Culture condition optimization method and arithmetic processing device

Abstract

Provided are a culture condition optimization method and an arithmetic processing device that can optimize culture conditions when culturing cells using a culture medium. In this arithmetic processing device 5, from among a plurality of culture medium constituents identified on the basis of quantification results from a quantification step, culture medium constituents that are correlated with culture results obtained in a culture results acquisition step are selected by a feature quantity extraction module 10 as important constituents. The arithmetic processing device 5 presents, as choice candidate constituents, culture medium constituents which are other than the important constituents selected in the selection step and which should be focused on when cells are cultured. The arithmetic processing device 5 allows a user to choose, from among the important constituents and the choice candidate constituents, an important constituent and a choice candidate constituent that are to be optimized, as optimization constituents, sets concentration conditions for each of the optimization constituents, and formulates culture conditions that satisfy the concentration conditions and include the optimization constituents on the basis of an optimization module 13 (formulation step).

Description

Method for optimizing culture conditions and processing device

[0001] The present invention relates to a method for optimizing culture conditions and a processing device.

[0002] Conventionally, systems have been developed that combine laboratory robots and Bayesian optimization to optimize experimental procedures (see, for example, non-patent literature). This non-patent literature aims to realize an autonomous search for optimal culture conditions using experimental design by combining an AI system that independently evaluates experimental results and plans the next experiment.

[0003] "Robotic search for optimal cell culture in regenerative medicine" Kanda, Tsuzuki et al, eLife 2022;11:e77007.DOI [searched May 20, 2020], Internet <URL: https: / / doi.org / 10.7554 / eLife.77007>

[0004] However, when optimizing the culture conditions and medium composition for cell culture, there are a large number of parameters to be considered, including medium components as well as environmental factors such as temperature, making it difficult to efficiently obtain a solution using experimental design depending on the number of parameters to be optimized. Therefore, there is a need to develop new methods for optimizing the culture conditions when culturing cells using medium.

[0005] The present invention has been made in consideration of the above problems, and aims to provide a method for optimizing culture conditions and a processing device that can optimize culture conditions when culturing cells using a culture medium.

[0006] The method for optimizing culture conditions of the present invention is a method for optimizing culture conditions when culturing cells using a culture medium, and includes: a culture result acquisition step for acquiring culture results obtained when culturing the cells using the culture medium under predetermined culture conditions; a selection step in a computing device for selecting, as important components, from a plurality of culture medium components using a feature extraction module, those culture medium components that are correlated with the culture result; a presentation step in the computing device for presenting, as selection candidate components, medium components other than the important components selected in the selection step that should be noted when culturing the cells; a selection step in the computing device for selecting, as optimization components, the important components and the selection candidate components to be targeted for optimization from the important components and the selection candidate components; and a formulation step in the computing device for formulating the culture conditions including the optimization components using an optimization module.

[0007] According to the present invention, it is possible to provide a method for optimizing culture conditions and a processing device that can optimize culture conditions when culturing cells using a medium.

[0008] 10 is a block diagram showing the overall configuration of a culture condition optimization system according to the present invention. FIG. 11 is a flowchart showing the optimization process procedure for culture conditions according to the first embodiment. FIG. 12 is a flowchart showing the optimization process procedure from step S15 onwards. FIG. 13 is a table showing the culture conditions when a verification test was conducted. FIG. 14 is a table showing the culture results. FIG. 15 is a table showing the importance list created in the verification test. FIG. 16 is a table showing the optimization component list created in the verification test. FIG. 17 is a table showing the concentration of each medium component in the "1x concentration feed medium", the concentration of each medium component in the "7-variable optimization", and the concentration of each medium component in the "13-variable optimization". FIG. 18 is a table showing the antibody production amount obtained as a result of a culture test using the medium shown in FIG. 8. FIG. 19 is a flowchart showing the optimization process procedure for culture conditions according to the second embodiment. FIG. 19 is a flowchart showing the optimization process procedure for culture conditions according to the third embodiment.

[0009] An embodiment of the present invention will be described in detail below with reference to the drawings. In the following description, the same components are designated by the same reference numerals, and duplicated descriptions will be omitted.

[0010] (1) First Embodiment <Configuration of Culture Condition Optimization System> FIG. 1 is a block diagram showing the overall configuration of a culture condition optimization system 1. The culture condition optimization system 1 according to this embodiment includes an analyzer 2, a culture result measurement device 3, a processor 5, a database 6, an operation unit 7, and a display unit 8. The analyzer 2 is a mass spectrometer such as a liquid chromatograph mass spectrometer (LC-MS) or a gas chromatograph mass spectrometer (GC-MS), which quantifies medium components in a culture medium and measures the abundance of compounds (medium components) in the medium used for cell culture. The analyzer 2 outputs mass spectrometry data obtained as the quantification results to the processor 5. Note that the input to the processor 5 does not have to be the quantification results obtained by the analyzer 2; data such as the medium components and their concentrations or composition ratios may be prepared in advance and input to the processor 5.

[0011] Furthermore, two or more quantitative results obtained by the analyzer 2 may be input to the arithmetic processing device 5. For example, the same sample (culture medium) may be quantified multiple times at intervals, and the multiple quantitative results obtained in time series may be used for analysis. Alternatively, medium components may be quantified once or more for each of different samples (culture medium), and the multiple quantitative results obtained may be used for analysis. This allows the arithmetic processing device 5 to identify the medium components present in the culture medium in which cells are cultured (which may be absolute values ​​such as concentrations, or relative values ​​such as the percentage content of each component) based on the mass spectrometry data.

[0012] The culture result measuring device 3 is a measuring device such as a liquid chromatograph (LC), an ultraviolet-visible spectrophotometer, an automated culture environment analyzer also equipped with a viable cell analysis function such as BioProfile FLEX2 (Nova Biomedical), an automated culture environment analyzer also equipped with an antibody quantification function such as Cedex Bio (Roche Diagnostics), or a viable cell analyzer such as Vi-Cell (Beckman Coulter), and measures the viability and viable cell density of the cultured cells, the amount of a target substance (here, an antibody produced by the cells) present in the culture supernatant, and a quality specification value of the target substance (e.g., the proportion of antibodies having a desired glycostructure, etc.) as measurement results. The culture result measuring device 3 outputs the measurement results of antibody production amount and viable cell density (here, antibody production amount) obtained as measurement results to the calculation processing device 5 as culture results.

[0013] The input to the arithmetic processing device 5 does not have to be the quantitative results acquired by the culture result measuring device 3, and data known from prior knowledge such as the type of cultured cells or the type of culture medium may be input. Furthermore, a set of values ​​obtained by measuring one or more types of culture supernatants at least once using one or more types of culture result measuring devices 3 may be input as the culture result to the arithmetic processing device 5. The elements constituting the culture result are not limited to scalar quantities such as the amount of antibody present, but may also be character strings representing vector quantities such as spectral data acquired by the analyzer 2 or the like, or the type of measurement target.

[0014] The database 6 is a hard disk drive (HDD), flash memory, or the like, which is capable of saving and reading data by writing it, and is connected to the arithmetic processing device 5. The database 6 saves and reads data in response to requests from the arithmetic processing device 5, and transmits and receives data to and from the arithmetic processing device 5. The database 6 stores various data such as mass spectrometry data and culture results output to the arithmetic processing device 5, data used for arithmetic processing in the arithmetic processing device 5, arithmetic processing results obtained by the arithmetic processing device 5, and the like.

[0015] The operation unit 7 is an operation input unit that allows the user to operate the arithmetic processing device 5, and is composed of, for example, switches and knobs provided on the arithmetic processing device 5, or a keyboard and mouse. When the user operates the operation unit 7 to input various execution instructions, the arithmetic processing device 5 executes various arithmetic processes based on predetermined software in accordance with the execution instructions.

[0016] The display unit 8 is a means for displaying information such as the operating status of the arithmetic processing device 5, a list of selectable items, mass spectrometry data acquired from the analyzer 2, culture results acquired from the culture result measuring device 3, arithmetic processing results obtained by the arithmetic processing device 5, and various data read from the database 6, and is configured, for example, by a display. Based on the various information displayed on the display unit 8, the user can check the medium components, the antibody production amount (culture results), the importance list (described later in FIG. 6 ), the optimized component list (described later in FIG. 7 ), the optimized culture conditions that are the arithmetic processing results, and the like.

[0017] The arithmetic processing device 5 has a computer configuration including a memory for storing programs and a processor that executes the programs, which are software that realizes each function. When the processor executes the programs, the arithmetic processing device 5 causes various circuit units to function as a feature extraction module 10, an optimization module 13, etc., to perform various arithmetic processing. In this case, the arithmetic processing device 5 includes an acquisition unit 9, a feature extraction module 10, a concentration condition setting unit 12, and an optimization module 13. The feature extraction module 10 includes a data processing unit 101, an important component selection unit 102, a list generation unit 104, a group setting unit 105, and a presentation processing unit 106. The arithmetic processing device 5 controls the feature extraction module 10, etc., to perform optimization processing of the culture conditions.

[0018] <Culture Condition Optimization Processing Procedure> Next, the culture condition optimization processing procedure executed by the arithmetic processing device 5 will be described with reference to the flowcharts in Figures 2 and 3. In Figures 2 and 3, the main tasks manually performed by the user are indicated in the box labeled "User's Tasks," and the main arithmetic processing executed by the arithmetic processing device 5 is indicated in the box labeled "Arithmetic Processing Device."

[0019] In step S1, an executing entity in an execution environment cultures cells using a culture medium under predetermined culture conditions (culture step). Here, the execution environment refers to an environment in which a culture test is actually performed, such as a factory or a laboratory. The executing entity in the execution environment performs a culture test in which cells are cultured in a culture medium under predetermined culture conditions and in a set procedure.

[0020] In the culture test in step S1, cells are cultured in a medium to produce antibodies from the cells, and the antibodies present in the culture supernatant of the medium are obtained as target substances. The execution entity varies depending on the type of culture test, and may be, for example, a user (human), or a mechanical device operated by a user, a robot arm, an incubator, an automated laboratory device, etc.

[0021] The culture conditions initially defined in the culture step in step S1 are randomly set by the arithmetic processing device 5 or the user.

[0022] The concentration conditions of the additive components (feed medium components) to be added to the medium during the culture test are output as additive component concentration conditions to the arithmetic processing device 5. As shown in Fig. 1 , the arithmetic processing device 5 includes an acquisition unit 9 connectable to an external device, and acquires the additive component concentration conditions via the acquisition unit 9 from, for example, an external device or the like in which the additive component concentration conditions are stored, and stores the acquired conditions in the database 6 (additive component concentration condition acquisition step).

[0023] In step S2, the user measures a specific target substance obtained when culturing the cells using the culture result measuring device 3 (culture result measuring step). Here, the culture result measuring device 3 measures the production amount (antibody concentration) of the antibody obtained by the culture, and outputs the obtained measurement result (antibody production amount) as the culture result to the arithmetic processing device 5. As a result, the arithmetic processing device 5 acquires the culture result from the culture result measuring device 3 (culture result acquiring step).

[0024] In step S3, the user obtains quantitative results of the medium components in the culture supernatant obtained by culturing cells using the analyzer 2 (quantification step). The analyzer 2 outputs the obtained mass spectrometry data as the quantitative results to the arithmetic processing device 5. As a result, the arithmetic processing device 5 obtains the quantitative results from the analyzer 2 (quantification result obtaining step).

[0025] In the feature extraction module 10, the data processing unit 101 combines the added component concentration conditions, the culture results, and the mass spectrometry data, performs preprocessing (correction of notation variations, normalization of measurement data, completion of missing data, and deletion of unnecessary data), and then performs range selection and filtering (narrowing down of data by culture date and time, ammonia concentration, pH, osmotic pressure, viable cell rate, etc.).

[0026] In step S4, the feature extraction module 10 uses the important component selection unit 102 to identify medium components contained in the culture supernatant based on the mass spectrometry data obtained in the quantification step, and selects, from the identified plurality of medium components, medium components that are relatively highly correlated with the culture results obtained in the culture result measurement step as important components (selection step). Here, a medium component that is relatively highly correlated with the culture results can be defined as, for example, a medium component whose index indicating correlation with the culture results is large when compared to a certain threshold, or a medium component whose correlation with the culture results is in the top a% (a is a predetermined value) when the correlation of the medium component with N factors is calculated. Here, medium components that are suspected to contribute to increased antibody production (increased antibody concentration) are selected as important components. The important component selection unit 102 uses multivariate analysis to select, from the plurality of medium components identified based on the mass spectrometry data in the quantification step, medium components that are suspected to contribute to increased antibody production.

[0027] Note that the object (objective variable) compared with the medium components in the selection step is not limited to the culture results themselves. The objective variable may be the culture results after a calculation such as logarithmic transformation, or a new element generated by performing some calculation (e.g., similarity calculation) on two or more elements contained in the culture results. The following cases are examples of using the similarity calculation described above as the objective variable. For example, if the quantification result of strain A1, which is being cultured very well, is X(A1), and the quantification result of strain B1, which is being cultured poorly, is X(B1), the objective is to maximize the productivity of strain B. In this case, the working hypothesis is that the productivity of strain B may be improved by making the state of strain B1 similar to that of strain A1. Therefore, the similarity r(AB) between the quantification result X(A1) of strain A1 and the quantification result X(B1) of strain B1 may be used as the objective variable, and a key medium that maximizes this similarity may be determined by multivariate analysis.

[0028] In addition, methods for obtaining the correlation between the culture results and medium components in the selection step include methods for obtaining the correlation between two variables such as the Pearson product-moment correlation coefficient and the maximum information content coefficient, methods for analyzing the variable importance of linear regression such as partial least squares (PLS), ElasticNet regression, and canonical correlation analysis, methods for analyzing the variable importance of nonlinear regression such as neural networks and xgboost, and methods for analyzing the variable importance of regression methods that consider multiple target variables (multitask regression).

[0029] In the multivariate analysis according to this embodiment, for example, the culture results indicating the amount of antibody production are used as the objective variable, and mass spectrometry data obtained by quantifying multiple medium components contained in the culture supernatant obtained by cell culture are used as the explanatory variables, and by applying multivariate regression (linear regression or nonlinear regression) or correlation analysis, it is possible to analyze the medium components that affect the increase in antibody production from among multiple medium components.

[0030] As an example of multivariate regression, naive partial least squares (PLS) regression can be applied, in which the culture results indicating the amount of antibody production are used as the response variable and the mass spectrometry data obtained by quantifying the medium components are used as the explanatory variables.As an example of correlation analysis, comprehensive calculation of Pearson product-moment correlation coefficients for each component can be applied, in which the culture results indicating the amount of antibody production are used as the response variable and each component of the mass spectrometry data obtained by quantifying the medium components is used as the explanatory variables.

[0031] The feature extraction module 10 according to this embodiment calculates the variable importance and correlation coefficient for each medium component quantified in the quantification step with respect to the target variable (antibody production amount) using, for example, two different types of multivariate analysis, i.e., multivariate regression and correlation analysis. Based on the obtained variable importance and correlation coefficient, the feature extraction module 10 identifies medium components that are predicted to contribute to increased antibody production and selects them as important components. For example, the feature extraction module 10 may determine importance based on the variable importance and correlation coefficient, and then rank and select medium components with high importance as important components. By selecting medium components with common coefficients from the analysis results obtained from multiple different types of multivariate analysis in this way as important components, the reliability of the selection results can be improved.

[0032] In this embodiment, two multivariate analyses are applied to determine the importance and to select important components from among the medium components. However, the present invention is not limited to this. The importance may be determined from only one multivariate analysis, or the importance may be determined by applying three or more multivariate analyses to select important components from among the medium components.

[0033] Here, importance refers to the degree to which a component is presumed to be correlated with the culture results (for example, "an index showing the correlation with the culture results is large when compared with a certain threshold" or "when the correlation of a medium component with N factors is calculated, the correlation with the culture results is in the top a% (a is a predetermined value)"). The higher the index value, such as variable importance, regression coefficient, or correlation coefficient, which is the analysis result of multivariate analysis, the more likely it is to be correlated with the culture results (contributing to increased antibody production). Therefore, a medium component with a higher importance can be said to be a more effective medium component for increasing antibody production, and is an important medium component that is correlated with the culture results.

[0034] For example, suppose there is a certain relationship between the culture results of a culture test and each medium component, and when the conditions (concentration of each medium component) are varied while satisfying that relationship, a high importance level is obtained (a case in which an evaluation result for the culture results of the medium component can be obtained). In such a case, for multiple medium components that are ranked high in importance, the interrelationships between each medium component can be learned using a feature transformation function, and then the culture results and medium components can be sequentially optimized in the space specified by the feature transformation function, thereby allowing for the search for medium components with high importance.

[0035] Here, as an example of calculating the importance of medium components using multivariate analysis, when conducting a culture test, two or more media with the same composition but different concentrations of the medium components are prepared, and the antibody production amount obtained when cells are cultured in each of the two or more prepared media is measured. Then, the feature extraction module 10 uses the measured antibody production amount (culture result) as the objective variable and performs multiple multivariate analyses (naive PLS regression and correlation analysis) to compare the medium components between the two or more media with different concentrations of the medium components. For each of the multiple multivariate analyses, the feature extraction module 10 identifies medium components that commonly cause differences in the culture results between the two or more media with different concentrations of the medium components.

[0036] The important component selection unit 102 according to this embodiment uses naive PLS regression and correlation analysis to obtain index values ​​(e.g., variable importance and correlation coefficient) representing the strength of the relationship between antibody production and each medium component for each medium component. It can be inferred that the larger the variable importance value, the more likely it is to contribute to increased antibody production. It can be inferred that the larger the positive correlation coefficient value, the more likely it is to be positively correlated with increased antibody production, and the larger the negative correlation coefficient value, the more likely it is to be negatively correlated with increased antibody production. For example, by sorting the medium components in descending order of correlation coefficient value, the important component selection unit 102 can sort the important medium components that are likely to contribute to increased antibody production in order of importance as important components. After creating an importance list in which the medium components are sorted in order of importance, the important component selection unit 102 outputs the list to the presentation processing unit 106. In addition, by presenting the variable importance obtained for each medium component separately from the correlation coefficient, the user can determine the importance using not only the correlation coefficient but also the variable importance as a guide and select important components.

[0037] In this embodiment, different index values ​​representing medium components correlated with the culture results are calculated for the same culture results using different multivariate analyses, and the importance is determined by one of the index values ​​or a combination of multiple index values, and the medium component with the highest importance score is preferably selected as the important component.

[0038] Figure 6 shows an example of an importance list. In the importance list shown in Figure 6, the "substance_name" column shows examples of medium components identified based on mass spectrometry data, and "corr day5" shows examples of correlation coefficients for medium components obtained by correlation analysis from the medium on day 5. "corr day6" shows examples of correlation coefficients for medium components obtained by correlation analysis from the medium on day 6. "corr day7" shows examples of correlation coefficients for medium components obtained by correlation analysis from the medium on day 7.

[0039] "vip5" shows an example of the variable importance for prediction (VIP) scores for each medium component obtained from the culture medium on day 5 by naive PLS regression, "vip6" shows an example of the VIP scores for each medium component obtained from the culture medium on day 6 by naive PLS regression, and "vip7" shows an example of the VIP scores for each medium component obtained from the culture medium on day 7 by naive PLS regression.

[0040] In the importance list shown in FIG. 6 , medium components are listed in descending order of correlation coefficient on the fifth day after the start of culture. The important component selection unit 102 selects important components from the multiple medium components listed in the importance list based on predetermined criteria. Examples of criteria include determining important components based on a predetermined correlation coefficient value, or automatically determining the top several medium components on the importance list as important components. Furthermore, a threshold value may be set for the VIP score, and medium components with a VIP score equal to or greater than the threshold may be selected as important components. This example shows an example in which the top seven variables (medium components No. 1 to No. 7) on the importance list are selected as important components based on the correlation coefficient value on the fifth day after the start of culture.

[0041] In step S5, the presentation processing unit 106 presents, as candidate components for selection, medium components other than the important components that are predicted to contribute to an increase in antibody production (presentation step). In this embodiment, the medium components No. 1 to No. 7 in the importance list shown in Figure 6 are highlighted as important components, and medium components No. 8 and onward are presented as candidate components for selection, distinguishing them from the other medium components.

[0042] In the above-described embodiment, all medium components other than the important components selected from the importance list are presented as candidate components for selection, but the present invention is not limited to this. In another embodiment, the presentation processing unit 106 may automatically present, as candidate components for selection, one or more medium components that may contribute to increasing antibody production, from among the medium components other than the important components selected in the importance list of Fig. 6, in accordance with predetermined selection rules based on domain knowledge information.

[0043] In addition, for example, the presentation processing unit 106 may automatically select medium components other than those in the importance list as candidate components for selection in accordance with predetermined selection rules based on domain knowledge information stored in the database 6, and present these multiple candidate components for selection to the user.

[0044] The domain knowledge information is information collected based on information such as publicly known literature, academic papers, patent publications, and published patent publications that contain knowledge about cell culture, such as the same type of cells, the same type of antibodies, the same type of culture medium, and the same type of culture conditions (pH, temperature, culture time, etc.) used in the culture test, and medium components that are presumed to have an effect on the increase in antibody concentration during the culture step are identified in advance.

[0045] Note that the selection criteria for selecting candidate components for selection from known medium components are preferably defined to select, for example, medium components already known to contribute to an increase in the production yield of the same type of antibody, medium components already known to contribute to an increase in the number of host cells producing the antibody, medium components that affect the profiling results of glycosylation structural isomers and charge isomers that are important for antibody quality, medium components that contribute to the reduction of host-derived proteins (HCPs) that serve as contaminants, medium components that contribute to the reduction of growth inhibitors such as ammonia and lactic acid, medium components that contribute to the viability of cells, etc. Furthermore, the number of candidate components for selection read from the domain knowledge information may be a predetermined number, or may be all of the relevant candidate components for selection.

[0046] The presentation processing unit 106 outputs an importance list in which the plurality of medium components in the importance list are sorted into important components and selection candidate components to the display unit 8. The display unit 8 displays the received importance list, allowing the user to visually confirm these important components and selection candidate components.

[0047] In step S6, the optimization system 1 prompts the user to select an optimized component from the selected important components (selection step). In step S7, the optimization system 1 prompts the user to select an optimized component from the presented selection candidate components (selection step). In the selection steps of steps S6 and S7, the optimized component may be selected from the important components and the selection candidate components based on the user's domain knowledge, or the selection may be performed automatically by the presentation processing unit 106 based on predetermined selection criteria. Examples of selection criteria include setting a threshold for the important components based on variable importance or correlation coefficient and selecting them in descending order, and ranking the selection candidate components in advance by the user and selecting those with the highest rankings, or selecting components that have been reported in large numbers in publicly known literature, academic papers, patent publications, published patent applications, etc.

[0048] Furthermore, in step S8, the optimization system 1 allows the user to select, from among the important components and other medium components not selected as candidate components for selection, other medium components that are inferred to be important based on the user's domain knowledge as optimization components (selection step).

[0049] In each of the selection steps of steps S6, S7, and S8, an optimized component is selected based on the user's operation of the operation unit 7. Information on the optimized component selected in the selection steps of steps S6, S7, and S8 is output to the list generation unit 104.

[0050] In step S9, the feature extraction module 10 creates an optimization component list in which the optimization components selected in the selection step are sorted in order of priority using the list generation unit 104. The method of determining the priority of the optimization components executed by the list generation unit 104 may, for example, specify that the larger the correlation coefficient of the optimization component selected from the importance list, the higher the priority, or the user may freely determine the priority of the optimization component using the operation unit 7 at their own discretion. Furthermore, priorities may also be specified in advance between the selection steps of steps S6, S7, and S8, for example, so that the priority of the optimization component selected in the selection step of step 6 is higher than the priority of the optimization component selected in the selection step of step S7.

[0051] Here, Fig. 7 shows an example of an optimization component list. The optimization component list shown in Fig. 7 shows an example in which all of the important components of the top seven variables selected in the importance list shown in Fig. 6 are selected as optimization components in step S6. The seven variables No. 1 to No. 7 in the optimization component list are the top seven variables in the importance list shown in Fig. 6.

[0052] 7 shows an example in which four variables No. 10, No. 15, No. 17, and No. 19 are selected by the user as optimization components from among the candidate components other than the important components in the importance list in step S7. The four variables No. 8 to No. 11 in the optimization component list are candidate components No. 10, No. 15, No. 17, and No. 19 in the importance list shown in FIG.

[0053] 7 shows an example in which glucose, which was not selected as an important component or a candidate component in step S8, is selected as the optimized component based on the user's domain knowledge. The single variable No. 12 in the optimized component list indicates the optimized component selected in step S8.

[0054] Next, in step S10, the optimization system 1 prompts the user to set the number of optimization components (the number of optimization variables) for formulating culture conditions in a formulation step executed by the optimization module 13, which will be described later. The number of optimization variables is set based on the user's operation of the operation unit 7.

[0055] In step S11, the feature extraction module 10 generates grouped components by grouping multiple optimized components in the optimized component list with one variable, according to a predetermined grouping rule, using the group setting unit 105 (group generation step). The optimized component list shown in FIG. 7 shows an example in which a lower-level variable with low importance in the importance list is used as a grouped component. As a result, the group setting unit 105 creates an optimized component list consisting of multiple optimized components with higher priorities and the grouped components.

[0056] As a grouping definition for grouping multiple optimized components as one variable, it is desirable to define as grouping components medium components that have the same at least one of the following: the attributes (categories) of the medium components, the concentrations of the medium components quantified by the analysis device 2, the time-dependent change trends of the quantification results of the medium components quantified by the analysis device 2 when culturing cells (pattern of increase or decrease in components during culture), the degree of correlation with the culture results obtained in the culture result measurement step, and metabolic pathways.

[0057] Next, in step S12, the optimization system 1 sets the concentration ranges (upper and lower concentration limits) of each optimization component and grouping component as concentration conditions in the optimization component list created in step S11, for the number of optimization variables in step S10 (concentration condition setting step). In this example, the concentration conditions are set based on the user's operation of the operation unit 7 (such as numerical input or selection). A grouping component is a combination of multiple optimization components considered as a single medium component, and a grouping component is considered as one variable, with one lower concentration limit and one upper concentration limit set for the grouping component.

[0058] The lower and upper concentration limits of the optimization component and grouping component are set arbitrarily by the user, and the determination method may be based on preliminary testing or preliminary studies, or may take into account the concentrations of each medium component in commercially available media and the convenience of the culture test (concentrations that can be produced), or may be determined based on domain knowledge, or may be determined based on the user's previous experimental experience.

[0059] Next, as shown in FIG. 3 , in step S15, the optimization system 1 prompts the user to set the number of parallel experiments. The number of parallel experiments indicates the number of culture experiments with different culture conditions that can be performed in parallel. For example, if "2" is set as the number of parallel experiments, this indicates that two culture tests with different culture conditions can be performed in parallel at the same time, and that culture experiments can be performed under two conditions at the same time. The number of parallel experiments is set based on the user's operation of the operation unit 7.

[0060] Next, in step S16, the optimization module 13 formulates culture conditions for the culture step that satisfy the concentration conditions set in the concentration condition setting step based on the experimental design generation process (formulation step). The optimization module 13 performs multivariate regression analysis as the experimental design generation process, and can determine, by calculation, the currently optimal culture conditions that are expected to increase the antibody production amount in the culture test.

[0061] Here, the culture test involves, for example, steps such as medium preparation, cell cultivation using the medium, and target substance extraction, which are performed in a predetermined order. In each step of medium preparation, cell cultivation, etc., various culture conditions are determined, such as the composition of the basal medium and feed medium, the concentration of each medium component, pH, culture period, culture temperature, humidity, etc. Since the culture conditions, such as the composition of the basal medium and feed medium and the concentration of each medium component, that maximize the gain obtained from the culture test are not obvious, it is necessary to repeatedly perform a large number of culture experiments using a naive method and through trial and error.

[0062] The more complex the culture test, the more complex and multiple the culture conditions become, making it more difficult to explore various culture conditions and find the optimal one. Even in small-scale, multi-variety culture tests, automation is being attempted using robots, but due to differences in the type of target substance and individual differences in cells, it becomes necessary to find the optimal culture conditions for each case, which results in a high total cost for exploring culture conditions.

[0063] Generally, when searching for culture conditions, as the number and types of entities performing culture tests, such as mechanical devices, robots, and users (humans), increase, it becomes necessary to sequentially generate appropriate culture conditions for each entity, which requires a great deal of effort.

[0064] The experimental design generation process executed by the optimization module 13 according to this embodiment takes the above points into consideration and enables the execution entity to identify, with a small number of experiments, optimal culture conditions that will yield as great a gain as possible (here, an increase in antibody production) when actually executing a culture test in an execution environment according to a predetermined process.

[0065] Here, we will explain the experimental design generation process executed by the optimization module 13. The optimization module 13 stores experimental design information that specifies multiple steps to be performed in sequence to culture cells for each of the various culture tests performed in step S1 (e.g., a culture medium preparation step, a cell culture step using the prepared culture medium, etc.). The experimental design information specifies various culture conditions for each step as necessary, such as the composition of medium components, the concentration of each medium component, pH, culture period, culture temperature, and humidity, and the user can perform culture tests based on the culture conditions specified in the experimental design information.

[0066] The experimental design information specifies that the concentration value of each medium component, which is one of the culture conditions, can be varied. The optimization module 13 sets the optimization component and grouping component identified by the feature extraction module 10 to the corresponding item (item indicating the medium component) in the experimental design information. The optimization module 13 also sets the concentration conditions, i.e., the lower and upper limit values ​​of the optimization component and grouping component set in the concentration condition setting step, as a concentration search range in the item indicating the concentration of the medium component in the experimental design information.

[0067] Here, the optimization module 13 selects multiple concentrations from within the respective concentration search ranges for each of the optimization component and the grouping component according to a given procedure (e.g., Bayesian optimization, orthogonal array, Latin hypercube, etc.) based on the culture conditions and culture results of one or more past culture tests. At this time, the concentrations may be selected by projecting the culture conditions of the past culture tests onto a search space representing the respective concentration search ranges for the optimization component and the grouping component.

[0068] In the optimization module 13 according to this embodiment, for example, a regression model (response surface) is generated in advance from the culture conditions and culture results of one or more past culture tests. The optimization module 13 uses this regression model to select one or more concentrations from within the concentration search range, for example, by Bayesian optimization or multitask Bayesian optimization. The number of concentrations selected at one time from within the concentration search range by the optimization module 13 is preferably the same as the number of parallel experiments set in step S15.

[0069] The optimization module 13 generates experimental design information by writing the concentrations of each optimization component and each grouping component selected by the regression model into corresponding items in the experimental design information. Since the optimization module 13 selects the same number of concentrations for each optimization component and grouping component as the number of parallel experiments using the regression model, the optimization module 13 generates the same number of experimental design information as the number of parallel experiments. In this way, the optimization module 13 generates multiple pieces of experimental design information with different concentrations for each optimization component and grouping component. As described above, in the formulation step, the optimization module 13 formulates multiple culture conditions and presents each of the formulated culture conditions to the execution entity as experimental design information.

[0070] The optimization module 13 outputs the experimental design information to the display unit 8 and causes the contents of the experimental design information to be displayed on the display unit 8. This allows the user to check each piece of experimental design information displayed on the display unit 8, and in step S1, a new culture test can be performed with the experimental parallel number in accordance with each piece of experimental design information.

[0071] In this way, in the formulation step, the arithmetic processing device 5 determines, using the optimization module 13, concentrations that satisfy the concentration conditions estimated to be necessary to obtain the expected culture results, and presents the determined concentrations to the user (concentration condition presentation step). At this time, the arithmetic processing device 5 allows the user to change the concentrations presented in the concentration condition presentation step to other concentration conditions (change step). This allows the user to modify the contents of each experimental design information displayed on the display unit 8 based on their own domain knowledge. Note that changes to concentrations in the experimental design information can be set based on the user's operation of the operation unit 7 (such as by entering or selecting a value).

[0072] Then, again in step S2, the user measures the antibody production amount obtained as a result of a culture test performed under culture conditions in which the concentrations of each optimization component and grouping component have been optimized, using the culture result measuring device 3. The culture result measuring device 3 outputs the obtained measurement result to the calculation processing device 5 as the culture result. Then, again in step S3, the user quantifies the medium components in the culture supernatant in which the cells were cultured, using the analysis device 2. The analysis device 2 outputs the obtained mass spectrometry data to the calculation processing device 5.

[0073] As described above, in the feature extraction module 10, the data processing unit 101 combines the added component concentration conditions obtained in step S16, the culture results obtained in step S2, and the mass spectrometry data obtained in step S3, and performs preprocessing and range selection / filtering processing.

[0074] Next, in step S19, the arithmetic processing device 5 performs a determination process (determination step) of whether or not it is necessary to repeat the experimental design generation process of step S16, using the optimization module 13. In the determination step, for example, the culture result obtained in the culture result measurement step based on the previously executed culture step is used as the determination criterion, and when the culture result obtained in the newly executed culture step is obtained, it is determined whether or not the difference between the acquired culture result and the determination criterion exceeds a predetermined specified range.

[0075] The determination in the determination step as to whether the difference from the determination criterion exceeds a specified range is made based on, for example, whether improvement in the objective variable is no longer observed, whether the culture conditions have converged, whether the culture result has achieved a target improvement rate, etc. When the calculation processing device 5 determines that the difference between the culture result of the newly executed culture test in step S1 and the determination criterion exceeds a specified range, it notifies the user by displaying the completion of the experimental plan generation process on the display unit 8, assuming that a predetermined culture result has been obtained.

[0076] On the other hand, if the difference between the culture results of the newly executed culture test in step S1 and the judgment criteria does not exceed the specified range, the process returns to steps S15 and S16, and in step S19, the setting of the number of parallel experiments in step S15, the execution of the experimental plan generation process in step S16, the culture test by the user in step S1, the measurement of antibody production amount in step S2, and the quantification of culture medium components in step S3 are repeated until the difference between the culture results of the culture test and the judgment criteria exceeds the specified range.

[0077] <Actions and Effects> According to the above configuration, the method for optimizing culture conditions includes a culture step in which an executing entity cultures cells using a culture medium under predetermined culture conditions in an execution environment, a culture result measurement step in which the culture results obtained when the cells are cultured are measured by a culture result measurement device 3, and a quantification step in which medium components in the medium used for cell culture are quantified by an analysis device 2. The method for optimizing culture conditions also includes a culture result acquisition step in which the calculation processing device 5 acquires the culture results, and a quantification result acquisition step in which the calculation processing device 5 acquires the quantification results.

[0078] Furthermore, the computing device 5 selects, from among the multiple medium components identified based on the acquired quantification results, medium components that are correlated with the culture results acquired in the culture result acquisition step as important components using the feature extraction module 10 (selection step).The computing device 5 presents, as selection candidate components, medium components other than the important components selected in the selection step that should be noted when culturing cells (presentation step).

[0079] As a result, the arithmetic processing device 5 allows the user to select important components and selection candidate components to be targeted for optimization from among the important components and selection candidate components as optimization components (selection step).

[0080] The arithmetic processing device 5 generates grouping components by grouping some of the multiple optimization components selected in the selection step in accordance with predetermined grouping rules (group generation step), sets concentration conditions for the optimization component and the grouping components (concentration condition setting step), and formulates culture conditions including the optimization component and the grouping components that satisfy the concentration conditions set in the concentration condition setting step, based on the optimization module 13 (formulation step). In this way, the culture condition optimization method can optimize the culture conditions when culturing cells using a medium.

[0081] <Verification Test> Next, as a verification test, an actual importance list was created ( FIG. 6 ), and the optimized components and candidate components for selection were determined based on the importance list. An optimized component list was also created, setting multiple candidate components as grouping components ( FIG. 7 ). Next, the number of optimization variables (medium components) to be optimized by the optimization module 13 was determined from the optimized component list, and the concentrations of each optimization variable were optimized by the optimization module 13. Then, multiple media were prepared using the concentrations determined by changing the number of optimization variables in the optimization module 13. A culture test was performed using each medium, and a verification test was performed in which the obtained culture results were compared. The culture conditions for the culture test performed in this verification test are shown in FIG. 4.

[0082] The importance list was created as follows. Here, commercially available feed medium A powder (medium powder) was prepared, and a 1x concentration feed medium A was prepared using the specified amount of medium powder according to the instruction manual (negative control). Separately, the same feed medium A powder (medium powder) was prepared, and a 2x concentration feed medium A was prepared using twice the specified amount of medium powder according to the instruction manual.

[0083] The cell types were then cultured in the prepared medium according to the culture conditions shown in Figure 4, and the amounts of antibody produced in each culture supernatant were measured on days 5, 6, and 7 from the start of culture. The culture results shown in Figure 5 were obtained. Figure 5 shows the culture results when a 1x feed medium was used and when a 2x feed medium was used. Also in Figure 5, the column "Antibody production on Day 5 [g / L]" shows the amount of antibody produced on day 5 from the start of culture, the column "Antibody production on Day 6 [g / L]" shows the amount of antibody produced on day 6, and the column "Antibody production on Day 7 [g / L]" shows the amount of antibody produced on day 7.

[0084] Next, antibody production on Day 5 (Day 5) and Day 6 (Day 6) was compared between medium conditions using a 2x feed medium and medium conditions using a 1x feed medium using naive PLS regression and correlation analysis, respectively. Based on the correlation coefficients obtained by correlation analysis, medium components that showed differences between the two groups on Days 5 and 6 were selected as important components, and the others were designated as candidate components. An importance list was then created by sorting the correlation coefficients of the medium components obtained from the medium on Day 5 in descending order. The results are shown in Figure 6. In the importance list shown in Figure 6, seven variables, No. 1 to No. 7, were selected as important components. Furthermore, in this importance list, medium components other than the important components that require attention when culturing cells are listed as candidate components from No. 8 onward.

[0085] Next, based on the importance list shown in Figure 6, an optimized component list was created as shown in Figure 7. In the optimized component list shown in Figure 7, the user selected all of the top seven variables selected as important components in the importance list shown in Figure 6 as optimized components, resulting in seven variables No. 1 to No. 7 with increasing priority. In addition, in the optimized component list shown in Figure 7, the user selected candidate components No. 10, No. 15, No. 17, and No. 19 other than the important components in the importance list shown in Figure 6 as optimized components, resulting in four variables No. 8 to No. 11.

[0086] Furthermore, based on the user's domain knowledge, glucose, which was not presented as an important component or a candidate component in the importance list, was selected as the optimized component and made one variable No. 12 in the optimized component list.

[0087] It should be noted that glucose and glutamine, selected as the optimization components, are generally known to have a significant impact on antibody production yield and viable cell density in the culture and antibody production processes of animal cell lines, including CHO cells. This is common knowledge to the extent that the concentrations of these components added are always specified in culture protocols. For this reason, glucose and glutamine were added as optimization components in this verification test.

[0088] Furthermore, it is common knowledge that the selected optimized components, asparagine, aspartic acid, and glutamic acid, are closely related to glutamine metabolism. Findings strongly suggesting this are also found in intracellular metabolite tracer experiments using 13C-labeled substrates and CHO cell culture systems (https: / / doi.org / 10.1002 / bit.27993). Furthermore, in CHO cell culture systems, these three components have been used as substitutes for glutamine to suppress the production of ammonia (a component that adversely affects culture) during glutamine metabolism (https: / / doi.org / 10.1002 / btpr.1658). Therefore, in this validation study, asparagine, aspartic acid, and glutamic acid were added as optimized components.

[0089] The optimized component No. 13 in the optimized component list is a grouped component created by grouping together the eight candidate components in the importance list (No. 8, No. 9, No. 11 to No. 14, No. 16, and No. 18) at the discretion of the user. In this way, in this verification test, an optimized component list with a total of 13 variables was created, consisting of the top seven variables selected as important components in the importance list, five medium components other than the important components in the importance list that were arbitrarily selected based on the user's domain knowledge, and one variable created by grouping together the eight variables in the importance list.

[0090] Next, a lower concentration limit and an upper concentration limit were set for each optimized component in the optimized component list. One lower concentration limit and one upper concentration limit were set for each grouping component in the optimized component list. The lower concentration limit and upper concentration limit for each optimized component and grouping component were manually set by the user based on the user's domain knowledge.

[0091] Next, only the top seven variables in the optimization component list were set as optimization variables whose concentrations were optimized in the optimization module 13, and only the concentrations of these seven variables were optimized in the optimization module 13. The results are shown in Figure 8. The "1x concentration feed medium (negative control)" in Figure 8 shows the concentration of each medium component obtained by quantitatively analyzing a medium prepared with the medium powder in the amount specified in the instruction manual for commercially available feed medium A using a liquid chromatography mass spectrometer (LC-MS) (negative control). Note that the concentration of "glucose" at 10 [g / L] was determined by the user.

[0092] In the "7-variable optimization" of Fig. 8, the concentrations of only the top 7 variables in the optimization component list were optimized by the optimization module 13, and the remaining variables were set to the same concentrations as the "1x concentration feed medium (negative control)". The 8 variables that were group constituent components that make up the grouping components were also set to the same concentrations as the "1x concentration feed medium (negative control)".

[0093] In optimization module 13, a multivariate regression showing the correlation between medium components and culture results, which was generated based on the concentration of each medium component and the culture results of previous cell cultures using a medium containing that medium component, was used to determine the optimal concentration for each of the top seven variables from within the range of each concentration condition of the top seven variables (search concentration range).

[0094] 8, the "13 variable optimization" determines the concentration of each of the 13 variables in the optimization component list by the optimization module 13. Here, for the concentration of the grouping component, the optimization module 13 determines one concentration that is estimated to be optimal from the range of concentration conditions from the lower limit value to the upper limit value.

[0095] The concentrations of the grouping components were identified by using the optimization module 13 to search for concentrations using multivariate regression for medium components with the same attributes as the grouping components (8 variables).

[0096] In this verification test, in the "13-variable optimization," the user presets individual concentration ratios for each group constituent component that makes up the grouping component relative to the concentration of the grouping component. The individual concentrations of each group constituent component were calculated individually from the concentration of the grouping component based on the preset concentration ratios.

[0097] Next, three culture media (n=3) were prepared, each with a concentration specified by "1x feed medium (negative control)" in Figure 8, a medium with a concentration specified by "7-variable optimization," and a medium with a concentration specified by "13-variable optimization," and a culture test was performed under the culture conditions shown in Figure 4. As a result, the culture results shown in Figure 9 were obtained. Figure 9 shows the measurement results of the antibody production amount [g / L] in the culture supernatant when a culture test was performed using each of the media shown in Figure 8.

[0098] As shown in Figure 9, it was confirmed that the "7-variable optimization" and "13-variable optimization" medium conditions increased antibody production amounts compared to the medium condition of "1x concentration feed medium (negative control)." It was also confirmed that the "13-variable optimization" medium conditions increased antibody production amounts compared to the "7-variable optimization" medium conditions.

[0099] (2) Second Embodiment In the first embodiment described above, a method for optimizing culture conditions was described that includes both a presentation step of presenting, as selection candidate components, medium components other than the important components that should be noted when culturing cells, and a group generation step of generating grouped components by grouping some of the multiple optimization components. However, the present invention is not limited to this. For example, the method may also include a method for optimizing culture conditions that does not include the group generation step.

[0100] <Culture Condition Optimization Processing Procedure According to the Second Embodiment> Here, a culture condition optimization method according to the second embodiment, which does not include a group generation step, will be described. Fig. 10 is a flowchart showing the optimization processing procedure according to the second embodiment, in which the same components as those in Fig. 2 are assigned the same reference numerals. The flowchart shown in Fig. 10 differs from the first embodiment described above in that it does not include the group generation step of step S11 shown in Fig. 2. However, the processing procedures up to step S10, step S12, and step S15 and subsequent steps shown in Fig. 3 are the same as those of the first embodiment described above, and therefore the following description will focus mainly on the differences.

[0101] In the optimization method according to the second embodiment, in step S9, the list generation unit 104 creates an optimization component list in which the optimization components selected in the selection step are sorted in order of priority. In step S10, the user is prompted to set the number of optimization variables for formulating the culture conditions in the formulation step executed by the optimization module 13 based on the optimization component list.

[0102] Next, in step S12, the optimization system 1 sets the concentration range (upper and lower concentration limits) as concentration conditions for each optimization component that was set as an optimization variable in step S10 in the optimization component list created in step S9 (concentration condition setting step). In this example, the concentration conditions are set based on the user's operation of the operation unit 7 (such as inputting or selecting a numerical value).

[0103] Next, from step S15 onwards shown in Figure 3, the procedure proceeds in the same manner as in the first embodiment described above, and in step S19, the setting of the number of parallel experiments in step S15, the execution of the experimental plan generation process in step S16, the culture test by the user in step S1, the measurement of the antibody production amount in step S2, and the quantification of the medium components in step S3 are repeated until the difference between the culture result obtained in the culture result measurement step based on the newly executed culture step and the judgment criterion exceeds a predetermined specified range.

[0104] <Actions and Effects> According to the above configuration, the optimization method for culture conditions according to the second embodiment also includes a culture step in which an executing entity in an execution environment cultures cells using a culture medium under predetermined culture conditions, a culture result measurement step in which the culture results obtained when the cells are cultured are measured by the culture result measurement device 3, and a quantification step in which the medium components in the medium used for cell culture are quantified by the analysis device 2. The optimization method for culture conditions according to the second embodiment also includes a culture result acquisition step in which the calculation processing device 5 acquires the culture results, and a quantification result acquisition step in which the calculation processing device 5 acquires the quantification results.

[0105] Furthermore, the computing device 5 selects, from among the multiple medium components identified based on the acquired quantification results, medium components that are correlated with the culture results acquired in the culture result acquisition step as important components using the feature extraction module 10 (selection step).The computing device 5 presents, as selection candidate components, medium components other than the important components selected in the selection step that should be noted when culturing cells (presentation step).

[0106] As a result, the arithmetic processing device 5 allows the user to select important components and selection candidate components to be targeted for optimization from among the important components and selection candidate components as optimization components (selection step).

[0107] The arithmetic processing device 5 sets the concentration conditions of the optimization components (concentration condition setting step), and formulates culture conditions including the optimization components that satisfy the concentration conditions set in the concentration condition setting step, based on the optimization module 13 (formulation step). In this way, the culture condition optimization method can optimize the culture conditions when culturing cells using a medium.

[0108] (3) Third Embodiment <Optimization Processing Procedure According to Third Embodiment> Next, a method for optimizing culture conditions according to a third embodiment will be described, which does not have a presentation step of presenting medium components other than important components and which should be noted when culturing cells as candidate components for selection, but has a group generation step of generating grouping components by grouping some of the multiple optimization components.

[0109] Fig. 11 is a flowchart showing an optimization processing procedure according to the third embodiment, in which the same components as those in Fig. 2 are assigned the same reference numerals. The flowchart shown in Fig. 11 differs from the first embodiment described above in that it does not include the presentation step of step S5 and the selection steps of steps S7 and S8 shown in Fig. 2. The following description will mainly focus on the differences from the first embodiment described above.

[0110] In the optimization method according to the third embodiment, in step S4, the important component selection unit 102 of the feature extraction module 10 identifies medium components contained in the culture supernatant based on the mass spectrometry data obtained in the quantification step, and selects, from the identified plurality of medium components, medium components that correlate with the culture results obtained in the culture result measurement step as important components (selection step), and then the method proceeds to step S6.

[0111] Next, in step S6, the optimization system 1 prompts the user to select an optimization component from the selected important components (selection step), and then proceeds to step S9. Then, in the next step S9, the feature extraction module 10 uses the list generation unit 104 to create an optimization component list in which the optimization components selected in the selection step are sorted in order of priority.

[0112] Next, in step S10, the optimization system 1 prompts the user to set the number of optimization components (the number of optimization variables) for which culture conditions are to be formulated in the formulation step executed by the optimization module 13. In step S11, the feature extraction module 10 causes the group setting unit 105 to generate grouping components by grouping multiple optimization components in the optimization component list using one variable in accordance with predetermined grouping rules (group generation step). As a result, the group setting unit 105 creates an optimization component list consisting of multiple optimization components with higher priorities and the grouping components.

[0113] Next, in step S12, the optimization system 1 sets the concentration ranges (upper and lower concentration limits) as concentration conditions for each optimization component and grouping component that were set as optimization variables in step S10 in the optimization component list created in step S11 (concentration condition setting step). In this example, the concentration conditions are set based on the user's operation of the operation unit 7 (such as inputting or selecting a numerical value).

[0114] Next, from step S15 onwards shown in Figure 3, the procedure proceeds in the same manner as in the first embodiment described above, and in step S19, the setting of the number of parallel experiments in step S15, the execution of the experimental plan generation process in step S16, the culture test by the user in step S1, the measurement of the antibody production amount in step S2, and the quantification of the medium components in step S3 are repeated until the difference between the culture result obtained in the culture result measurement step based on the newly executed culture step and the judgment criterion exceeds a predetermined specified range.

[0115] <Actions and Effects> According to the above configuration, the optimization method for culture conditions according to the third embodiment also includes a culture step in which an executing entity in an execution environment cultures cells using a culture medium under predetermined culture conditions, a culture result measurement step in which the execution entity measures the culture results obtained when the cells are cultured using the culture result measurement device 3, and a quantification step in which the medium components in the medium used for cell culture are quantified using the analysis device 2. The optimization method for culture conditions according to the third embodiment also includes a culture result acquisition step in which the calculation processing device 5 acquires the culture results, and a quantification result acquisition step in which the calculation processing device 5 acquires the quantification results.

[0116] Furthermore, the arithmetic processing device 5 selects, from among the plurality of medium components identified based on the acquired quantification results, medium components that are correlated with the culture results acquired in the culture result acquisition step as important components using the feature extraction module 10 (selection step). As a result, the arithmetic processing device 5 allows the user to select, from among the important components, important components to be targeted for optimization as optimization components (selection step).

[0117] The arithmetic processing device 5 sets grouping components by grouping some of the optimization components in accordance with predetermined grouping rules (group setting step), sets concentration conditions for the optimization component and the grouping components (concentration condition setting step), and formulates culture conditions including the optimization component and the grouping components that satisfy the concentration conditions set in the concentration condition setting step, based on the optimization module 13 (formulation step). In this way, the culture condition optimization method according to the third embodiment can also optimize the culture conditions when culturing cells using a medium.

[0118] The present invention allows for various embodiments and modifications without departing from the broad spirit and scope of the present invention. Furthermore, the above-described embodiments are intended to illustrate the present invention and are not intended to limit the scope of the present invention. For example, in the above-described embodiments, the culture result measured in the culture result measurement step is the production amount of antibodies produced by cells. However, the present invention is not limited to this. The culture result may be measured as the number of (viable) cells per unit volume cultured in a medium, the growth rate of cells over a certain culture period, or their shape. In this case, in the above-described embodiments, "increased antibody production amount" is replaced with "increased density of (viable) cells" and "optimized cell shape," respectively.

[0119] As described above, in addition to embodiments for increasing the amount of cells, antibodies, or other produced substances, the present invention can also be applied to embodiments for increasing the uniformity and physiological activity of high-quality cells and produced substances, or for reducing the amount of impurities or growth inhibitors produced. For example, the differentiation efficiency from pluripotent stem cells such as ES cells to differentiated cells may be used as the culture result, and culture conditions may be designed to optimize the differentiation efficiency based on the culture results. Furthermore, with regard to the sugar chain structure conferred on antibodies, the proportion of antibodies having a desired sugar chain structure may be used as the culture result, and culture conditions may be designed to improve the proportion of such antibodies based on the culture results. Furthermore, the amount of substances, such as impurities or inhibitors, that adversely affect cell growth or antibody production may be used as the culture result, and culture conditions may be designed to reduce the amount of such substances based on the culture results.

[0120] That is, in the above case, the culture results obtained in the culture result obtaining step will be any one or more of the production amount of the antibody produced by the cells, the cell density, or the cell morphology, cell quality, uniformity of the produced substance, physiological activity, differentiation efficiency from pluripotent stem cells to differentiated cells, the proportion of antibodies having a desired sugar chain structure conferred to the antibody, and the amount of a substance that adversely affects cell growth or antibody production. Also, in this case, in the selection step, medium components that correlate with any one or more of the production amount of the antibody produced by the cells, cell density, cell morphology, cell quality, uniformity of the produced substance, physiological activity, differentiation efficiency from pluripotent stem cells to differentiated cells, the proportion of antibodies having a desired sugar chain structure conferred to the antibody, and the amount of a substance that adversely affects cell growth or antibody production are selected as important components.

[0121] REFERENCE SIGNS LIST 1 Culture condition optimization system 2 Analysis device 3 Culture result measurement device 5 Processing device 10 Feature extraction module 13 Optimization module

Claims

1. A method for optimizing culture conditions for culturing cells using a culture medium, the method comprising: a culture result acquisition step for acquiring culture results obtained when culturing cells using the culture medium under predetermined culture conditions; a selection step in a computing device for selecting, as important components, from a plurality of culture medium components using a feature extraction module, those culture medium components that are correlated with the culture results; a presentation step in the computing device for presenting, as selection candidate components, those culture medium components other than the important components selected in the selection step that should be noted when culturing the cells; a selection step in the computing device for selecting, as optimization components, the important components and the selection candidate components to be targeted for optimization from the important components and the selection candidate components; and a formulation step in the computing device for formulating the culture conditions including the optimization components using an optimization module.

2. The method for optimizing culture conditions according to claim 1, wherein the culture conditions for the initial culture performed to obtain the culture result are randomly set by the arithmetic processing device or a user.

3. The method for optimizing culture conditions according to claim 1, wherein in the selection step, the optimization component is selected from the important components and the selection candidate components based on domain knowledge.

4. The method for optimizing culture conditions according to claim 1, wherein the formulation step formulates a plurality of culture conditions using the optimization module, and presents the plurality of culture conditions formulated in the formulation step.

5. The method for optimizing culture conditions according to claim 1, wherein in the selection step, medium components that correlate with one or more of the production amount of antibodies produced by the cells, the density of the cells, the shape of the cells, the quality of the cells, the uniformity of the produced substances, physiological activity, the differentiation efficiency from pluripotent stem cells to differentiated cells, the proportion of antibodies with a desired glycan structure, and the amount of substances that adversely affect the growth of the cells or the production amount of antibodies are selected as the important components.

6. A method for optimizing culture conditions as described in claim 1, further comprising, before the formulation step, a concentration condition setting step in the arithmetic processing device for setting concentration conditions for the optimization component, wherein the formulation step involves the optimization module formulating the culture conditions that satisfy the concentration conditions for the optimization component set in the concentration condition setting step.

7. A method for optimizing culture conditions for culturing cells using a culture medium, the method comprising: a culture result acquisition step for acquiring culture results obtained when cells are cultured using the culture medium under predetermined culture conditions; a selection step in a computing device for selecting, from a plurality of culture medium components, medium components that are correlated with the culture results as important components using a feature extraction module; a selection step in the computing device for selecting, from the plurality of important components, the important components to be optimized as optimization components; a group setting step in the computing device for setting grouping components by grouping some of the optimization components in accordance with predetermined grouping rules; and a formulation step in the computing device for formulating the culture conditions including the optimization components and the grouping components based on the optimization module.

8. The method for optimizing culture conditions according to claim 7, wherein the grouping is defined as grouping components that have the same at least one of the following criteria: the attributes of the culture medium components, the concentration values ​​of the culture medium components quantified by an analyzer, the tendency of changes over time in the quantification results of the culture medium components quantified by the analyzer when culturing the cells, the degree of correlation with the culture results obtained in the culture result acquisition step, and metabolic pathways.

9. A method for optimizing culture conditions for culturing cells using a culture medium, the method comprising: a culture result acquisition step for acquiring culture results obtained when culturing the cells using the culture medium under the specified culture conditions; a selection step in a computing device for selecting, as important components, from a plurality of culture medium components using a feature extraction module, those culture medium components that are correlated with the culture result; a presentation step in the computing device for presenting, as selection candidate components, those culture medium components other than the important components selected in the selection step that should be noted when culturing the cells; a selection step in the computing device for selecting, as optimization components, the important components and the selection candidate components to be targeted for optimization from the important components and the selection candidate components; a group generation step in the computing device for generating grouping components by grouping some of the plurality of optimization components selected in the selection step in accordance with predetermined grouping rules; and a formulation step in the computing device for formulating the culture conditions including the optimization components and the grouping components based on an optimization module.

10. A method for optimizing culture conditions according to any one of claims 1 to 9, wherein the culture results obtained in the culture result obtaining step are one or more of the production amount of antibodies produced by the cells, the density of the cells, the shape of the cells, the quality of the cells, the uniformity of the produced substances, physiological activity, the differentiation efficiency from pluripotent stem cells to differentiated cells, the proportion of antibodies having a desired glycan structure, or the amount of a substance that has an adverse effect on the growth of the cells or the production amount of antibodies.

11. A method for optimizing culture conditions according to any one of claims 1 to 9, wherein the steps from the culture result acquisition step to the formulation step are repeatedly performed until a predetermined culture result is obtained, thereby optimizing the culture conditions for culturing the cells.

12. The method for optimizing culture conditions described in claim 11, further comprising a judgment step in the arithmetic processing device, in which a predetermined culture result is used as a judgment criterion, and it is judged whether the difference between the acquired culture result and the judgment criterion exceeds a predetermined specified range, and if it is judged by the judgment step that the difference between the acquired culture result and the judgment criterion does not exceed the specified range, it is determined that the predetermined culture result has been obtained, and the repetition from the culture result acquisition step to the formulation step is terminated.

13. A computing device comprising: a memory for storing a program; and a processor for executing the program, wherein, by executing the program, the processor performs: a culture result acquisition step for acquiring culture results obtained when cells are cultured using a culture medium under specified culture conditions; a selection step for selecting, from a plurality of culture medium components, medium components that are correlated with the culture result as important components; a presentation step for presenting, as selection candidate components, medium components other than the important components selected in the selection step that should be noted when culturing the cells; a selection step for selecting, from the important components and the selection candidate components, the important components and the selection candidate components to be targeted for optimization as optimization components; and a formulation step for formulating culture conditions including the optimization components.

Images