Cell characteristic property prediction device, method for operating cell characteristic property prediction device, and operation program for cell characteristic property prediction device

Abstract

Provided is a cell characteristic property prediction device for predicting a characteristic property of a target cell that is obtained through monoclonalization from a host cell population that is a collection of host cells having a plurality of different subtypes. The cell characteristic property prediction device includes a processor. The processor acquires subtype information representing a subtype of a target cell derived from a subtype of a host cell, and performs prediction according to the subtype information.

Description

Cell property prediction device, operation method of cell property prediction device, and operation program of cell property prediction device

[0001] The technology of the present disclosure relates to a cell property prediction device, an operating method for a cell property prediction device, and an operating program for a cell property prediction device.

[0002] Currently, the cultivation of antibody-producing cells for the production of antibody drugs, or cells such as cardiomyocytes and nerve cells differentiated from iPS (induced pluripotent stem) cells, is being actively carried out. In the field of cell cultivation, various technologies have been proposed that predict cell characteristics without actually conducting long-term tests by analyzing various data related to the cells using computers.

[0003] For example, JP-A-2022-550083 describes a technique in which antibody-producing cells (referred to as clone cells in JP-A-2022-550083) into which an antibody gene has been incorporated and which have been monoclonalized (made into single cells) are subjected to karyotype analysis, a genomic instability value of the antibody-producing cells is derived as a cellular characteristic, and antibody-producing cells that stably produce antibodies are selected based on the genomic instability value.

[0004] Furthermore, JP 2022-533003 A describes a technology for predicting the value of a product quality attribute of antibody-producing cells as a cell characteristic by analyzing, using a machine learning-based regression estimator, a first attribute value of antibody-producing cells measured by an opto-electronic cell line generation and analysis system and a second attribute value of antibody-producing cells measured in a cell pool screening stage.

[0005] Some cells have multiple different subtypes. The characteristics of the cells are greatly influenced by the subtype. However, the techniques described in JP-A-2022-550083 and JP-A-2022-533003 do not take subtypes into account when predicting cell characteristics. Therefore, there is a possibility that the accuracy of predicting cell characteristics is insufficient.

[0006] One embodiment of the technology of the present disclosure provides a cell property prediction device, an operating method for the cell property prediction device, and an operating program for the cell property prediction device, which are capable of improving the accuracy of predicting cell properties.

[0007] The cell characteristic prediction device disclosed herein is a cell characteristic prediction device that predicts the characteristics of a target cell, which is a cell obtained through monoclonalization from a host cell population, which is a collection of host cells having multiple different subtypes, and is equipped with a processor. The processor obtains subtype information that represents the subtype of the target cell derived from the subtype of the host cell, and makes a prediction based on the subtype information.

[0008] Preferably, the processor acquires target cell information regarding the target cells, determines the subtype of the target cells based on the target cell information, and acquires the determination result of the subtype of the target cells as subtype information.

[0009] Preferably, the processor inputs the target cell information into a first machine learning model and causes the first machine learning model to output a subtype discrimination result.

[0010] The target cell information preferably includes at least one of genetic data of the target cell, culture data of the target cell, and shape data of the target cell.

[0011] Preferably, the processor obtains target cell information regarding the target cell and predicts characteristics of the target cell based on the target cell information.

[0012] Preferably, the processor inputs the target cell information into a second machine learning model and causes the second machine learning model to output a prediction result of the target cell characteristics.

[0013] The target cell information preferably includes at least one of genetic data of the target cell, culture data of the target cell, and shape data of the target cell.

[0014] There are a number of tools for predicting the properties of target cells, and it is preferable that the processor selects a matching tool from the plurality of tools that matches the subtype information and makes the prediction using the matching tool.

[0015] The processor preferably presents the predicted results of the properties of the target cell to the user.

[0016] The processor preferably presents the subtype information to the user.

[0017] Preferably, the processor presents the user with a confidence level for the prediction result.

[0018] The processor preferably determines subsequent culture conditions for the target cells based on at least one of the subtype information and the predicted results of the properties of the target cells.

[0019] Preferably, the host cells are of mammalian origin.

[0020] The target cells are cells that produce a substance that will be an active ingredient in a biopharmaceutical, and the processor preferably predicts the stability of production of the substance as a property of the target cells.

[0021] Preferably, the agent is an antibody.

[0022] The processor preferably predicts the culture conditions of the target cells as the properties of the target cells.

[0023] The method of operating the cell characteristic prediction device disclosed herein is a method of operating the cell characteristic prediction device for predicting the characteristics of a target cell, which is a cell obtained through monoclonalization from a host cell population, which is a collection of host cells having multiple different subtypes, and includes obtaining subtype information representing the subtype of the target cell derived from the subtype of the host cell, and making a prediction based on the subtype information.

[0024] The operating program of the cell characteristic prediction device disclosed herein is an operating program of the cell characteristic prediction device that predicts the characteristics of a target cell, which is a cell obtained through monoclonalization from a host cell population, which is a collection of host cells having multiple different subtypes, and causes a computer to execute processes including obtaining subtype information representing the subtype of the target cell derived from the subtype of the host cell, and making a prediction based on the subtype information.

[0025] According to the technology of the present disclosure, it is possible to provide a cell property prediction device, an operating method for a cell property prediction device, and an operating program for a cell property prediction device that can improve the accuracy of predicting cell properties.

[0026] 1 is a diagram showing an overview of the manufacturing process of an antibody pharmaceutical. FIG. 2 is a diagram showing detailed procedures for the clone production process. FIG. 3 is a diagram showing detailed procedures for the clone production process. FIG. 4 is a diagram showing a production stability prediction server and a user terminal. FIG. 5 is a diagram showing an antibody-producing cell information group. FIG. 6 is a block diagram showing a computer constituting the production stability prediction server and a user terminal. FIG. 7 is a block diagram showing a processing unit of the CPU of the production stability prediction server. FIG. 8 is a diagram showing processing of the discrimination unit. FIG. 9 is a diagram showing prediction models constituting a prediction model group. FIG. 10 is a diagram showing how a prediction model that matches the discrimination result is selected. FIG. 11 is a block diagram showing processing of the prediction unit. FIG. 12 is a block diagram showing an information input screen. FIG. 13 is a diagram showing a prediction result display screen. FIG. 14 is a flowchart showing the processing procedure of the production stability prediction server. FIG. 15 is a diagram showing a prediction result display screen of the second embodiment. FIG. 16 is a table showing the reliability of the prediction results of each prediction model. FIG. 17 is a diagram showing a third embodiment in which culture conditions are determined based on subtype. FIG. 18 is a diagram showing a fourth embodiment in which culture conditions for antibody-producing cells are predicted. FIG. 19 is a performance comparison table between the prediction models of the comparative examples and the prediction models of the examples.

[0027] 1 , an antibody pharmaceutical 10 is manufactured through a clone production process 11, a process development process 12, and a GMP (Good Manufacturing Practice) manufacturing process 13. The antibody pharmaceutical 10 is an example of a "biopharmaceutical" according to the technology of the present disclosure.

[0028] Clone production step 11 is a step of producing multiple clones 24 (see FIG. 2 for all) of antibody-producing cells 23 that produce antibodies 25, and selecting clones 24 with superior properties from the multiple clones 24. Here, clones 24 refer to a population of genetically identical antibody-producing cells 23. Antibody-producing cells 23 are an example of "target cells" and "cells that produce a substance that serves as an active ingredient of a biopharmaceutical" according to the technology of the present disclosure. Furthermore, antibody 25 is an example of "a substance that serves as an active ingredient of a biopharmaceutical" according to the technology of the present disclosure.

[0029] Process development step 12 is a step in which selected clone 24 is used to develop culture and purification conditions for GMP production step 13. GMP production step 13 is a step in which antibody 25 is produced by clone 24 under the culture and purification conditions developed in process development step 12, and antibody 25 is purified and formulated to obtain antibody pharmaceutical 10. The technology of the present disclosure relates to clone production step 11, which is one of clone production step 11, process development step 12, and GMP production step 13. Therefore, the detailed procedure of clone production step 11 will be described below.

[0030] As shown in FIG. 2 , the clone production process 11 begins with preparing a host cell population 21, which is a collection of multiple host cells 20 (step ST10). The host cells 20 are vertebrate-derived cells, more specifically mammalian-derived cells, in this case Chinese Hamster Ovary (CHO) cells. The host cells 20 have congenital or acquired heterogeneity and have three different subtypes: subtypes A, B, and C. Subtypes A and B are predominant in number of host cells 20, while subtype C is rarer than subtypes A and B. The host cells 20 are an example of a “host cell” according to the technology of the present disclosure, and the host cell population 21 is an example of a “host cell population” according to the technology of the present disclosure. The acquired heterogeneity may be intentional or may occur naturally during culture.

[0031] Subtypes may differ, for example, in genetic information. Alternatively, subtypes may differ in characteristics such as RNA (ribonucleic acid) expression level, size, shape, or proliferation rate. Subtypes may represent multiple similar subtypes, such as subtype A including subtypes A1 and A2, and subtype B including subtypes B1, B2, and B3. Furthermore, as described above, the number of subtypes may be biased or approximately equal.

[0032] The host cells 20 may be subjected to single-cell analysis, and subtypes of the host cells 20 may be identified by PCA clustering, which combines principal component analysis (PCA) with cluster analysis. Alternatively, multiple clones 24 of the antibody-producing cells 23 may be produced, and the subtypes of the host cells 20 may be identified based on common factors (RNA, SNP (Single Nucleotide Polymorphism), CNV (Copy Number Variation), etc.) of the clones 24 that have different measurement data such as gene expression levels or proliferation rates.

[0033] Next, antibody genes 22 are incorporated into each of the host cells 20 that make up the host cell population 21 (step ST15). In this way, antibody-producing cells 23 are generated from the host cells 20. Subsequently, the plurality of antibody-producing cells 23 are separated one by one and monoclonalized (step ST20).

[0034] The monoclonalized antibody-producing cells 23 are cultured individually to produce a plurality of clones 24. Each of the plurality of clones 24 is then caused to produce an antibody 25 (step ST25). This step ST25 is a standard test conducted to confirm the antibody production ability of each clone 24, and takes about two weeks.

[0035] The antibody production ability of each clone 24 is not constant; some clones 24 have high antibody production ability and others have low antibody production ability. Some clones 24 hardly produce antibodies 25. Therefore, after the specification test in step ST25 is completed, a first selection is performed to select clones 24 whose numerical value indicating antibody production ability is equal to or greater than a predetermined threshold value (step ST30). In other words, the first selection is a process of excluding clones 24 whose antibody production ability is relatively low. The numerical value indicating antibody production ability is, for example, the amount of antibody produced per unit time during the specification test period.

[0036] Conventionally, a test to confirm the production stability of antibody 25 from clone 24 (production stability test) is then conducted for a specified period, for example, two to three months, which has been a hindrance to the production of antibody pharmaceutical 10.

[0037] Therefore, in the technology of the present disclosure, as shown in step ST35 of FIG. 3 , various data related to the clone 24 (antibody-producing cells 23) are analyzed by a computer to predict the production stability of the antibody 25 from the clone 24 (antibody-producing cells 23) without actually conducting a two- to three-month production stability test (hereinafter referred to as "production stability prediction"). Then, a prediction result 28 of the production stability prediction is presented to a user U (see FIG. 4 ). The start point of the specified period is the start point of the specification test in step ST25 or the end point of the first selection in step ST30. The specified period may be in months, such as the aforementioned two to three months, or may be the period until a predetermined number of passages are performed. Alternatively, the specified period may be determined based on the proliferation capacity of the antibody-producing cells 23 or the actual culture period in the GMP production process 13. The production stability of the antibody 25 is an example of a "characteristic of the target cells" according to the technology of the present disclosure.

[0038] Production stability is determined based on the amount of antibody 25 produced by the clone 24, i.e., the degree of change in antibody production amount over a specified period. Specifically, the degree of change in antibody production amount over a specified period is the rate of increase or decrease in antibody production amount between the start and end of the specified period. The rate of increase or decrease in antibody production amount is obtained by dividing the antibody production amount at the end of the specified period by the antibody production amount at the start. Because antibody production amount often decreases, the rate of increase or decrease in antibody production amount rarely exceeds 100% and is usually less than 100%. Stable production stability refers to a case where the rate of increase or decrease in antibody production amount is equal to or greater than a predetermined threshold. On the other hand, unstable production stability refers to a case where the rate of increase or decrease in antibody production amount is less than a threshold. The threshold is, for example, 70% or 80%.

[0039] After the prediction of production stability in step ST35, a second selection is performed to select clones 24 predicted to have stable production stability (step ST40). In other words, the second selection is a process to remove clones 24 predicted to have unstable production stability. The clones 24 selected in this second selection are used in the subsequent process development step 12 and GMP production step 13.

[0040] As an example, as shown in FIG. 4 , the production stability prediction server 30 is connected to a user terminal 31 via a network 32. The production stability prediction server 30 is a server responsible for the production stability prediction in step ST35 shown in FIG. 3 and is an example of a "cell characteristic prediction device" according to the technology of the present disclosure. The user terminal 31 is installed at a pharmaceutical company developing a biopharmaceutical or an organization contracted by a pharmaceutical company to develop a biopharmaceutical, i.e., a contract research organization (CRO). The user terminal 31 is operated by a user U involved in the development of a biopharmaceutical at the pharmaceutical company or CRO. The network 32 is, for example, a wide area network (WAN) such as the Internet or a public communication network. Note that while only one user terminal 31 is connected to the production stability prediction server 30 in FIG. 4 , in reality, multiple user terminals 31 from multiple pharmaceutical companies or CROs are connected to the production stability prediction server 30.

[0041] The user terminal 31 transmits a prediction request 33 to the production stability prediction server 30. The prediction request 33 is a request for the production stability prediction server 30 to predict production stability. The prediction request 33 includes an antibody-producing cell information group 34. The antibody-producing cell information group 34 is a collection of antibody-producing cell information 40 (see FIG. 5 ) related to the antibody-producing cells 23 that constitute each clone 24 after the first selection. The antibody-producing cell information 40 is an example of "target cell information" according to the technology of the present disclosure. Although not shown in the figure, the prediction request 33 also includes a terminal ID (identification data) for uniquely identifying the user terminal 31 that sent the prediction request 33.

[0042] When the prediction request 33 is received, the production stability prediction server 30 predicts the production stability of the antibody-producing cells 23 of each clone 24 after the first selection. Then, a prediction result group 35, which is a collection of the prediction results 28, is distributed to the user terminal 31 that sent the prediction request 33. When the prediction result group 35 is received, the user terminal 31 makes the prediction result group 35 available for viewing by the user U.

[0043] As an example, as shown in FIG. 5 , antibody-producing cell information 40 includes an antibody-producing cell ID for uniquely identifying the antibody-producing cells 23 constituting each clone 24, genetic data 41 of the antibody-producing cells 23, culture data 42 of the antibody-producing cells 23, and shape data 43 of the antibody-producing cells 23. The genetic data 41 is, for example, gene expression levels of the antibody-producing cells 23 obtained by RNA sequence analysis or quantitative polymerase chain reaction (qPCR) analysis. The gene expression levels are count values ​​that take positive integers and can be used after logarithmic transformation. The culture data 42 is, for example, the cell count and antibody production amount of the antibody-producing cells 23 measured after the standard test in step ST25 shown in FIG. 2 . The shape data 43 is, for example, shape feature values ​​of the antibody-producing cells 23 measured after the standard test and images of the antibody-producing cells 23 taken after the standard test. The shape feature amount is, for example, a representative value of the major axis, minor axis, and / or area of ​​the antibody-producing cells 23. In the case of an image, the shape data may be a representative value of pixel values ​​instead of the image itself. The representative value may be the average value, maximum value, minimum value, mode value, and / or median value.

[0044] 6 , the computers constituting the production stability prediction server 30 and the user terminal 31 basically have the same configuration, and include a storage 45, a memory 46, a CPU (Central Processing Unit) 47, a communication unit 48, a display 49, and an input device 50. These are interconnected via a bus line 51.

[0045] The storage 45 is a hard disk drive built into the computer constituting the production stability prediction server 30 and the user terminal 31, or connected via a cable or network. Alternatively, the storage 45 is a disk array consisting of multiple hard disk drives. The storage 45 stores control programs such as an operating system, various application programs (hereinafter referred to as APs (Application Programs)), and various data associated with these programs. Note that a solid state drive may be used instead of a hard disk drive.

[0046] The memory 46 is a work memory for the CPU 47 to execute processing. The CPU 47 loads programs stored in the storage 45 into the memory 46 and executes processing in accordance with the programs. In this way, the CPU 47 comprehensively controls each part of the computer. The CPU 47 is an example of a "processor" according to the technology of the present disclosure. The memory 46 may be built into the CPU 47.

[0047] The communication unit 48 is a network interface that controls the transmission of various information via the network 32, etc. The display 49 displays various screens. The various screens are equipped with an operation function using a GUI (Graphical User Interface). The computers that make up the production stability prediction server 30 and the user terminal 31 accept input of operation instructions from an input device 50 through the various screens. The input device 50 is a keyboard, a mouse, a touch panel, a microphone for voice input, etc.

[0048] In the following explanation, the parts of the computer that make up the production stability prediction server 30 (storage 45 and CPU 47) are distinguished by adding the suffix "A" to their symbols, and the parts of the computer that make up the user terminal 31 (storage 45, CPU 47, display 49, and input device 50) are distinguished by adding the suffix "B" to their symbols.

[0049] 7 , an operating program 60 is stored in the storage 45A of the production stability prediction server 30. The operating program 60 is an AP for causing a computer to function as the production stability prediction server 30. In other words, the operating program 60 is an example of an "operating program for a cell characteristic prediction device" according to the technology of the present disclosure. The storage 45A also stores an antibody-producing cell information group 34, a discrimination model 61, a prediction model group 62, and the like.

[0050] When the operating program 60 is started, the CPU 47A of the computer constituting the production stability prediction server 30 works in cooperation with the memory 46 etc. to function as a request receiving unit 65, an RW control unit 66, a discrimination unit 67, a prediction unit 68, and a screen distribution control unit 69.

[0051] The request receiving unit 65 receives various requests from the user terminal 31. In particular, the request receiving unit 65 receives a prediction request 33 from the user terminal 31. As described above, the prediction request 33 includes the antibody-producing cell information group 34. The antibody-producing cell information group 34 is a collection of multiple pieces of antibody-producing cell information 40. Therefore, by receiving the prediction request 33, the request receiving unit 65 acquires the antibody-producing cell information 40. When the prediction request 33 is received, the request receiving unit 65 outputs the antibody-producing cell information group 34 included in the prediction request 33 to the RW control unit 66. Furthermore, the request receiving unit 65 outputs the terminal ID of the user terminal 31 included in the prediction request 33 to the screen distribution control unit 69.

[0052] The RW control unit 66 controls the storage of various data in the storage 45A and the reading of various data from the storage 45A. For example, the RW control unit 66 stores the antibody-producing cell information group 34 from the request receiving unit 65 in the storage 45A. The RW control unit 66 also reads the antibody-producing cell information group 34 from the storage 45A and outputs the read antibody-producing cell information group 34 to the discrimination unit 67 and the prediction unit 68. The RW control unit 66 also reads the discrimination model 61 from the storage 45A and outputs the read discrimination model 61 to the discrimination unit 67. The RW control unit 66 also reads the prediction model group 62 from the storage 45A and outputs the read prediction model group 62 to the prediction unit 68.

[0053] The discrimination unit 67 uses the discrimination model 61 to discriminate the subtype of the antibody-producing cells 23 that constitute each clone 24. The subtype of the antibody-producing cells 23 is derived from the subtype of the original host cell 20. The discrimination unit 67 outputs a discrimination result group 72, which is a collection of subtype discrimination results 75 (see FIG. 8 ), to the prediction unit 68 and the screen distribution control unit 69.

[0054] The prediction unit 68 uses the prediction model group 62 to predict the production stability of each clone 24. The prediction unit 68 outputs a prediction result group 35, which is a collection of prediction results 28 of the production stability prediction of each clone 24, to the screen distribution control unit 69.

[0055] The screen delivery control unit 69 controls the delivery of various screens to the user terminal 31. Specifically, the screen delivery control unit 69 delivers and outputs various screens to the user terminal 31 that has sent the various requests in the form of screen data for web delivery created using a markup language such as XML (Extensible Markup Language). At this time, the screen delivery control unit 69 identifies the user terminal 31 that has sent the various requests based on the terminal ID from the request receiving unit 65. Note that instead of XML, other data description languages ​​such as JSON (Javascript (registered trademark) Object Notation) may be used.

[0056] The various screens include an information input screen 85 (see FIG. 13) for inputting antibody-producing cell information 40, and a prediction result display screen 95 (see FIG. 14) for displaying prediction results 28. In addition to these processing units 65 to 69, the CPU 47A also includes an instruction receiving unit that receives various operation instructions from the input device 50.

[0057] As an example, as shown in FIG. 8 , the discrimination unit 67 inputs antibody-producing cell information 40 (including genetic data 41, culture data 42, and shape data 43) into a discrimination model 61, and causes the discrimination model 61 to output a discrimination result 75. The discrimination result 75 includes the antibody-producing cell ID and the subtype of the antibody-producing cell 23. In other words, the discrimination result 75 is an example of the "subtype information" according to the technology of the present disclosure.

[0058] The discrimination model 61 is a machine learning model trained to discriminate the subtype of antibody-producing cells 23 based on the antibody-producing cell information 40. The training data for the discrimination model 61 is a combination of antibody-producing cell information 40 of previously generated antibody-producing cells 23 and the subtype (correct answer data) determined by detailed genetic analysis of the antibody-producing cells 23. One or more of the multiple elements constituting the antibody-producing cell information 40 may be removed from the input to the discrimination model 61, a learning discrimination result 75 may be output from the discrimination model 61, and the discrimination results 75 with and without the removed elements may be compared to calculate the contribution of the removed elements to the discrimination result 75, and elements with high contributions may be narrowed down to the elements to be ultimately input to the discrimination model 61. For example, for the genetic data 41 of the antibody-producing cell information 40, the gene expression levels of several hundred genes may be narrowed down to the gene expression levels of several to several dozen genes with high contributions (so-called marker genes) as the gene expression levels to be ultimately input to the discrimination model 61. In this case, genes that show statistically significant differences in gene expression levels among the subtypes may be first narrowed down, and then the genes may be narrowed down based on their contribution.

[0059] The discrimination model 61 outputs, for example, the probability that the antibody-producing cells 23 are subtype A, the probability that they are subtype B, and the probability that they are subtype C, and discriminates the subtype with the highest probability as the subtype of the antibody-producing cells 23. The probabilities that the antibody-producing cells 23 are subtype A, the probability that they are subtype B, and the probability that they are subtype C add up to 100%. The discrimination model 61 is an example of a "first machine learning model" according to the technology of the present disclosure.

[0060] 9 , the prediction model group 62 is composed of three prediction models 78: a subtype A prediction model 78A, a subtype B prediction model 78B, and a subtype C prediction model 78C. The subtype A prediction model 78A, the subtype B prediction model 78B, and the subtype C prediction model 78C are, as the name suggests, models that predict the production stability of subtype A antibody-producing cells 23, subtype B antibody-producing cells 23, and subtype C antibody-producing cells 23, respectively. The subtype A prediction model 78A, the subtype B prediction model 78B, and the subtype C prediction model 78C are examples of a "second machine learning model" and "multiple tools" according to the technology of the present disclosure.

[0061] 10A, if the subtype of the antibody-producing cells 23 according to the discrimination result 75 is subtype A, the prediction unit 68 selects a subtype A prediction model 78A. As shown in (B), if the subtype of the antibody-producing cells 23 according to the discrimination result 75 is subtype B, the prediction unit 68 selects a subtype B prediction model 78B. As shown in (C), if the subtype of the antibody-producing cells 23 according to the discrimination result 75 is subtype C, the prediction unit 68 selects a subtype C prediction model 78C. The subtype A prediction model 78A in the case of (A), the subtype B prediction model 78B in the case of (B), and the subtype C prediction model 78C in the case of (C) are examples of "matching tools" according to the technology of the present disclosure.

[0062] As an example, as shown in FIG. 11 , the prediction unit 68 inputs antibody-producing cell information 40 (the genetic data 41, culture data 42, and shape data 43) into a prediction model 78 selected as shown in FIG. 10 and causes the prediction model 78 to output a prediction result 28. The prediction model 78 is a machine learning model trained to predict production stability based on the antibody-producing cell information 40. The training data for the prediction model 78 is a set of antibody-producing cell information 40 of previously generated antibody-producing cells 23 and the production stability (correct answer data) determined by actually conducting a production stability test on a clone 24 of the antibody-producing cells 23. The training data for the subtype A prediction model 78A consists only of the set of antibody-producing cell information 40 and production stability of previously generated subtype A antibody-producing cells 23. Similarly, the training data for the subtype B prediction model 78B consists only of the set of antibody-producing cell information 40 and production stability of previously generated subtype B antibody-producing cells 23. Furthermore, the training data for the subtype C prediction model 78C consists only of sets of antibody-producing cell information 40 and production stability for subtype C antibody-producing cells 23 that were generated in the past. The prediction model 78 is preferably a model that robustly predicts production stability for antibody-producing cells 23 that produce unknown antibodies 25. In other words, the prediction model 78 is preferably a model with domain generalizability, with a domain being a substance that serves as an active ingredient of a biopharmaceutical, such as antibody 25.

[0063] As in the case of the discrimination model 61, one or more of the multiple elements constituting the antibody-producing cell information 40 may be excluded and input to the prediction model 78, and a learning prediction result 28 may be output from the prediction model 78. The prediction results 28 with and without the elements excluded may be compared to calculate the contribution of the excluded elements to the prediction result 28, and elements with high contributions may be narrowed down to the elements to be finally input to the prediction model 78. For example, for the genetic data 41 of the antibody-producing cell information 40, the gene expression levels of several to several tens of genes with high contributions may be narrowed down to the gene expression levels to be finally input to the prediction model 78 from the gene expression levels of hundreds of genes. In this case, genes that show a statistically significant difference in gene expression level depending on whether production stability is stable or unstable may first be narrowed down, and then the narrowing down based on contribution may be performed.

[0064] The prediction model 78 outputs, for example, the probability that the production stability of the antibody 25 in the antibody-producing cells 23 is stable and the probability that it is unstable, and adopts the larger probability as the prediction result 28. The probability that the production stability of the antibody 25 in the antibody-producing cells 23 is stable and the probability that it is unstable add up to 100%.

[0065] Here, the elements of the antibody-producing cell information 40 input into the discrimination model 61 may all be the same as the elements of the antibody-producing cell information 40 input into the prediction model 78, or some or all may be different. Similarly, the elements of the antibody-producing cell information 40 input into the subtype A prediction model 78A, the subtype B prediction model 78B, and the subtype C prediction model 78C may all be the same, or some or all may be different. In any case, to save time and effort, it is preferable to obtain all elements of the antibody-producing cell information 40 at once without leaving anything out.

[0066] As an example, as shown in FIG. 12 , a prediction AP 80 is stored in the storage 45B of the user terminal 31. The prediction AP 80 is installed in the user terminal 31 by the user U. The prediction AP 80 is an AP for predicting production stability. When the prediction AP 80 is launched, the CPU 47B of the user terminal 31 functions as a browser control unit 82 in cooperation with the memory 46 and the like. The browser control unit 82 controls the operation of a web browser dedicated to the prediction AP 80.

[0067] The browser control unit 82 reproduces various screens based on various screen data from the production stability prediction server 30 and displays the reproduced various screens on the display 49B. The browser control unit 82 also accepts various operation instructions input by the user U from the input device 50B via the various screens. The browser control unit 82 transmits various requests, including the prediction request 33, to the production stability prediction server 30 in response to the operation instructions.

[0068] When the prediction AP 80 is launched, an information input screen 85, as shown in FIG. 13 as an example, is displayed on the display 49B under the control of the browser control unit 82. The information input screen 85 is provided with an input box 86 for the antibody-producing cell information 40 for each clone 24. A file of the antibody-producing cell information 40 can be dropped into the input box 86. The input box 86 can be added by selecting add buttons 87A and 87B at the bottom. The add button 87A is a button for adding one input box 86 at a time, and the add button 87B is a button for adding ten input boxes 86 at a time.

[0069] The user U inputs the desired antibody-producing cell information 40 in the input box 86, and then selects the prediction button 88. When the prediction button 88 is selected, the browser control unit 82 generates a prediction request 33 including an antibody-producing cell information group 34, which is a collection of the antibody-producing cell information 40 input in the input box 86, and transmits the generated prediction request 33 to the production stability prediction server 30.

[0070] Furthermore, when production stability prediction is performed in the production stability prediction server 30, a prediction result display screen 95 shown in Fig. 14 as an example is displayed on the display 49B under the control of the browser control unit 82. The prediction result display screen 95 displays a list 96 summarizing the prediction results 28 and discrimination results 75 for each clone 24. In this way, the prediction results 28 are presented to the user U in the form of screen data distribution.

[0071] A save button 97 and an OK button 98 are provided at the bottom of the prediction result display screen 95. When the save button 97 is selected, the contents of the list 96 are stored in the storage 45B of the user terminal 31. When the OK button 98 is selected, the display of the prediction result display screen 95 is cleared.

[0072] Next, the operation of the above configuration will be described with reference to the flowchart shown in Fig. 15 as an example. When the operating program 60 is started in the production stability prediction server 30, the CPU 47A of the production stability prediction server 30 functions as a request receiving unit 65, an RW control unit 66, a discrimination unit 67, a prediction unit 68, and a screen distribution control unit 69, as shown in Fig. 7. When the prediction AP 80 is started in the user terminal 31, the CPU 47B of the user terminal 31 functions as a browser control unit 82, as shown in Fig. 12.

[0073] 13 is displayed on the display 49B of the user terminal 31 under the control of the browser control unit 82. When the user U inputs the desired antibody-producing cell information 40 into the input box 86 on the information input screen 85 and selects the prediction button 88, a prediction request 33 is sent from the browser control unit 82 to the production stability prediction server 30. As shown in FIG. 1, the prediction request 33 includes an antibody-producing cell information group 34, which is a collection of the antibody-producing cell information 40, the terminal ID of the user terminal 31, etc.

[0074] In the production stability prediction server 30, the request receiving unit 65 receives the prediction request 33, thereby acquiring the antibody-producing cell information 40 of the antibody-producing cell information group 34 included in the prediction request 33 (YES in step ST100). The antibody-producing cell information group 34 is output from the request receiving unit 65 to the RW control unit 66 and stored in storage 45A under the control of the RW control unit 66 (step ST110). In addition, the terminal ID of the user terminal 31 included in the prediction request 33 is output from the request receiving unit 65 to the screen distribution control unit 69.

[0075] The antibody-producing cell information group 34 is read out from storage 45A by the RW control unit 66 (step ST120). The antibody-producing cell information group 34 is output from the RW control unit 66 to the discrimination unit 67 and the prediction unit 68. The RW control unit 66 also reads out the discrimination model 61 from storage 45A and outputs the read-out discrimination model 61 to the discrimination unit 67. The RW control unit 66 also reads out the prediction model group 62 from storage 45A and outputs the read-out prediction model group 62 to the prediction unit 68.

[0076] In the discrimination unit 67, as shown in Fig. 8, the antibody-producing cell information 40 is input to the discrimination model 61. As a result, a discrimination result 75 of the subtype of the antibody-producing cells 23 is output from the discrimination model 61 (step ST130). Discrimination of the subtype of the antibody-producing cells 23 based on this antibody-producing cell information 40 is performed on the antibody-producing cells 23 that make up each clone 24. A discrimination result group 72, which is a collection of the discrimination results 75 of the antibody-producing cells 23 that make up each clone 24, is output from the discrimination unit 67 to the prediction unit 68 and the screen distribution control unit 69.

[0077] As shown in FIG. 10 , the prediction unit 68 selects one prediction model 78 that matches the discrimination result 75 from among the subtype A prediction model 78A, the subtype B prediction model 78B, and the subtype C prediction model 78C. Then, as shown in FIG. 11 , the antibody-producing cell information 40 is input to the selected prediction model 78. As a result, a prediction result 28 is output from the prediction model 78 (step ST140). A prediction of the production stability of the antibody-producing cells 23 based on this antibody-producing cell information 40 is performed for the antibody-producing cells 23 that constitute each clone 24. A prediction result group 35, which is a collection of prediction results 28 for the antibody-producing cells 23 that constitute each clone 24, is output from the prediction unit 68 to the screen distribution control unit 69.

[0078] The screen delivery control unit 69 generates screen data for the prediction result display screen 95 shown in Fig. 14 based on the prediction result group 35 and the discrimination result group 72. Under the control of the screen delivery control unit 69, the screen data for the prediction result display screen 95 is delivered to the user terminal 31 that is the sender of the prediction request 33 (step ST150).

[0079] In the user terminal 31, under the control of the browser control unit 82, the screen data of the prediction result display screen 95 is reproduced, and the reproduced prediction result display screen 95 is displayed on the display 49B. In this way, the prediction result 28 is presented to the user U.

[0080] As described above, the CPU 47A of the production stability prediction server 30 includes a discrimination unit 67 and a prediction unit 68. The discrimination unit 67 generates and obtains a discrimination result 75 indicating the subtype of the antibody-producing cells 23 derived from the subtype of the host cell 20. The prediction unit 68 predicts production stability in accordance with the discrimination result 75. Therefore, it is possible to improve the prediction accuracy of the production stability of the antibody-producing cells 23 compared to when the subtype is not taken into consideration.

[0081] The request receiving unit 65 acquires antibody-producing cell information 40 regarding the antibody-producing cells 23. The discrimination unit 67 discriminates the subtype of the antibody-producing cells 23 based on the antibody-producing cell information 40 and acquires the discrimination result 75 of the subtype of the antibody-producing cells 23 as subtype information. This eliminates the need for the user U to discriminate the subtype based on the antibody-producing cell information 40 and input the discriminated subtype into the production stability prediction server 30 via the user terminal 31.

[0082] The discrimination unit 67 inputs the antibody-producing cell information 40 into the discrimination model 61, and causes the discrimination model 61 to output the subtype discrimination result 75. Therefore, the subtype discrimination result 75 can be easily obtained.

[0083] The prediction unit 68 predicts production stability based on the antibody-producing cell information 40. More specifically, the prediction unit 68 inputs the antibody-producing cell information 40 into a prediction model 78, and causes the prediction model 78 to output a production stability prediction result 28. This makes it possible to easily obtain a subtype discrimination result 75.

[0084] The antibody-producing cell information 40 includes genetic data 41 of the antibody-producing cells 23, culture data 42 of the antibody-producing cells 23, and shape data 43 of the antibody-producing cells 23. The genetic data 41, culture data 42, and shape data 43 are very useful data for understanding the characteristics of the antibody-producing cells 23. This can improve the accuracy of the discrimination model 61 in discriminating subtypes and the accuracy of the prediction model 78 in predicting production stability. The antibody-producing cell information 40 only needs to include at least one of the genetic data 41, culture data 42, and shape data 43.

[0085] The prediction unit 68 selects a prediction model 78 that matches the discrimination result 75 from among the subtype A prediction model 78A, the subtype B prediction model 78B, and the subtype C prediction model 78C, and predicts production stability using the selected prediction model 78. By predicting production stability using a prediction model 78 specialized for each subtype in this way, it is possible to improve the prediction accuracy of the production stability of antibody-producing cells 23 compared to when a single prediction model is used regardless of the subtype (see FIG. 20 ).

[0086] The screen distribution control unit 69 distributes the screen data of the prediction result display screen 95 to the user terminal 31, thereby presenting the prediction result 28 of the production stability of the antibody-producing cells 23 to the user U. By referring to the prediction result 28, the user U can easily complete the second selection in step ST40 shown in FIG.

[0087] Furthermore, the screen distribution control unit 69 distributes screen data of the prediction result display screen 95 to the user terminal 31, thereby presenting the subtype discrimination result 75 of the antibody-producing cells 23 to the user U. The user U can know the subtype in addition to the prediction result 28. The user U can take various measures according to the subtype, such as setting culture conditions in the subsequent process development step 12 etc. according to the subtype.

[0088] The host cells 20 are mammalian-derived cells, which are widely used in the production of antibody pharmaceuticals 10. This increases the versatility of the technology of the present disclosure.

[0089] In this embodiment, the target cells are antibody-producing cells 23 that produce antibody 25, which is the active ingredient of the antibody pharmaceutical 10. Antibody pharmaceuticals 10 containing antibody 25 as the active ingredient are widely used in the treatment of chronic diseases such as cancer, diabetes, and rheumatoid arthritis, as well as rare diseases such as hemophilia and Crohn's disease. Therefore, this embodiment, in which the substance is antibody 25, can facilitate the production of antibody pharmaceuticals 10 that are widely used in the treatment of various diseases.

[0090] The prediction unit 68 then predicts the production stability of the antibody 25 as a property of the target cells, thereby eliminating the need for production stability testing and greatly facilitating the production of the antibody pharmaceutical 10.

[0091] The substance that serves as the active ingredient of the biopharmaceutical is not limited to the illustrated antibody 25. It may also be an antibody-like protein, a peptide, a virus, or the like. It may also be a cytokine (interferon, interleukin, etc.), a hormone (insulin, glucagon, follicle-stimulating hormone, erythropoietin, etc.), a growth factor (IGF (insulin-like growth factor)-1, bFGF (basic fibroblast growth factor), etc.), a blood coagulation factor (factor 7, factor 8, factor 9, etc.), an enzyme (lysosomal enzyme, DNA (deoxyribonucleic acid) degrading enzyme, etc.), an Fc (fragment crystalline) fusion protein, a receptor, albumin, or a protein vaccine. The antibody 25 also includes bispecific antibodies, antibody-drug conjugates, low molecular weight antibodies, glycosylated antibodies, and the like.

[0092] Second Embodiment In the second embodiment, the screen delivery control unit 69 delivers a prediction result display screen 105 shown in FIG. 16 as an example to the user terminal 31 that has transmitted the prediction request 33. The prediction result display screen 105 is displayed on the display 49B under the control of the browser control unit 82. The prediction result display screen 105 displays a list 107 including the prediction result 28, the discrimination result 75, and the reliability 106 of the prediction result 28.

[0093] The reliability 106 is assigned by the prediction unit 68 only to prediction results 28 that indicate stable production stability. The prediction unit 68 derives the reliability 106 based on a table 110 shown in FIG. 17 , for example. As shown in table 110, the reliability 106 is determined by a combination of the type of prediction model 78 used and the probability P of stable production stability output by the prediction model 78. The reliability 106 is divided into three levels: high, medium, and low. The reliability 106 of the subtype A prediction model 78A is set relatively higher than the reliability 106 of the subtype B prediction model 78B and the subtype C prediction model 78C. For example, if the prediction model 78 used is the subtype A prediction model 78A and the probability P is greater than 75% and less than or equal to 100% (75% < P ≦ 100%), the reliability 106 is high. Furthermore, if the prediction model 78 used is a prediction model 78B for subtype B or a prediction model 78C for subtype C, and the probability P is greater than 50% and less than or equal to 75% (50% < P ≦ 75%), the reliability 106 is low.

[0094] The reliability 106 of the subtype A prediction model 78A is set relatively higher than the reliability 106 of the subtype B prediction model 78B and the subtype C prediction model 78C for the following reasons: The number of pieces of training data for the subtype B prediction model 78B and the subtype C prediction model 78C is less than the number of pieces of training data for the subtype A prediction model 78A, and the training of the subtype B prediction model 78B and the subtype C prediction model 78C is insufficient.

[0095] As described above, in the second embodiment, the screen distribution control unit 69 presents the reliability 106 of the prediction result 28 to the user U. The user U can perform the second selection by referring to not only the prediction result 28 but also the reliability 106. For example, in the case of FIG. 16 , the prediction result 28 indicates that clones 1, 2, 5, 6, 8, and 10 all have stable production stability, but clones 1 and 2 with high reliability 106 can be preferentially selected. Furthermore, for clones with low reliability 106, an actual production stability test can be conducted to confirm production stability.

[0096] In addition, without taking into consideration the probability P that the production stability output by the prediction model 78 is stable, the reliability 106 may simply be set to high if the prediction model 78 used is a prediction model 78A for subtype A, medium if the prediction model 78B for subtype B, or low if the prediction model 78C for subtype C.

[0097] [Third Embodiment] In the third embodiment, as shown in FIG. 18 as an example, the prediction unit 68 determines basic culture conditions 115 for antibody-producing cells 23 in the process development process 12 and GMP production process 13 subsequent to the clone production process 11 based on the subtype information indicated by the discrimination result 75. The culture conditions 115 include the type of culture medium, the hydrogen ion exponent (pH: Potential Hydrogen) of the culture medium, the temperature of the culture environment, and the carbon dioxide concentration of the culture environment. The culture conditions 115 are preset according to the subtype and stored in the storage 45A. (A) shows the culture conditions 115 for subtype A, (B) for subtype B, and (C) for subtype C. The culture conditions 115 are displayed on the prediction result display screen 95 or 105 in response to the selection of a display button, and can be viewed by the user U.

[0098] As described above, in the third embodiment, the prediction unit 68 determines the culture conditions 115 for the antibody-producing cells 23 after the clone production step 11 based on the subtype information. This saves the user U the trouble of determining the culture conditions 115. Furthermore, the subsequent steps can be carried out under culture conditions 115 that are suitable for each subtype of antibody-producing cells 23. Note that the culture conditions 115 are merely basic conditions, and are conditions that are subject to various fine adjustments in the process development step 12 and the GMP production step 13.

[0099] [Fourth Embodiment] In each of the above embodiments, the production stability of the antibody 25 is predicted as a characteristic of the antibody-producing cells 23, but this is not limiting. As an example, as shown in the prediction model 120 and prediction result 121 in Fig. 19, the culture conditions of the antibody-producing cells 23 after the clone production step 11, in this case the pH of the medium, may be predicted as a characteristic. In this case, the prediction unit 68 predicts the culture conditions of the antibody-producing cells 23, thereby determining the culture conditions of the antibody-producing cells 23 after the clone production step 11.

[0100] As described above, in the fourth embodiment, the prediction unit 68 predicts the culture conditions for the antibody-producing cells 23 as the characteristics of the antibody-producing cells 23. This saves the user U the trouble of determining the culture conditions. Furthermore, the subsequent steps can be carried out under culture conditions that are suitable for the antibody-producing cells 23.

[0101] The predicted characteristics of the antibody-producing cells 23 may be the amount of antibody 25 produced, the quality of the antibody 25, or the like.

[0102] [Example] In this example, the host cells were CHO cells, and the substance was an antibody. As evaluation samples, multiple clones of each of the CHO cells producing five types of antibodies were prepared. Two subtypes of CHO cells were identified, and hereinafter, these two subtypes are referred to as subtype X and subtype Y. After the standard test was completed, the gene expression levels of all genes in each sample were measured by RNA sequencing analysis. Note that the standard test was performed with a seeding number of clones of 5 x 10 ５ The culture was carried out at a concentration of 1000 cells / mL in suspension culture in a 40 mL flask.

[0103] As an example, a logistic regression model was prepared for each subtype X and Y, in which the gene expression levels of 100 genes were used as explanatory variables and production stability was classified into two classes: stable and unstable. Five-fold cross-validation was performed to train the prediction model. Furthermore, genes showing a statistically significant difference in gene expression levels depending on whether production stability was stable or unstable were narrowed down to 300 to 400 genes, and then the number of genes to be input was narrowed down based on contribution to 100.

[0104] As a comparative example, one prediction model was prepared that predicts production stability without distinguishing between subtypes X and Y. Five-fold cross-validation was also performed when training the prediction model of this comparative example. In both the examples and comparative examples, the five-fold cross-validation was performed by dividing the sets of gene expression levels and correct data for production stability for each of the five types of antibodies prepared as evaluation samples, and using the sets of gene expression levels and correct data for production stability for untrained antibodies as test data to evaluate the prediction accuracy of the prediction model. More specifically, the sets of gene expression levels and correct data for production stability for four types of antibodies were used as training data, and the set of gene expression levels and correct data for production stability for the remaining one type of antibody was used as test data.

[0105] The results of the performance evaluation of the prediction models of this example and the comparative example are shown in Table 125 of Fig. 20. The values ​​in Table 125 are PR-AUC (Area Under the Precision-Recall Curve), which allows for comprehensive evaluation of the performance of the machine learning model.

[0106] According to Table 125, the prediction model of this example had higher overall performance than the prediction model of the comparative example. The reason for this is thought to be that the prediction model of the comparative example had relatively low performance for subtype Y and relatively low prediction accuracy for production stability in the case of subtype Y, whereas the prediction model of this example had significantly improved performance for subtype Y compared to the comparative example and relatively high prediction accuracy for production stability in the case of subtype Y.

[0107] When the prediction model of the Comparative Example is used, if the subtypes of the clones used to predict production stability are mostly subtype Y for some reason, there is a high risk of errors in the second selection if the prediction results are trusted. In contrast, the prediction model of this Example can reduce the risk of errors in the second selection.

[0108] The predictive model of this embodiment and the predictive model of the comparative example both have the same performance for subtype X. In other words, there is no difference in performance for subtype X between the predictive model of this embodiment and the predictive model of the comparative example. This result indicates that it is acceptable to use the predictive model of the comparative example as the predictive model for subtype X of this embodiment, and furthermore, it is acceptable to use training data of a clone of subtype Y as training data for the predictive model for subtype X of this embodiment. In other words, this indicates that the predictive model of this embodiment is highly flexible.

[0109] From the above, it was confirmed that the technology disclosed herein makes it possible to improve the accuracy of predicting cell properties.

[0110] One or more genes capable of distinguishing subtypes may be searched for, and the subtype may be distinguished based on the gene expression levels of the genes. In other words, the discrimination model 61 does not necessarily have to be used to distinguish subtypes. The prediction model 78 does not necessarily have to be used to predict the characteristics of the antibody-producing cells 23. A rule base may be used instead of the discrimination model 61 and the prediction model 78.

[0111] The user U may input subtype information through the user terminal 31, and the subtype information input by the user U may be acquired by the production stability prediction server 30.

[0112] In the above embodiments, a prediction model is prepared for each subtype, but this is not limited to this. For example, two prediction models may be prepared for three subtypes, A, B, and C: a prediction model for subtype A and a prediction model for both subtypes B and C. Alternatively, a single prediction model may be used, and subtype information may be included in the input of the prediction model.

[0113] The host cells 20 are not limited to the exemplified CHO cells, but may also be HEK (Human Embryonic Kidney) cells. Furthermore, the host cells 20 are not limited to mammalian-derived cells, but may also be insect cells, etc.

[0114] The host cell is not limited to the exemplified host cell 20. The target cell is also not limited to the exemplified antibody-producing cell 23. For example, an iPS cell may be used as the host cell, and a cardiomyocyte, nerve cell, or the like, differentiated from the iPS cell may be used as the target cell. In this case, the characteristics of the target cell to be predicted include the difficulty of differentiation induction, the proliferation ability of the differentiated cell, or the subsequent culture conditions for the differentiated cell, etc.

[0115] The discrimination model 61 and the prediction model 78 may continue to be trained even after being stored in the storage 45A of the production stability prediction server 30.

[0116] The production stability prediction server 30 may be installed in a pharmaceutical company or a pharmaceutical development contract organization, or may be installed in a data center independent of the pharmaceutical company or pharmaceutical development contract organization.

[0117] Instead of delivering screen data of the prediction result display screen 95 including the prediction result 28, the discrimination result 75, etc. to the user terminal 31, the prediction result 28, etc. themselves may be delivered to the user terminal 31. In this case, the user terminal 31 generates the prediction result display screen 95 based on the prediction result 28, etc. under the control of the browser control unit 82.

[0118] The method of presenting the prediction results 28, etc. to the user U is not limited to the example of delivering screen data. The prediction results 28, etc. may be presented to the user U by printing them on a paper medium, or by attaching them to an email and sending them to the user terminal 31.

[0119] The hardware configuration of the computer constituting the production stability prediction server 30 according to the technology of the present disclosure can be modified in various ways. For example, the production stability prediction server 30 can be configured with multiple computers separated as hardware in order to improve processing power and reliability. For example, the functions of the request reception unit 65 and the RW control unit 66 and the functions of the discrimination unit 67, the prediction unit 68, and the screen distribution control unit 69 can be distributed and performed by two computers. In this case, the production stability prediction server 30 is configured with two computers.

[0120] In this way, the hardware configuration of the computer of the production stability prediction server 30 can be changed as appropriate depending on the required performance such as processing power, safety, reliability, etc. Furthermore, not only the hardware but also APs such as the operating program 60 can be duplicated or stored in a distributed manner in multiple storage devices in order to ensure safety and reliability.

[0121] In each of the above embodiments, the hardware structure of the processing unit that executes various processes, such as the request receiving unit 65, the RW control unit 66, the discrimination unit 67, the prediction unit 68, the screen delivery control unit 69, and the browser control unit 82, can be any of the various processors listed below. As described above, the various processors include the CPUs 47A and 47B, which are general-purpose processors that execute software (the operating program 60 and the prediction AP 80) and function as various processing units, as well as programmable logic devices (PLDs) that are processors whose circuit configuration can be changed after manufacture, such as an FPGA (Field Programmable Gate Array), and dedicated electrical circuits that are processors having a circuit configuration designed specifically for executing specific processing, such as an ASIC (Application Specific Integrated Circuit).

[0122] A single processing unit may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (e.g., a combination of multiple FPGAs and / or a combination of a CPU and an FPGA).Furthermore, multiple processing units may be configured with a single processor.

[0123] Examples of configuring multiple processing units with a single processor include, first, a form in which one processor is configured with a combination of one or more CPUs and software, as typified by computers such as client and server, and this processor functions as multiple processing units. Second, a form in which a processor is used to realize the functions of the entire system including multiple processing units with a single IC (Integrated Circuit) chip, as typified by systems on chips (SoCs). In this way, various processing units are configured using one or more of the above-mentioned various processors as a hardware structure.

[0124] Furthermore, more specifically, the hardware structure of these various processors can be an electric circuit (circuitry) that combines circuit elements such as semiconductor elements.

[0125] From the above description, the technology described in the following supplementary paragraphs can be understood.

[0126] [Supplementary Item 1] A cell characteristic prediction device for predicting the characteristics of a target cell, which is a cell obtained through monoclonalization from a host cell population, which is a collection of host cells having a plurality of different subtypes, comprising a processor, wherein the processor acquires subtype information indicating a subtype of the target cell derived from a subtype of the host cell, and performs the prediction according to the subtype information. [Supplementary Item 2] The cell characteristic prediction device of Supplementary Item 1, wherein the processor acquires target cell information regarding the target cell, determines the subtype of the target cell based on the target cell information, and obtains the determination result of the subtype of the target cell as the subtype information. [Supplementary Item 3] The cell characteristic prediction device of Supplementary Item 2, wherein the processor inputs the target cell information to a first machine learning model and causes the first machine learning model to output the determination result of the subtype. [Supplementary Item 4] The cell characteristic prediction device of Supplementary Item 2 or Supplementary Item 3, wherein the target cell information includes at least one of genetic data of the target cell, culture data of the target cell, and shape data of the target cell. [Supplementary Item 5] The cell property prediction device of any one of Supplementary Items 1 to 4, wherein the processor acquires target cell information about the target cell, and predicts the properties of the target cell based on the target cell information. [Supplementary Item 6] The cell property prediction device of Supplementary Item 5, wherein the processor inputs the target cell information into a second machine learning model and causes the second machine learning model to output a prediction result of the property of the target cell. [Supplementary Item 7] The cell property prediction device of Supplementary Item 5 or 6, wherein the target cell information includes at least one of genetic data of the target cell, culture data of the target cell, and shape data of the target cell. [Supplementary Item 8] The cell property prediction device of any one of Supplementary Items 1 to 7, wherein there are a plurality of tools for predicting the properties of the target cell, and the processor selects a compatible tool compatible with the subtype information from the plurality of tools, and performs the prediction using the compatible tool.[Supplementary Item 9] The cell property prediction device of any one of Supplementary Items 1 to 8, wherein the processor presents a predicted result of the property of the target cell to a user. [Supplementary Item 10] The cell property prediction device of Supplementary Item 9, wherein the processor presents the subtype information to the user. [Supplementary Item 11] The cell property prediction device of Supplementary Item 9 or 10, wherein the processor presents a reliability of the prediction result to the user. [Supplementary Item 12] The cell property prediction device of any one of Supplementary Items 1 to 11, wherein the processor determines subsequent culture conditions for the target cell based on at least one of the subtype information and the predicted result of the property of the target cell. [Supplementary Item 13] The cell property prediction device of any one of Supplementary Items 1 to 12, wherein the host cell is a mammalian-derived cell. [Supplementary Item 14] The cell characteristic prediction device according to any one of Supplementary Items 1 to 13, wherein the target cell is a cell that produces a substance that will be an active ingredient in a biopharmaceutical, and the processor predicts the production stability of the substance as a property of the target cell. [Supplementary Item 15] The cell characteristic prediction device according to Supplementary Item 14, wherein the substance is an antibody. [Supplementary Item 16] The cell characteristic prediction device according to any one of Supplementary Items 1 to 15, wherein the processor predicts the culture conditions of the target cell as a property of the target cell.

[0127] The technology of the present disclosure can be appropriately combined with the various embodiments and / or various modified examples described above. Furthermore, it is not limited to the above embodiments, and various configurations can be adopted without departing from the spirit of the present disclosure. Furthermore, the technology of the present disclosure extends not only to programs, but also to storage media that non-temporarily store programs, and computer program products that include programs.

[0128] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[0129] In this specification, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed by connecting them with "and / or."

[0130] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

Claims

1. A cell characteristic prediction device for predicting the characteristics of a target cell, which is a cell obtained through monoclonalization from a host cell population, which is a collection of host cells having a plurality of different subtypes, comprising a processor, wherein the processor obtains subtype information representing a subtype of the target cell derived from a subtype of the host cells, and performs the prediction according to the subtype information.

2. The cell characteristic prediction device of claim 1, wherein the processor acquires target cell information regarding the target cell, discriminates a subtype of the target cell based on the target cell information, and acquires the discrimination result of the subtype of the target cell as the subtype information.

3. The cell characteristic prediction device of claim 2, wherein the processor inputs the target cell information into a first machine learning model and outputs the subtype discrimination result from the first machine learning model.

4. The cell characteristic prediction device according to claim 2, wherein the target cell information includes at least one of genetic data of the target cell, culture data of the target cell, and shape data of the target cell.

5. The cell characteristic prediction device of claim 1, wherein the processor acquires target cell information regarding the target cell, and predicts the characteristics of the target cell based on the target cell information.

6. The cell characteristic prediction device described in claim 5, wherein the processor inputs the target cell information into a second machine learning model and outputs a prediction result of the characteristics of the target cell from the second machine learning model.

7. The cell characteristic prediction device according to claim 5, wherein the target cell information includes at least one of genetic data of the target cell, culture data of the target cell, and shape data of the target cell.

8. The cell characteristic prediction device of claim 1, wherein there are a plurality of tools for predicting the characteristics of the target cell, and the processor selects a compatible tool from the plurality of tools that is compatible with the subtype information, and performs the prediction using the compatible tool.

9. The cell characteristic prediction device according to claim 1, wherein the processor presents a prediction result of the characteristics of the target cell to a user.

10. The cell characteristic prediction device according to claim 9, wherein the processor presents the subtype information to the user.

11. The cell characteristic prediction device according to claim 9, wherein the processor presents the reliability of the prediction result to the user.

12. The cell characteristic prediction device according to claim 1, wherein the processor determines subsequent culture conditions for the target cell based on at least one of the subtype information and the predicted results of the characteristics of the target cell.

13. The cell characteristic prediction device according to claim 1, wherein the host cell is a cell derived from a mammal.

14. The cell characteristic prediction device of claim 1, wherein the target cell is a cell that produces a substance that is an active ingredient of a biopharmaceutical, and the processor predicts the production stability of the substance as a characteristic of the target cell.

15. The cell characteristic prediction device according to claim 14, wherein the substance is an antibody.

16. The cell characteristic prediction device according to claim 1, wherein the processor predicts culture conditions of the target cell as a characteristic of the target cell.

17. A method for operating a cell characteristic prediction device that predicts the characteristics of a target cell, which is a cell obtained through monoclonalization from a host cell population, which is a collection of host cells having a plurality of different subtypes, comprising: obtaining subtype information representing the subtype of the target cell derived from the subtype of the host cells; and performing the prediction in accordance with the subtype information.

18. An operating program for a cell characteristic prediction device that predicts the characteristics of a target cell, which is a cell obtained through monoclonalization from a host cell population, which is a collection of host cells having a plurality of different subtypes, the operating program for the cell characteristic prediction device causing a computer to execute processes including: obtaining subtype information representing the subtype of the target cell derived from the subtype of the host cells; and making the prediction according to the subtype information.

Images