 143, and the evaluation protocol DB 144.

In the culture device 28, the control unit 22 reads the program 271 from the auxiliary storage unit 27, induces differentiation of induced pluripotent stem cells, and produces the desired differentiated cells.
The cell preparation device 29 has, for example, a built-in centrifuge, and for example, the cell suspension can be centrifuged in the process of producing differentiated cells to separate the supernatant from the cells, so that the cell group and the culture supernatant are collected. Used as a device.
Examples of the cell analysis device 30 include a cell counter that counts the number of differentiated cells, a microscope for acquiring cell morphology, a surface antigen quantitative analysis device, and the like.

FIG. 11 is a block diagram showing an example of the configuration of the terminal 3. The terminal 3 includes a control unit 31, a main storage unit 32, a communication unit 33, an operation unit 34, a display panel 35, and an auxiliary storage unit 36 that control the entire device. The terminal 3 can be composed of, for example, a desktop computer, a notebook personal computer, a tablet, a smartphone, a clock-type device, or the like. The control unit 31 can be composed of a CPU, a ROM, a RAM, and the like. The main storage unit 32, the communication unit 33, the operation unit 34, and the display panel 35 have the same configurations as the main storage unit 23, the communication unit 24, the operation unit 25, and the display panel 26 of the control device 21.
The auxiliary storage unit 36 is a large-capacity memory, a hard disk, or the like, and stores a program 361 required for the control unit 31 to execute a process. In addition, the auxiliary storage unit 36 stores the manufacturing protocol DB 362 and the evaluation protocol DB 363 that each store the manufacturing protocol and the evaluation protocol acquired from the evaluation estimation device 1, the compound information DB 364, and the usage history information DB 365. The production protocol DB 362, the evaluation protocol DB 363, the compound information DB 364, and the usage history DB 365 have the same configurations as the production protocol DB 142, the evaluation protocol DB 144, the compound information DB 145, and the usage history DB 148.
The terminal 3 acquires the analysis result from an analyzer 39 such as an absorptiometer. The case is not limited to the case where the terminal 3 and the analyzer 39 are connected.
The terminal 4 has the same configuration as the terminal 3, and the control unit 41 controls each unit.

FIG. 12 is a flowchart showing a procedure for inducing differentiation from induced pluripotent stem cells to produce nerve cells.
Obtain artificial pluripotent stem cells derived from somatic cells of a patient having a predetermined disease (S101).
Induced pluripotent stem cells are subcultured (S102).
The medium is exchanged (S103).
Nerve progenitor cells are cultured (S104).
The medium is exchanged (S105).
Nerve cells are cultured (S106).
A compound as a drug is added (S107).
Obtain the physical characteristics before and after the addition of the compound (S108).
The effect and toxicity of the compound are evaluated based on the amount of change in physical properties before and after the addition of the compound (S109).
The same applies to the case of producing other cells such as cardiomyocytes by inducing differentiation from induced pluripotent stem cells.
In the present embodiment, the control unit 11 acquires the physical property information of the compound by the physical property information output model.

FIG. 13 is a schematic diagram showing the configuration of the physical characteristic information output model.
The physical property information output model is a learning model that is expected to be used as a program module that is a part of artificial intelligence software, and a multi-layer neural network (deep learning) can be used. For example, a convolutional neural network (Convolutional Neural Network) can be used, but a recurrent neural network (RNN) may be used as deep learning. In addition, other machine learning such as decision trees, random forests, and support vector machines may be used. The control unit 11 receives commands from the physical characteristic information output model for basic cell information, clinical information, culture conditions, compound information, image data, marker values, and cell characteristic information input to the input layer of the physical characteristic information output model. It operates to perform calculations and output physical property information. The cell information is not limited to the case where all of the basic cell information, clinical information, culture conditions, compound information, image data, marker values, and cell characteristic information are input.

The control unit 11 adds image data before and after the addition of the compound, basic cell information, clinical information, genetic information, culture conditions, compound information, marker values before and after the addition of the compound, and cell characteristic information to the physical property information output model. And enter. Image data is input to a plurality of convolution layers and pooling layers and compressed. The image data of the pooling layer is combined with basic cell information, clinical information, genetic information, culture conditions, compound information, marker values, and cell characteristic information, and is input to and compressed in a plurality of convolution layers and pooling layers. Output physical property information. The output layer has a plurality of neurons that output physical property information. For example, the evaluation items are Aβ38 amount, Aβ40 amount, and Aβ42 amount. In the physical characteristic information output model of No. 2, the physical characteristic information output from the neuron is a combination of Aβ38 amount, Aβ40 amount, and Aβ42 amount and a probability value of each combination when the softmax function is used.
For example, Aβ38 amount 10 pmol / L, Aβ40 amount 600 pmol / L, Aβ42 amount 180 pmol / L ... 0.92 (probability value)
Aβ38 amount 9 pmol / L, Aβ40 amount 650 pmol / L, Aβ42 amount 150 pmol / L ... 0.04
Output as ...
The physical characteristic information output model is not limited to the case where the image data is compressed and then combined with other cell information.

FIG. 14 is a flowchart showing an example of a procedure for generating a physical characteristic information output model by the control unit 11.
The control unit 11 reads the teacher data from the teacher data DB 146 (step S111). The control unit 11 labels the physical property information on the cell basic information, clinical information, genetic information, culture conditions, compound information, image data, marker value and cell characteristic information of each row of the teacher data DB 146, and acquires the teacher data. ..
The control unit 11 uses the teacher data to output physical property information when inputting basic cell information, clinical information, genetic information, culture conditions, compound information, image data, marker values, and cell characteristic information. Is generated (step S112). The control unit 11 stores the generated physical characteristic information output model in the physical characteristic information output model DB 147, and ends a series of processing.
The physical property information output model may be generated by the control unit 22 or the control unit 31.

FIG. 15 is a flowchart showing an example of a procedure for processing the acquisition and transmission of physical property information by the control unit 11.
The control unit 31 transmits basic cell information, clinical information, and genetic information to the evaluation estimation device 1 (S301).
The control unit 11 receives the basic cell information, clinical information, and genetic information and stores them in the usage history DB 148 (S121).
The control unit 31 transmits the culture conditions to the evaluation estimation device 1 (S302).
The control unit 11 receives the culture conditions and stores them in the usage history DB 148 (S122).
The control unit 31 transmits the compound information to the evaluation estimation device 1 (S303).
The control unit 11 receives the compound information and stores it in the usage history DB 148 (S123).

The control unit 31 transmits the image data before addition to the evaluation estimation device 1 (S304).
The control unit 11 receives the image data before addition and stores it in the usage history DB 148 (S124). The control unit 11 may acquire image data directly from the microscope as the cell analysis device 30 of the culture device 2.
The control unit 31 transmits the marker value and the cell characteristic information before addition to the evaluation estimation device 1 (S305). The control unit 31 transmits, for example, the amount of Aβ in the medium.
The control unit 11 receives the marker value and the cell characteristic information before addition and stores them in the usage history DB 148 (S125).
The control unit 31 transmits the added image data to the evaluation estimation device 1 (S306).
The control unit 11 receives the image data after the addition and stores it in the usage history DB 148 (S126).
The control unit 31 transmits the marker value and the cell characteristic information after the addition to the evaluation estimation device 1 (S307).
The control unit 11 receives the marker value and the cell characteristic information after the addition and stores them in the usage history DB 148 (S127).

The control unit 11 reads out the physical characteristic information output model DB 147 and selects the physical characteristic information output model based on the evaluation items (S128).
The control unit 11 inputs cell information and compound information to the physical characteristic information output model (S129).
The control unit 11 acquires the physical property information of the physical characteristic information output model (S130).
The control unit 11 stores the physical property information in the usage history DB 148 (S131).
The control unit 11 transmits the physical property information to the terminal 3 (S132), and ends the process.
The control unit 31 receives the physical property information (S308) and stores it in the usage history DB 365.

The control unit 11 can relearn the physical characteristic information output model based on the physical characteristic information output by the physical characteristic information output model and the measured value thereof so that the reliability of the estimation result of the physical characteristic information is improved.
FIG. 16 is a flowchart showing a procedure of re-learning processing of the physical characteristic information output model by the control unit 11.
The control unit 11 reads the usage history DB 148 (S141).
The control unit 11 has an ID No. 1 having physical property information and an actually measured value. The usage history of is acquired as teacher data (142).
The control unit 11 uses the teacher data to output physical property information when inputting basic cell information, clinical information, genetic information, culture conditions, compound information, image data, marker values, and cell characteristic information. Is generated (step S143). The control unit 11 stores the generated physical characteristic information output model in the physical characteristic information output model DB 147, and ends a series of processing.

According to the present embodiment, since the cell information of the differentiated cells is input to the physical characteristic information output model and the physical characteristic information is acquired, the costly and complicated evaluation step of the compound can be omitted. The accuracy of compound evaluation can also be grasped by comparing positive subjects at the time of cell production (for each lot).
Re-learning improves the accuracy of the output physical property information.
In re-learning, statistical processing such as normal distribution, significance level, confidence interval, mean and variance estimation and test, setting of reference value, etc. using each teacher data (condition, information, image data, marker value, etc.) as a population. Can be performed to sort the priorities for the results of each teacher data.

(Embodiment 2)
FIG. 17 is a block diagram showing an example of the configuration of the evaluation estimation device 1 of the information processing system 10 according to the second embodiment, and FIG. 18 is a block diagram showing an example of the configuration of the terminal 3 according to the second embodiment. In FIGS. 17 and 18, the same parts as those in FIGS. 2 and 11 are designated by the same reference numerals, and abbreviated description thereof will be omitted.
The auxiliary storage unit 14 of the evaluation estimation device 1 has a first physical characteristic information output model DB 149 and a second physical characteristic information output model DB 151 that store the first physical characteristic information output model and the second physical characteristic information output model information, respectively. The first physical characteristic information output model outputs the first physical characteristic information when cell information including image data and medium data before adding the compound is input. The second physical characteristic information output model outputs the second physical characteristic information when the cell information including the image data and the medium data after the addition of the compound is input.
The terminal 3 installs the first physical characteristic information output model and the second physical characteristic information output model of the evaluation estimation device 1, and corresponds to the first physical characteristic information output model DB 149 and the second physical characteristic information output model DB 151 in the auxiliary storage unit 36. The first physical characteristic information output model DB366 and the second physical characteristic information output model DB367 are stored. The control unit 11 or 31 acquires the amount of change in the physical characteristics before and after the addition of the compound by the difference between the first physical characteristic information and the second physical characteristic information acquired by the first physical characteristic information output model and the second physical characteristic information output model.

FIG. 19 is an explanatory diagram showing an example of the record layout of the teacher data DB 146. The usage history is stored for each incubator 2. The teacher data DB 146 has an ID No. A column, a cell basic information column, a clinical information column, a genetic information column, a culture condition column, a compound information column, an image data column, a marker value column, a cell characteristic information column, and a first physical property information column. , The second physical property information string, and the difference information string are stored. ID No. The columns, cell basic information sequence, clinical information sequence, genetic information sequence, culture condition column, compound information sequence, image data sequence, marker value sequence, and cell characteristic information sequence are the ID Nos. Of the teacher data DB 146 according to the first embodiment. It has the same structure as the column, cell basic information sequence, clinical information sequence, genetic information sequence, culture condition sequence, compound information sequence, image data sequence, marker value sequence, and cell characteristic information sequence. The first physical characteristic information sequence stores the first physical characteristic information obtained by using the analyzer before the addition of the compound. The second physical characteristic information sequence stores the second physical characteristic information obtained by using the analyzer after the addition of the compound. Specifically, ID No. 1, No. In the case of row 2, the first physical property information is Aβ38 / Aβ40 and Aβ42 / Aβ40 per cell number of differentiated cells measured before the addition of the compound, and the second physical property information is per cell number of differentiated cells measured after the addition of the compound. Aβ38 / Aβ40 and Aβ42 / Aβ40. The first physical property information and the second physical property information may be the above-mentioned (Aβ38 / Aβ40) / total Aβ amount and (Aβ42 / Aβ40) / total Aβ amount. When the image data is not input as the cell information, the first physical property information and the second physical property information may be image data of differentiated cells.

FIG. 20 is an explanatory diagram showing an example of the record layout of the usage history DB 148. The usage history is stored for each incubator 2. The usage history DB 148 has an ID No. A column, a cell basic information column, a clinical information column, a genetic information column, a culture condition column, a compound information column, an image data column, a marker value column, a cell characteristic information column, a date and time column, and the first. It stores one physical property information string, a second physical property information string, a difference information string, and an actually measured value string. ID No. The columns, cell basic information sequence, clinical information column, genetic information column, culture condition column, compound information column, image data column, marker value column, cell characteristic information column, and date and time column are usage history DB148 according to the first embodiment. ID No. It has the same structure as the column, cell basic information column, clinical information column, genetic information column, culture condition column, compound information column, image data column, marker value column, cell characteristic information column, and date and time column. The first physical characteristic information string stores the first physical characteristic information output by the first physical characteristic information output model. The second physical characteristic information string stores the second physical characteristic information output by the second physical characteristic information output model. The difference information column stores the difference between the first physical property information and the second physical property information. The actually measured value sequence stores the actually measured values of the first and second physical property information obtained by using the analyzer, and the difference information of the actually measured first and second physical property information. ID No. In the case of row 2, the first physical characteristic information is Aβ38 / Aβ40 and Aβ42 / Aβ40 per number of differentiated cells obtained by the first physical characteristic information output model before the addition of the compound, and the second physical characteristic information is after the addition of the compound. It is Aβ38 / Aβ40 and Aβ42 / Aβ40 per the number of differentiated cells acquired by the second physical characteristic information output model. ID No. In line 2, Aβ38 / Aβ40 and Aβ42 / Aβ40 obtained by actual measurement values before and after the addition of the compound are also stored in the same differentiated cells.

FIG. 21 is a schematic diagram showing the configurations of the first physical characteristic information output model and the second physical characteristic information output model.
The control unit 11 inputs the image data before the addition of the compound, the basic cell information, the clinical information, the genetic information, the culture conditions, the marker value and the cell characteristic information before the addition of the compound into the first physical property information output model. To do. Image data is input to a plurality of convolution layers and pooling layers and compressed. The image data of the pooling layer is combined with basic cell information, clinical information, genetic information, culture conditions, compound information, marker values, and cell characteristic information, and is input to and compressed in a plurality of convolution layers and pooling layers. The first physical property information is output. The output layer has a plurality of neurons that output physical property information. In the first physical property information output from the neuron, the evaluation items are Aβ38 / Aβ40 and Aβ42 / Aβ40 per cell number.
When the softmax function is used, it is a combination of each value of Aβ38 / Aβ40 and Aβ42 / Aβ40 and a probability value of each combination.
For example, Aβ38 / Aβ40 0.017 per cell number, Aβ42 / Aβ40 0.30 ... 0.94 per cell number.
Aβ38 / Aβ40 0.019 per cell number, Aβ40 / Aβ40 0.35 ... 0.03 per cell number
Output as ...
The control unit 11 acquires 0.017 as Aβ38 / Aβ40 and 0.30 as Aβ42 / Aβ40 based on the combination having the highest probability value.

When the first physical property information is (Aβ38 / Aβ40) / total Aβ amount and (Aβ42 / Aβ40) / total Aβ amount
For example, (Aβ38 / Aβ40) / total Aβ amount 2.1 × 10 ^-5 , no Aβ aggregation, (Aβ42 / Aβ40) / total Aβ amount 3.8 × 10 ^{-4 with} Aβ aggregation,… 0.92
(Aβ38 / Aβ40) / total Aβ amount 1.9 × 10 ^-5 without Aβ aggregation, (Aβ42 / Aβ40) / total Aβ amount 3.4 × 10 ^-4 , with Aβ aggregation… 0.02
Output as ...

The control unit 11 adds image data after the addition of the compound, basic cell information, clinical information, genetic information, culture conditions, compound information, marker value and cell characteristic information after the addition of the compound to the second physical property information output model. And enter. Image data is input to a plurality of convolution layers and pooling layers and compressed. The image data of the pooling layer is combined with basic cell information, clinical information, genetic information, culture conditions, compound information, marker values, and cell characteristic information, and is input to and compressed in a plurality of convolution layers and pooling layers. The second physical property information is output. The output layer has a plurality of neurons that output the second physical characteristic information. The evaluation items of the second physical property information output from the neuron are the amount of Aβ and image data, and when the softmax function is used, the amount of Aβ, the Aβ cohesiveness, and the probability value thereof.
For example, Aβ38 / Aβ40 0.010 per cell number, Aβ42 / Aβ40 0.10 0.93 per cell number.
Aβ38 / Aβ40 0.016 per cell number, Aβ42 / Aβ40 0.26 0.04 per cell number
Output as ... The control unit 11 acquires 0.010 as Aβ38 / Aβ40 and 0.10 as Aβ42 / Aβ40.

FIG. 22 is a flowchart showing an example of a procedure for generating the first physical characteristic information output model by the control unit 11. The control unit 11 generates a first physical characteristic information output model for each evaluation item.
The control unit 11 reads the teacher data according to the evaluation item from the teacher data DB 146 (step S151). The control unit 11 labels the first physical property information with respect to the cell basic information, clinical information, genetic information, culture conditions, image data before addition, marker value before addition, and cell characteristic information of each row of the teacher data DB 146. , Get teacher data.
When the control unit 11 inputs basic cell information, clinical information, genetic information, culture conditions, compound information, image data before addition, marker value before addition, and cell characteristic information using teacher data, the first physical property information Is generated (step S152). The control unit 11 stores the generated first physical characteristic information output model in the first physical characteristic information output DB 149, and ends a series of processing.
The control unit 11 also generates the second physical characteristic information output model in the same manner and stores it in the second physical characteristic information output DB 151.

The terminal 3 installs the first physical characteristic information output DB 149 and the second physical characteristic information output DB 151 of the evaluation estimation device 1, and stores them in the auxiliary storage unit 36 as the first physical characteristic information output DB 366 and the second physical characteristic information output DB 367. Hereinafter, a case where the terminal 3 estimates the first physical property information and the second physical property information will be described. In this case, the control unit 31 functions as an acquisition unit, a selection unit, an estimation unit, and an output unit.

FIG. 23 is a flowchart showing an example of a procedure for processing the acquisition and transmission of physical property information by the control unit 31.
The control unit 31 stores basic cell information, clinical information, and genetic information in the usage history DB 365 (S311).
The control unit 31 stores the culture conditions in the usage history DB 365 (S312).
The control unit 31 stores the compound information in the usage history DB 365 (S313).
The control unit 31 acquires the image data before addition and stores it in the usage history DB 365 (S314).
The control unit 31 acquires the marker value and the cell characteristic information before the addition and stores them in the usage history DB 365 (S315).
The control unit 31 selects the first physical characteristic information output model from the first physical characteristic information output model DB366 according to the evaluation item (S316).
The control unit 31 inputs basic cell information, clinical information, genetic information, culture conditions, image data before addition, marker values, and cell characteristic information into the first physical characteristic information output model (S317).
The control unit 31 acquires the first physical property information (S318).
The control unit 31 determines whether or not to add the compound based on the first physical property information (S319). The control unit 31 determines whether or not the compound can be added to obtain physical property information based on, for example, the number of cells or the survival rate.
When the control unit 31 determines that the compound is to be added based on the first physical property information (S319: YES), the control unit 31 outputs an instruction for adding the compound to the display panel 35 or the like (S320).
The control unit 31 acquires the image data after the addition and stores it in the usage history DB 365 (S321).
The control unit 31 acquires the marker value and the cell characteristic information after the addition and stores them in the usage history DB 365 (S322).
The control unit 31 selects the second physical characteristic information output model from the second physical characteristic information output model DB367 according to the evaluation item (S323).
The control unit 31 inputs basic cell information, clinical information, genetic information, culture conditions, compound information, image data after addition, marker values and cell characteristic information into the second physical property information output model (S324).
The control unit 31 acquires the second physical property information (S325).
The control unit 31 acquires the difference information based on the first physical property information and the second physical property information (S326). When the first physical property information and the second physical property information are Aβ42 / Aβ40 per number of cells before and after addition, the difference between Aβ42 / Aβ40 before and after addition is calculated. As described with reference to FIG. 21, the control unit 31 acquires 0.017 as Aβ38 / Aβ40 of the first physical characteristic information, 0.30 as Aβ42 / Aβ40, and 0.010 as Aβ38 / Aβ40 of the second physical characteristic information. When 0.10 is acquired as Aβ42 / Aβ40, the difference information (first physical characteristic information-2nd physical characteristic information) is 0.007 for Aβ38 / Aβ40 and 0.20 for Aβ42 / Aβ40.
The control unit 31 stores the first physical characteristic information, the second physical characteristic information, and the difference information in the usage history DB 365 (S327).
The control unit 31 transmits the first physical property information, the second physical characteristic information, and the difference information to the evaluation estimation device 1 and the terminal 4 (S328), and ends the process.

FIG. 24 is a graph showing Aβ42 / Aβ40 before and after the addition of the compound. As shown in FIG. 24, the addition of a predetermined compound reduces Aβ42 / Aβ40.
In the present embodiment, the control unit 31 displays Aβ38 / before the addition of the compound as the first physical characteristic information acquired by the first physical characteristic information output model corresponding to each well of the microplate on the display panel 35. Aβ40 and Aβ42 / Aβ40 are displayed. Then, the control unit 31 displays Aβ38 / Aβ40 and Aβ42 / Aβ40 after the addition of the compound as the second physical characteristic information acquired by the second physical characteristic information output model corresponding to each well.

FIG. 25 is a display screen showing Aβ38 / Aβ40 and Aβ42 / Aβ40 before the addition of the compound. Each square on the upper side of the display screen corresponds to each well on the lower microplate. The control unit 31 obtains the calculated values when the Aβ38 / Aβ40 and Aβ42 / Aβ40 are calculated by obtaining the Aβ38 amount, the Aβ40 amount, and the Aβ42 amount by the ELISA method or the like using the microplate by the first physical characteristic information output model. Then, the calculated value is displayed in each square corresponding to each well. The upper row of each square is Aβ38 / Aβ40, and the lower row is Aβ42 / Aβ40. Columns I and III of the display screen of FIG. 25 correspond to the same differentiated cell (1), and columns II and IV correspond to other differentiated cells (2) different from the differentiated cell (1). The control unit 31 inputs the cell information related to the differentiated cell (1) into the first physical characteristic information output model, and displays the acquired Aβ38 / Aβ40 and Aβ42 / Aβ40 in the squares of columns I and III. The control unit 31 inputs the cell information related to the differentiated cell (2) into the first physical characteristic information output model, and displays the acquired Aβ38 / Aβ40 and Aβ42 / Aβ40 in the squares of columns II and IV.

FIG. 26 is a display screen showing Aβ38 / Aβ40 and Aβ42 / Aβ40 after the addition of the compound. Rows III and IV of the display screen of FIG. 26 are rows of wells to which the compound has been added. The case where different compounds are added to the differentiated cells in the same row is shown. The control unit 31 inputs the cell information including the marker value after addition and the compound information into the second physical characteristic information output model corresponding to each well in the same row, and inputs the acquired Aβ38 / Aβ40 and Aβ42 / Aβ40 in row III. And display in the square of column IV.

The user can obtain the amount of change (difference) in the physical properties before and after the addition of the compound by comparing the display screens I and III of the display panel 35 after the addition with the II and IV columns. ..
The case of displaying physical property information for two types of differentiated cells is not limited. Further, the case is not limited to the case where different compounds are added to the same differentiated cells. Physical property information when the same compound is added to differentiated cells at different concentrations may be displayed on the display panel 35.

According to this embodiment, the effect / efficacy or toxicity / safety of the compound can be evaluated based on the difference between the second and first physical property information.
It is also possible to determine whether or not the evaluation result can be obtained by adding the compound to the differentiated cells based on the cell information before the addition of the compound.
In the present embodiment, the case where the first physical characteristic information output model and the second physical characteristic information output model are generated separately is described, but the present invention is not limited to this. In one physical characteristic information output model, the cell information before the addition of the compound may be input to output the first physical property information, and the cell information after the addition of the compound may be input to output the second physical characteristic information.

It should be considered that the embodiments disclosed this time are exemplary in all respects and are not restrictive. The scope of the present invention is indicated by the scope of claims, not the above-mentioned meaning, and is intended to include all modifications within the meaning and scope equivalent to the scope of claims.

1 Evaluation estimation device 2 Differentiation induction culture device 3, 4 terminals 10 Information processing system 11 Control unit (acquisition unit, selection unit, estimation unit, output unit)
21 Control device 22, 31 Control unit 28 Culture equipment 12, 23, 32 Main storage unit 13, 24, 33 Communication unit 25, 34 Operation unit 14, 27, 36 Auxiliary storage unit 141 Information processing program 142 Manufacturing protocol DB
143 Manufacturing program DB
144 Evaluation Protocol DB
145 Compound Information DB
146 Teacher data DB
147 Physical property information output model DB
148 Usage history DB
149 1st physical characteristic information output model DB
151 Second physical characteristic information output model DB

Claims (17)

Differentiation is induced from the induced pluripotent stem cell to a predetermined cell, and the cell information of the cell that has been induced to differentiate into the predetermined cell is acquired.
The teacher data including the cell information and the physical property information of the cell based on the measurement using the analyzer is acquired, and the teacher data is acquired.
A method for generating a learning model, which generates a learning model that outputs physical property information of the cells when the cell information of the cells induced to differentiate into a predetermined cell is input based on the teacher data. The cell information of the cell induced to differentiate into a predetermined cell and the physical property information based on the actual measurement of the cell are acquired.
The method for generating a learning model according to claim 1, wherein the learning model is relearned based on the cell information and the physical property information. Get compound information,
The teacher data includes compound information
The learning model according to claim 1 or 2, which generates a learning model that outputs physical property information when cell information and compound information of cells induced to differentiate into predetermined cells are input based on the teacher data. How to generate. The cell information is the information including the first information including the cell type, clinical information, genetic information, and culture conditions, and the second information including the cell image and the number of cells after culturing, according to claim 3. How to generate a learning model for. The physical property information is at least one kind of information selected from the number of cells, light absorption characteristics of a predetermined wavelength, chromosomal information, image information, activity potential, medium information, and marker value, according to claims 1 to 4. The method for generating a learning model according to any one of the items. The method for generating a learning model according to claim 5, wherein the physical property information indicates the amount of change over time. An acquisition unit that acquires cell information of cells that have been induced to differentiate into predetermined cells from induced pluripotent stem cells,
When the cell information of a cell whose differentiation has been induced into a predetermined cell is input, the cell information acquired by the acquisition unit is input to a learning model that outputs physical property information of the cell based on measurement using an analyzer. And an estimation unit that estimates physical property information,
An evaluation estimation device equipped with an output unit that outputs estimated physical property information. The acquisition unit acquires compound information and obtains compound information.
The learning model outputs physical property information when the compound information is input.
The evaluation estimation device according to claim 7, wherein the estimation unit inputs the compound information acquired by the acquisition unit into the learning model. The acquisition unit acquires the cell information of the cells before adding the compound, and obtains the cell information.
The learning model is a model that outputs the first physical property information when the cell information before adding the compound is input.
The evaluation estimation device according to claim 8, wherein the estimation unit estimates the first physical property information. The acquisition unit acquires the cell information of the cells after the compound is added, and obtains the cell information.
The learning model is a model that outputs second physical property information when the cell information after adding the compound is input.
The evaluation estimation device according to claim 9, wherein the estimation unit estimates the second physical property information. The evaluation estimation device according to claim 10, further comprising a calculation unit that calculates the physical property information of the compound based on the first physical property information and the second physical property information estimated by the estimation unit. The cell information is information including first information including cell type, clinical information, genetic information, and culture conditions, and second information including cell image and cell number after culturing, claims 7 to 11. The evaluation estimation device according to any one of the items up to. The physical property information is at least one kind of information selected from the number of cells, light absorption characteristics of a predetermined wavelength, chromosomal information, image information, activity potential, medium information, and marker value, according to claims 7 to 12. The evaluation estimation device according to any one item. The ninth aspect of the present invention, which comprises a determination unit for determining whether or not to add the compound based on a biomarker value, a surrogate marker value, or an image of the cell before adding the compound to the cell. Evaluation estimation device. According to claim 7, a display unit for displaying each physical property information estimated by the estimation unit is provided corresponding to each well of the microplate of the differentiation-inducing culture device that induces differentiation from induced pluripotent stem cells into predetermined cells. The evaluation estimation device according to any one of up to 14. An input layer in which cell information of cells induced to differentiate from induced pluripotent stem cells into predetermined cells is input,
An output layer that outputs physical property information of the cells based on measurement using an analyzer, and
It is provided with cell information of cells induced to differentiate from induced pluripotent stem cells into predetermined cells and an intermediate layer in which parameters are learned based on the physical property information of the cells.
When the cell information of one cell induced to differentiate into a predetermined cell is input from the induced pluripotent stem cell, the computer is set so that the physical property information of the cell is output from the output layer through the calculation by the intermediate layer. A learning model that works. Obtain cell information of cells induced to differentiate into predetermined cells from induced pluripotent stem cells,
When the cell information of a cell whose differentiation is induced into a predetermined cell is input, the acquired cell information is input to a learning model that outputs the physical property information of the cell based on the measurement using an analyzer to obtain the physical property information. A computer program that causes a computer to perform processing that outputs.

Images