Method for operating a cell culture condition determination device and cell culture condition determination device

Abstract

The present invention provides an apparatus for determining cell culture conditions using artificial intelligence and a method for operating the apparatus, comprising the steps of acquiring a first quality attribute and a first influence variable at a first time point from a cell culture medium, acquiring a first injection variable, applying at least one of the first quality attribute, the first influence variable, and the first injection variable to a culture simulation model to acquire a second quality attribute and a second influence variable at a second time point, and selectively changing the first injection variable so that the second quality attribute meets a predetermined condition.

Description

Technical Field

[0001] The present invention relates to an apparatus for determining cell culture conditions using artificial intelligence and a method of operating the apparatus. More specifically, the apparatus includes a method of predicting changes in influence variables due to injection variables and searching for optimal injection variables to obtain a target quality attribute.

Background Art

[0002] Bioreactors, which are apparatuses capable of performing biological reactions or processes on a laboratory or industrial scale, are widely used in the biopharmaceutical industry. Bioreactors are used to produce various types of bioproducts. Bioproducts may include, for example, cell culture media and beverages, biofuels, bioenergy, biochemicals, antibiotics, amino acids, enzymes, monoclonal antibodies, multi-specific antibodies, vaccines, blood products, plasma fractionation products, proteins, cell culture pharmaceuticals, cell therapy agents, gene therapy agents, tissue engineering products, advanced bio-fusion composite products, food culture media, biopolymers, alcohols, flavors, fragrances, and the like.

[0003] Note that cell culture of microorganisms, plant cells, animal cells, etc. can be performed by methods such as batch culture, fed-batch culture, continuous culture, and perfusion culture. In a specific process, in order to replenish nutrients contained in a fluid medium or to remove harmful by-products generated during the process, the liquid medium (culture solution) contained in the bioreactor can be periodically or continuously removed and re-supplied.

[0004] During cell culture in a bioreactor, regulating the major metabolites in the culture medium can directly affect the quality of the product produced. For example, lactate is one of the major metabolites that should be controlled during cell culture, especially when mammalian cells grow. As an example, a large amount of lactate is produced during the cell growth phase, while its consumption is observed when the cells enter the stationary phase. High levels of lactate can have a negative impact on cell culture. For example, lactate accumulation can have a negative impact on the quality attributes of the product, and in fact, extreme lactate accumulation may prevent the cell culture from being used commercially.

[0005] For this reason, a solution has been proposed to regulate one of the amounts of major metabolites in the culture medium. However, in the actual culture medium, the major metabolites affect each other and vary. Therefore, when optimizing all the major metabolites instead of just regulating a specific type of major metabolite, the culture conditions can be optimized.

[0006] However, optimizing all the metabolites or balancing them is a very difficult task. For example, Glucose, Glutamine, and Glutamate are all parameters that have a negative impact on cells when depleted. However, injecting a large amount of them through the Feed may increase the osmotic pressure and have a negative impact on the cells. Also, Glucose, Glutamine, and Glutamate themselves may have a negative impact when in excess. That is, balancing the metabolites is a very complex task. Summary of the Invention Problems to be Solved by the Invention

[0007] Accordingly, the present invention is related to a virtual bioreactor simulation device implemented by promoting machine learning that maximally reflects the interaction between parameters (influence variables and injection variables) that affect it in order to achieve desired cell quality attributes.

[0008] Further, the present invention is provided to predict and analyze the values of future important parameters from current parameter values (influence variables and injection variables) and derive optimal cell culture conditions and quality attributes.

Means for Solving the Problems

[0009] In this specification, the operation method of the cell culture condition determination device includes the steps of obtaining a first quality attribute and a first influence variable at a first time point from a cell culture solution, obtaining a first injection variable, applying at least one of the first quality attribute, the first influence variable, and the first injection variable to a culture simulation model to obtain a second quality attribute and a second influence variable at a second time point, and selectively changing the first injection variable so that the second quality attribute meets a predetermined condition, wherein the second time point is in the future of the first time point.

[0010] The first quality attribute may include a first VCD value and a first IgG amount.

[0011] The first influence variable may include at least one of the amount of first glutamine, the amount of first glutamate, the amount of first glucose, the amount of first lactate, the amount of first ammonia, and the first osmolarity.

[0012] The step of obtaining the first quality attribute and the first influencing variables at the first time point from the cell culture medium includes the step of obtaining a plurality of first influencing variables measured at the same time point, and the step of obtaining the second quality attribute and the second influencing variables at the second time point may include the step of obtaining the second influencing variables predicted based on the plurality of measured first influencing variables.

[0013] The first injection variable may include at least one of the amount of the first nutrient feed, the amount of glucose in the first nutrient medium, the amount of the first oxygen molecules, the amount of the first carbon dioxide molecules, the amount of the first nitrogen molecules, the amount of the first air, the first pH value, the first Agitator Speed, the first Vessel temperature, the first Media injection schedule, and the amount of the first Antifoam.

[0014] The step of selectively changing the first injection variable includes, in the first period, changing the first injection variable such that the second VCD value included in the second quality attribute has a maximum value, and in the second period, changing the first injection variable such that the amount of the second IgG included in the second quality attribute has a maximum value, and the second period may be after the first period.

[0015] The step of selectively changing the first injection variable includes, in the first period, changing the first injection variable such that the value obtained by multiplying the second VCD value included in the second quality attribute by e^(-k * the amount of the second ammonia) has a maximum value, and in the second period, changing the first injection variable such that the value obtained by multiplying the amount of the second IgG included in the second quality attribute by e^(-k * the amount of the second ammonia) has a maximum value, and k may be a positive number.

[0016] The step of selectively changing the first injection variable may include determining whether the amount of the second Lactate included in the second influencing variable is less than or equal to a first critical amount, and when the amount of the second Lactate is less than or equal to the first critical amount and within the first period, changing the first injection variable so that the second VCD value included in the second quality attribute has a maximum value, and when the amount of the second Lactate is less than or equal to the first critical amount and within the second period, changing the first injection variable so that the amount of the second IgG included in the second quality attribute has a maximum value.

[0017] Also, in this specification, the step of selectively changing the first injection variable may include determining whether the amount of the second Glutamate included in the second influencing variable is greater than or equal to a second critical amount and less than or equal to a third critical amount, and when the amount of the second Glutamate is greater than or equal to the second critical amount and less than or equal to the third critical amount and within the first period, changing the first injection variable so that the second VCD value included in the second quality attribute has a maximum value, and when the amount of the second Glutamate is greater than or equal to the second critical amount and less than or equal to the third critical amount and within the second period, changing the first injection variable so that the amount of the second IgG included in the second quality attribute has a maximum value.

[0018] Also, this specification may include the step of transmitting the changed first injection variable to the bioreactor.

[0019] In this specification, the cell culture condition determination device includes a processor and a memory. The processor acquires the first quality attribute and the first influencing variable at a first time point from the cell culture solution, acquires the first injection variable, applies at least one of the first quality attribute, the first influencing variable, and the first injection variable to a culture simulation model, acquires the second quality attribute and the second influencing variable at a second time point, and selectively changes the first injection variable so that the second quality attribute meets predetermined conditions. The second time point may be in the future of the first time point. BRIEF DESCRIPTION OF THE DRAWINGS

[0020]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Embodiments for Carrying Out the Invention

[0021] The advantages and features of the disclosed embodiments, and the methods for achieving them, will become clear by referring to the embodiments described below together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, and can be embodied in various different forms, and these embodiments are merely provided to complete the present invention and to fully inform those with ordinary knowledge in the technical field to which the present invention pertains of the scope of the invention.

[0022] The terms used in this specification will be briefly explained, and the disclosed embodiments will be specifically described.

[0023] The terms used in this specification are selected as general terms that are currently widely used as much as possible in consideration of the functions in the present invention, but this can change depending on the intentions or precedents of those skilled in the relevant fields, the emergence of new technologies, etc. Also, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the part of the description of the corresponding invention. Therefore, the terms used in the present invention should be defined based on the meaning of the terms and the overall content of the present invention.

[0024] As used herein, the singular forms also include the plural forms unless specifically stated otherwise in the context. Also, the plural forms include the singular form unless specifically stated otherwise in the context.

[0025] When any part of the specification states that a component "comprises", this means that it includes, without exclusion of other components, and may further include other components, unless specifically stated to the contrary.

[0026] Also, the term "module" as used in the specification means a software or hardware component, and a "module" performs a predetermined role. However, a "module" is not limited to meaning software or hardware. A "module" can also be configured to be located in a storage medium that can be addressed, or can be configured to cause one or more processors to execute. Thus, by way of example, a "module" includes components such as software components, object-oriented software components, class components and task components, and processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays and variables. The functions provided within a component and a "module" may be combined with a smaller number of components and "modules", or may be further separated into additional components and "modules".

[0027] According to an embodiment of the present invention, the "unit" may be implemented by a processor and a memory. The term "processor" should be broadly interpreted to include general-purpose processors, central processing units (CPUs), microprocessors, digital signal processors (DSPs), controllers, microcontrollers, state machines, and the like. In some environments, the "processor" may also refer to application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), and the like. The term "processor" may also refer to a combination of processing devices, such as a combination of a DSP and a microprocessor, a combination of multiple microprocessors, a combination of one or more microprocessors coupled with a DSP core, or any other such configuration combination.

[0028] The term "memory" should be broadly interpreted to include any electronic component capable of storing electronic information. The term "memory" may also refer to various types of processor-readable media, such as random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable PROM (EEPROM), flash memory, magnetic or optical data storage devices, registers, and the like. When a processor can read / retrieve information from the memory or record information in the memory, the memory is said to be in electronic communication with the processor. Memory integrated into a processor is in electronic communication with the processor.

[0029] Hereinafter, embodiments will be described in detail with reference to the accompanying drawings so that those with ordinary knowledge in the technical field to which the present invention pertains can easily implement them. Also, parts not related to the description are omitted in the drawings to clearly explain the present invention.

[0030] FIG. 1 is a block diagram of a cell culture condition determination device 100 according to an embodiment of the present invention. Referring to FIG. 1, a cell culture condition determination device 100 according to an embodiment may include at least one of a data learning unit 110 or a data recognition unit 120. The cell culture condition determination device 100 as described above may include a processor and a memory. In the present invention, the culture condition determination device 100 can be simply described as the device 100. Also, the device 100 may include a PC, a server, a tablet, a workstation, a smartphone, and the like. The device 100 may include an input unit and an output unit. The input unit is a device for receiving an input from a user, and the output unit may be a device for outputting information in the form of audio, video, or data. The device 100 may include a communication unit for transmitting and receiving information from an external device. The device 100 can receive information from a user terminal and transmit an analysis result to the user terminal.

[0031] The data learning unit 110 can learn a machine learning model for performing a target task using a data set. The data learning unit 110 can receive a data set and label information related to the target task. The data learning unit 110 can perform machine learning on the relationship between the data set and the label information to obtain a machine learning model. The machine learning model obtained by the data learning unit 110 can be a model for predicting label information using the data set.

[0032] The data recognition unit 120 may receive and store the machine learning model of the data learning unit 110. The data recognition unit 120 can output label information predicted by applying the machine learning model to the input data. Also, the data recognition unit 120 can use the input data, the label information, and the result output by the machine learning model to update the machine learning model.

[0033] At least one of the data learning unit 110 and the data recognition unit 120 may be fabricated in the form of at least one hardware chip and mounted on an electronic device. For example, at least one of the data learning unit 110 and the data recognition unit 120 can be fabricated in the form of a dedicated hardware chip for artificial intelligence (AI), or can be fabricated as part of an existing general-purpose processor (e.g., CPU or application processor) or a graphics dedicated processor (e.g., GPU), and can also be mounted on various electronic devices already described.

[0034] In addition, the data learning unit 110 and the data recognition unit 120 can also be respectively mounted on separate electronic devices. For example, one of the data learning unit 110 and the data recognition unit 120 may be included in an electronic device, and the remaining one may be included in a server. Also, the data learning unit 110 and the data recognition unit 120 can provide the machine learning model information constructed by the data learning unit 110 to the data recognition unit 120 via wired or wireless means, and the data input to the data recognition unit 120 can also be provided to the data learning unit 110 as additional learning data.

[0035] Note that at least one of the data learning unit 110 and the data recognition unit 120 may be embodied as a software module. When at least one of the data learning unit 110 and the data recognition unit 120 is embodied as a software module (or a program module including instructions), the software module may be stored in a memory or a non-transitory computer readable medium readable by a computer. Also, in this case, at least one software module may be provided by an OS (Operating System) or by a predetermined application. Or, a part of at least one software module may be provided by an OS (Operating System), and the remaining part may be provided by a predetermined application.

[0036] The data learning unit 110 according to an embodiment of the present invention may include a data acquisition unit 111, a preprocessing unit 112, a learning data selection unit 113, a model learning unit 114, and a model evaluation unit 115.

[0037] The data acquisition unit 111 can acquire data necessary for machine learning. Since a large amount of data is required for learning, the data acquisition unit 111 can receive a data set including a plurality of data.

[0038] Label information may be assigned to each of the plurality of data. The label information may be information that describes each of the plurality of data. The label information may be information that the target task is trying to derive. The label information can be obtained from user input, obtained from memory, or obtained from the results of a machine learning model. The label information may be ground truth information. For example, if the target task is to predict the quality attribute or influence variable at the second time point based on the quality attribute, influence variable, and injection variable at the first time point, the plurality of data used in machine learning will be the quality attribute, influence variable, and injection variable at the first time point, and the label information will be the quality attribute or influence variable at the second time point. The label information may be ground truth information, and may be directly input by the user or stored in the apparatus 100 in advance.

[0039] The preprocessing unit 112 can preprocess the acquired data so that the received data can be used for machine learning. The preprocessing unit 112 can process the acquired data set into a preset format so that the model learning unit 114 described later can use it.

[0040] The learning data selection unit 113 can select the data necessary for learning from the preprocessed data. The selected data can be provided to the model learning unit 114. The learning data selection unit 113 can select the data necessary for learning from the preprocessed data according to a preset criterion. Also, the learning data selection unit 113 can select data according to a preset criterion by learning by the model learning unit 114 described later.

[0041] The model learning unit 114 can learn the criteria for what label information to output based on the dataset. Also, the model learning unit 114 can perform machine learning using the dataset and the label information regarding the dataset as learning data. Further, the model learning unit 114 can perform machine learning using an already acquired machine learning model. In this case, the already acquired machine learning model can be a model constructed in advance. For example, the machine learning model can be a model constructed in advance by inputting basic learning data.

[0042] The machine learning model can be constructed in consideration of the application field of the learning model, the purpose of learning, or the computer performance of the device, etc. The machine learning model can be, for example, a model based on a neural network. For example, models such as Deep Neural Network (DNN), Recurrent Neural Network (RNN), Long Short-Term Memory models (LSTM), BRDNN (Bidirectional Recurrent Deep Neural Network), and Convolutional Neural Networks (CNN) can be used as the machine learning model, but are not limited thereto.

[0043] According to various embodiments, when there are a plurality of pre-constructed machine learning models, the model learning unit 114 can be determined as the machine learning model that learns the machine learning model having a high relevance between the input learning data and the basic learning data. In this case, the basic learning data may be pre-classified for each data type, and the machine learning model may be pre-constructed for each data type. For example, the basic learning data may be pre-classified according to various criteria such as the place where the learning data was generated, the time when the learning data was generated, the size of the learning data, the generator of the learning data, the type of object in the learning data, etc.

[0044] In addition, the model learning unit 114 can train a machine learning model using, for example, a learning algorithm including error back-propagation or gradient descent.

[0045] In addition, the model learning unit 114 can train a machine learning model using, for example, supervised learning with learning data as input values. Also, the model learning unit 114 can obtain a machine learning model using, for example, unsupervised learning that discovers criteria for a target task by learning on its own the types of data required for the target task without a separate teacher. Further, the model learning unit 114 can train a machine learning model using, for example, reinforcement learning that uses feedback on whether the result of a target task by learning is correct.

[0046] In addition, when the machine learning model is trained, the model learning unit 114 can save the trained machine learning model. In this case, the model learning unit 114 can save the trained machine learning model in the memory of the electronic device including the data recognition unit 120. Alternatively, the model learning unit 114 can also save the trained machine learning model in the memory of a server connected to the electronic device via a wired or wireless network.

[0047] The memory in which the trained machine learning model is saved can also save, for example, instructions or data related to at least one other component of the electronic device. Also, the memory can save software and / or programs. The program may include, for example, a kernel, middleware, an application programming interface (API), and / or an application program (or "application").

[0048] The model evaluation unit 115 can input evaluation data into the machine learning model and, if the result output from the evaluation data does not meet a predetermined standard, cause the model learning unit 114 to learn again. In this case, the evaluation data can be default data for evaluating the machine learning model.

[0049] For example, if the number or ratio of evaluation data with inaccurate recognition results among the results of the learned machine learning model for the evaluation data exceeds a preset threshold, the model evaluation unit 115 can evaluate that it does not meet the predetermined standard. For example, if the predetermined standard is defined as a ratio of 2%, and the learned machine learning model outputs incorrect recognition results for more than 20 out of a total of 1000 evaluation data, the model evaluation unit 115 can evaluate that the learned machine learning model is not suitable.

[0050] Note that when there are multiple learned machine learning models, the model evaluation unit 115 can evaluate whether each learned machine learning model meets a predetermined standard and determine the model that meets the predetermined standard as the final machine learning model. In this case, when there are multiple models that meet the predetermined standard, the model evaluation unit 115 can determine any one or a predetermined number of models set in descending order of evaluation scores as the final machine learning model.

[0051] Note that at least one of the data acquisition unit 111, preprocessing unit 112, learning data selection unit 113, model learning unit 114, and model evaluation unit 115 within the data learning unit 110 may be manufactured in the form of at least one hardware chip and mounted on an electronic device. For example, at least one of the data acquisition unit 111, preprocessing unit 112, learning data selection unit 113, model learning unit 114, and model evaluation unit 115 can be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), or can be manufactured as part of an existing general-purpose processor (e.g., CPU or application processor) or a graphics dedicated processor (e.g., GPU), and can also be mounted on the various electronic devices described above.

[0052] Also, the data acquisition unit 111, preprocessing unit 112, learning data selection unit 113, model learning unit 114, and model evaluation unit 115 can be mounted on one electronic device, or can be respectively mounted on separate electronic devices. For example, some of the data acquisition unit 111, preprocessing unit 112, learning data selection unit 113, model learning unit 114, and model evaluation unit 115 may be included in an electronic device, and the remaining part may be included in a server.

[0053] Further, at least one of the data acquisition unit 111, the preprocessing unit 112, the learning data selection unit 113, the model learning unit 114, and the model evaluation unit 115 may be embodied as a software module. When at least one of the data acquisition unit 111, the preprocessing unit 112, the learning data selection unit 113, the model learning unit 114, and the model evaluation unit 115 is embodied as a software module (or a program module including instructions), the software module may be stored in a non-transitory computer readable medium readable by a computer. Also, in this case, at least one software module may be provided by an OS (Operating System) or by a predetermined application. Or a part of at least one software module may be provided by an OS (Operating System), and the remaining part may be provided by a predetermined application.

[0054] The data recognition unit 120 according to an embodiment of the present invention may include a data acquisition unit 121, a preprocessing unit 122, a recognition data selection unit 123, a recognition result providing unit 124, and a model update unit 125.

[0055] The data acquisition unit 121 can receive input data. The preprocessing unit 122 can preprocess the acquired input data so that the acquired input data can be used by the recognition data selection unit 123 or the recognition result providing unit 124.

[0056] The recognition data selection unit 123 can select necessary data from the preprocessed data. The selected data may be provided to the recognition result providing unit 124. The recognition data selection unit 123 can select some or all of the preprocessed data according to a predetermined criterion. Also, the recognition data selection unit 123 can select data according to a predetermined criterion by learning by the model learning unit 114.

[0057] The recognition result providing unit 124 can apply the selected data to the machine learning model and obtain result data. The machine learning model can be the machine learning model generated by the model learning unit 114. The recognition result providing unit 124 can output the result data. For example, the recognition result providing unit 124 can receive the user's current biometric information and output the user's psychological information as the result data.

[0058] The model updating unit 125 can update the machine learning model based on the evaluation of the recognition result provided by the recognition result providing unit 124. For example, the model updating unit 125 can cause the model learning unit 114 to update the machine learning model by providing the recognition result provided by the recognition result providing unit 124 to the model learning unit 114.

[0059] Note that at least one of the data acquisition unit 121, the preprocessing unit 122, the recognition data selection unit 123, the recognition result providing unit 124, and the model updating unit 125 in the data recognition unit 120 may be manufactured in the form of at least one hardware chip and mounted on the electronic device. For example, at least one of the data acquisition unit 121, the preprocessing unit 122, the recognition data selection unit 123, the recognition result providing unit 124, and the model updating unit 125 can also be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), or can be manufactured as part of an existing general-purpose processor (e.g., CPU or application processor) or a graphics dedicated processor (e.g., GPU) and mounted on the various electronic devices described above.

[0060] Further, the data acquisition unit 121, the preprocessing unit 122, the recognition data selection unit 123, the recognition result providing unit 124, and the model update unit 125 can be mounted on one electronic device, or can be respectively mounted on separate electronic devices. For example, some of the data acquisition unit 121, the preprocessing unit 122, the recognition data selection unit 123, the recognition result providing unit 124, and the model update unit 125 may be included in an electronic device, and the remaining part may be included in a server.

[0061] Further, at least one of the data acquisition unit 121, the preprocessing unit 122, the recognition data selection unit 123, the recognition result providing unit 124, and the model update unit 125 may be embodied as a software module. When at least one of the data acquisition unit 121, the preprocessing unit 122, the recognition data selection unit 123, the recognition result providing unit 124, and the model update unit 125 is embodied as a software module (or a program module including instructions), the software module may be stored in a non-transitory computer readable medium readable by a computer. Also, in this case, at least one software module may be provided by an OS (Operating System) or by a predetermined application. Or a part of at least one software module may be provided by an OS (Operating System), and the remaining part may be provided by a predetermined application.

[0062] Hereinafter, a method and an apparatus for the data acquisition unit 111, the preprocessing unit 112, and the learning data selection unit 113 of the data learning unit 110 to receive and process learning data will be described in more detail.

[0063] FIG. 2 is a diagram showing a cell culture condition determination apparatus according to an embodiment of the present invention. The apparatus 100 may include a processor 210 and a memory 220. The processor 210 can execute the instruction words stored in the memory 220.

[0064] As described above, the apparatus 100 may include the data learning unit 110 or the data recognition unit 120. The data learning unit 110 or the data recognition unit 120 may be implemented by the processor 210 and the memory 220.

[0065] To describe the apparatus 100 in more detail, refer to FIG. 18 for a while.

[0066] FIG. 18 is a block diagram for explaining a bioreactor system according to an embodiment of the present invention. Referring to FIG. 18, an embodiment of a bioreactor system according to the present invention is shown. As shown in the figure, the cell culture solution is cultured in the bioreactor 1810 during the culture period and then harvested. During the culture, various parameters in the bioreactor 1810 are monitored. The parameters are measured by the measuring device 1820. According to the present invention, the measuring device 1820 periodically or continuously monitors the influencing variables and quality attributes transmitted to the apparatus 100.

[0067] As shown in FIG. 18, the apparatus 100 can receive predetermined conditions. The apparatus 100 can select the optimal injection variable among a plurality of injection variables so as to satisfy the predetermined conditions. The apparatus 100 may include one or more prediction models 730. The prediction model 730 may be generated based on the data learning unit 110 or the data recognition unit 120. The apparatus 100 may include at least one of a classification model or a prediction model.

[0068] A classification model can be a model for classifying an impact variable or a quality attribute into one of a plurality of classes. The class can be a range of the impact variable or the quality attribute. The classification model can predict the probability that the impact variable or the quality attribute belongs to a specific class. The classification model can use various multivariate analysis methods, including partial least squares analysis alone or together with linear discriminant analysis. Also, the classification model can use a classification tree, a support vector machine, etc. In one embodiment, the median of the percent probabilities generated from each classification model can be used as the final percent probability for the cell culture.

[0069] The apparatus 100 may include a prediction model. The prediction model can predict how the impact variable and the quality attribute will change in the future based on the injection variables. For example, as shown in FIG. 18, the apparatus 100 can acquire the injection variables. The apparatus 100 can predict the future impact variable and the future quality attribute based on the injection variables. The apparatus 100 can use the prediction model to predict the future impact variable and the future quality attribute. The future impact variable and the future quality attribute are predicted differently depending on the value of the injection variable. The apparatus 100 can repeatedly perform a number of simulations to determine whether a calibration measure is necessary in the cell culture fluid. That is, the apparatus 100 can repeatedly perform simulations for a plurality of injection variables that are different from each other. Also, the apparatus 100 can acquire the optimal injection variables for maximizing the production of the target substance in the bioreactor 1810, and can predict the future quality attribute and the future impact variable based on the optimal injection variables.

[0070] When the future actual impact variable or the future actual quality attribute is provided to the apparatus 100, the apparatus 100 can continue to perform simulations over the entire culture period in order to further change or modify the injection variables and change one or more conditions in the cell culture fluid.

[0071] FIG. 3 is a flowchart for explaining an operation method of a cell culture condition determination apparatus according to an embodiment of the present invention. Also, FIG. 19 is a flowchart for generally explaining an operation method of a cell culture condition determination apparatus according to an embodiment of the present invention.

[0072] First, the operation of the apparatus 100 of the present invention will be generally explained with reference to FIG. 19. The apparatus 100 can perform a step 1910 of obtaining a first quality attribute and a first influence variable at a first time point from a cell culture solution in order to determine a culture condition. Such a step 1910 can correspond to step 310 in FIG. 3.

[0073] The apparatus 100 can perform a step 1920 of obtaining a first injection variable. Such a step 1920 can correspond to step 320 in FIG. 3.

[0074] The apparatus 100 can perform a step 1930 of applying at least one of the first quality attribute, the first influence variable, and the first injection variable to a culture simulation model to obtain a second quality attribute and a second influence variable at a second time point. Such a step 1930 can correspond to step 330 in FIG. 3.

[0075] The second time point can be in the future of the first time point. The apparatus 100 can perform a step 1940 of selectively changing the first injection variable so that the second quality attribute meets a predetermined condition. Such a step 1940 can correspond to steps 340 and 350 in FIG. 3. That is, steps 340 and 350 can be performed for the apparatus 100 to perform the step 1940 of selectively changing the first injection variable.

[0076] The apparatus 100 can perform a plurality of tests using a plurality of first injection variables in order to selectively change the first injection variables. That is, the apparatus 100 can predict a second quality attribute and a second influence variable for each of the plurality of first injection variables. Further, the apparatus 100 can select a first injection variable that meets a predetermined condition among the plurality of first injection variables. The predetermined condition may mean that the predicted second quality attribute and the predicted second influence variable are within a specific range. The apparatus 100 can modify the culture conditions of the cell culture solution using the selected first injection variable and culture the cells. For example, the apparatus 100 can perform a step of transmitting the changed first injection variable to the bioreactor 1810. The bioreactor 1810 can change the culture conditions based on the changed first injection variable. Hereinafter, such a process will be described in more detail.

[0077] The apparatus 100 can perform a step 310 of acquiring a first quality attribute and a first influence variable from the cell culture solution.

[0078] The quality attribute is a numerical value related to the ultimate goal that the cell culture solution is to reach via the apparatus 100. For example, the quality attribute may include at least one of the VCD (viable cell density) value of the cell culture solution (medium) or the amount of IgG (Immunoglobulin G). The VCD value and the amount of IgG can be types of quality attributes. Further, the quality attribute may include an Antibody titer or Antibody quality attributes.

[0079] The influencing variables may include at least one of the amount of glutamine, the amount of glutamate, the amount of glucose, the amount of lactate, the amount of ammonia, and the osmolarity in the culture medium. The amount of glutamine, the amount of glutamate, the amount of glucose, the amount of lactate, the amount of ammonia, and the osmolarity may represent the types of influencing variables. In this specification, the amount may be the concentration of the substance in the culture medium. For example, the unit of the amount may be cells / mL, mmol / L, or g / L. However, it is not limited thereto, and in this specification, the amount may mean the absolute amount of the substance. For example, the unit of the amount may be the number of cells, mol, or g. Also, in this specification, the amount may mean the ratio of the substance to the total substance. For example, the unit may be %.

[0080] Also, the first quality attribute and the first influencing variable can indicate the quality attribute and the influencing variable at the first time point. The first time point may be a time point preceding the second time point. The first time point may mean a time point having a measured value. Also, the second time point is a future time point having no measured value, but may be a time point at which a predicted value can be obtained based on the measured value at the first time point. The second time point may be after a predetermined time has elapsed from the first time point. The predetermined time may be 10 minutes, 1 hour, 1 day, etc. Also, since the first time point is a time point at which a measured value can be measured, it can change as time passes. For example, the first time point may be the present. However, it is not limited thereto, and the first time point may be in the past or the future. The quality attribute and the influencing variable at the first time point may be measured values or predicted values for the future. The second time point can also change when the first time point changes after a predetermined time has elapsed from the first time point.

[0081] The first quality attribute may include at least one of the first VCD value and the amount of the first IgG. Also, the first influencing variable may include at least one of the amount of the first glutamine, the amount of the first glutamate, the amount of the first glucose, the amount of the first lactate, the amount of the first ammonia, and the first osmolarity.

[0082] The device 100 can acquire the first quality attribute and the first influencing variable from the user by using the input unit included in the device 100. Also, the device 100 can receive the first quality attribute and the first influencing variable from an external device. Also, the device 100 can measure the first quality attribute and the first influencing variable of the culture medium by using a sensor. Also, the device 100 can acquire the first quality attribute and the first influencing variable pre-stored in the memory. Also, the device 100 can acquire the first quality attribute and the first influencing variable measured by the measuring instrument 1820.

[0083] The device 100 can perform the step of acquiring two or more types of the first influencing variables measured at the same time point. That is, the device 100 can measure two or more types among the amount of the first glutamine, the amount of the first glutamate, the amount of the first glucose, the amount of the first lactate, the amount of the first ammonia, and the first osmolarity. However, it is not limited thereto, and the device 100 can also acquire one first influencing variable.

[0084] The device 100 can perform step 320 of acquiring a plurality of first injection variables included within a predetermined range. As already explained, the device 100 can perform step 1920 of acquiring the first injection variables, but different first injection variables can be used for testing. That is, the device 100 can acquire a plurality of first injection variables to select an optimal first injection variable. The injection variables can be variables that affect the influencing variables and quality attributes. The injection variables can be values controlled by the user. The user can control the injection variables and change the influencing variables and quality attributes. The culture simulation model of the present invention can predict future influencing variables or quality attributes based on the current injection variables. The injection variables can be numerical values related to substances controlled such that cells are cultured in a medium. The device 100 may receive an input of the injection variables from the user. Also, the device 100 can receive the injection variables from another device. Further, the device 100 can acquire the injection variables from a memory. Additionally, the device 100 can acquire the injection variables based on a predetermined algorithm. For example, the device 100 can acquire the injection variables randomly within a predetermined range.

[0085] The injection variables or sub-injection variables may include at least one of the amount of feed, the amount of glucose in the feed, the amount of oxygen molecules (O2), the amount of carbon dioxide (CO2) molecules, the amount of nitrogen molecules (N2), the amount of air, the pH value, the agitator speed, the vessel temperature, the media injection schedule, and the amount of antifoam. Here, the amount of oxygen molecules may mean the amount of dissolved oxygen in the medium.

[0086] Also, the first injection variable can indicate the injection variable at the first time point. The first time point can be a time point preceding the second time point. The interval between the first time point and the second time point can be 10 minutes, 1 hour, or 1 day, etc. For example, the first injection variable may include at least one of the amount of the first nutrient medium (feed), the amount of glucose in the first nutrient medium, the amount of the first oxygen molecules, the amount of the first carbon dioxide molecules, the amount of the first nitrogen molecules, the amount of the first air, the first pH value, the first Agitator Speed, the first Vessel temperature, the first Media injection schedule, and the amount of the first Antifoam.

[0087] In the present invention, the "plurality of" first injection variables may mean that there are a plurality of sets of first injection variables each including at least one of the amount of the first nutrient medium (feed), the amount of glucose in the first nutrient medium, the amount of the first oxygen molecules, the amount of the first carbon dioxide molecules, the amount of the first nitrogen molecules, the amount of the first air, the first pH value, the first Agitator Speed, the first Vessel temperature, the first Media injection schedule, and the amount of the first Antifoam.

[0088] The first injection variable may be included within a predetermined range. The predetermined range can mean the range of values of the first injection variable. The predetermined range may vary depending on the type of the first injection variable. The amount of the first nutrient medium (feed), the amount of glucose in the first nutrient medium, the amount of the first oxygen molecules, the amount of the first carbon dioxide molecules, the amount of the first nitrogen molecules, the amount of the first air, the first pH value, the first Agitator Speed, the first Vessel temperature, the first Media injection schedule, and the amount of the first Antifoam may each have a predetermined range set. The predetermined range is an experimentally obtained value and can be the range of the injection variable for maximizing the quality attribute. The predetermined range can be changed according to the user's settings.

[0089] The device 100 can use a plurality of first injection variables to find the optimal injection variables at the first time point. When the device 100 performs step 320 of obtaining a plurality of first injection variables included within a predetermined range, it can further perform the following steps. The device 100 can perform a step of selecting a predetermined number of sub-injection variables out of the plurality of sub-injection variables.

[0090] The injection variable may include sub-injection variables. The sub-injection variable may mean at least one type included in the injection variable. Or, the injection variable may mean a group of sub-injection variables. For example, the sub-injection variable may respectively mean the amount of feed in the injection variable, the amount of Glucose in the feed, the amount of oxygen molecules, the amount of carbon dioxide molecules, the amount of nitrogen molecules, the amount of air, the pH value, the Agitator Speed, the Vessel temperature, the Media injection schedule, and the amount of antifoam. The plurality of sub-injection variables may mean the amount of feed, the amount of Glucose in the feed, the amount of oxygen molecules, the amount of carbon dioxide molecules, the amount of nitrogen molecules, the amount of air, the pH value, the Agitator Speed, the Vessel temperature, the Media injection schedule, and the amount of antifoam.

[0091] The device 100 can select a predetermined number out of the sub-injection variables. For example, the predetermined number can be 2. The device 100 can select a predetermined number of sub-injection variables based on the user's input. Or, the device 100 can automatically select a predetermined number of sub-injection variables. For example, when the predetermined number is 2, the selected sub-injection variables can be the pH value and / or the amount of oxygen molecules.

[0092] In addition, the apparatus 100 can perform a step of obtaining candidate values for each selected sub-injection variable. The predetermined range may be different for each sub-injection variable. The apparatus 100 can obtain candidate values automatically or manually for each selected sub-injection variable. The apparatus 100 can select candidate values for the selected sub-injection variables within a predetermined range and obtain candidate values for each selected sub-injection variable. More specifically, the apparatus 100 can divide the predetermined range into a plurality of candidate ranges. The apparatus 100 can divide the predetermined range into n equal parts by the number n of the predetermined ranges. The number of the predetermined ranges may be different for each sub-injection variable. In addition, the apparatus 100 can select representative values for each of the plurality of candidate ranges and obtain candidate values. The representative value of the candidate range can be the median of the candidate range or the average of the minimum and maximum of the candidate range.

[0093] For example, the predetermined range of the amount of oxygen molecules, which is a sub-injection variable (dissolved oxygen amount in the medium, %), can be 15 or more and less than 55. The apparatus 100 can determine a plurality of candidate ranges as 15 - 25, 25 - 35, 35 - 45, and 45 - 55. The apparatus 100 can determine 20, 30, 40, 50, which are the median values of the plurality of candidate ranges, as the candidate values of the amount of oxygen molecules (%). Similarly, the predetermined range of the pH value, which is a sub-injection variable, can be 5.5 or more and less than 8.5. The apparatus 100 can determine a plurality of candidate ranges as 5.5 - 6.5, 6.5 - 7.5, 7.5 - 8.5. In addition, the apparatus 100 can determine 6, 7, 8, which are the median values of the plurality of candidate ranges, as the candidate values of the pH value. Similarly, the predetermined range of the Agitator Speed (RPM) can be 75 or more and 325 or less. In the above manner, the apparatus 100 can determine 100, 150, 200, 250, 300 as the candidate values of the Agitator Speed (RPM).

[0094] The apparatus 100 can perform the step of setting the sub-injection variables not selected among a plurality of sub-injection variables to predetermined basic values, setting the selected sub-injection variables to one of the candidate values, and obtaining a plurality of first injection variables. The predetermined basic values may vary depending on the type of sub-injection variables. For example, the basic value of Vessel temperature may be 36.95°C.

[0095] The apparatus 100 can obtain a plurality of first injection variables based on the combination of candidates of the selected sub-injection variables. For example, if the candidate values of the amount of oxygen molecules (%) are 4, such as 20, 30, 40, 50, the candidate values of the pH value are 3, such as 6, 7, 8, and the candidate values of Agitator Speed (RPM) are 5, such as 100, 150, 200, 250, 300, 60 combinations are created. That is, the plurality of first injection variables can be 60. The above example is the case where three sub-injection variables such as the amount of oxygen molecules (%), the pH value, and Agitator Speed (RPM) are selected. However, it is not limited thereto. That is, the number of selected sub-injection variables may be a value other than 3, and sub-injection variables other than the amount of oxygen molecules (%), the pH value, and Agitator Speed (RPM) can also be selected.

[0096] The apparatus 100 can perform tests on 60 injection variables to generate 60 predicted impact variables and 60 predicted quality attributes for each of the 60 injection variables. Also, the apparatus 100 can select the optimal first injection variable with the best quality attribute. Here, the test may mean steps 330 to 350. Briefly explained, the test may mean the process of applying a plurality of first injection variables to a culture simulation model, predicting a plurality of second quality attributes and a plurality of second impact variables at a second time point, and selecting the first injection variable corresponding to the highest value among the plurality of second quality attributes as the optimal first injection variable.

[0097] The device 100 can perform further tests on a predetermined range for the optimal first injection variable to determine an optimal first injection variable with better quality attributes. For example, the pH value included in the optimal first injection variable can be 7. However, 7 is only a representative value, and there may be an optimal value among the values greater than or equal to 6.5 and less than 7.5. The device 100 can divide the range of 6.5 or more and less than 7.5 into a plurality of candidate ranges, determine candidate values, and determine the optimal first injection variable. The process of determining the optimal first injection variable is the same as the process of determining the optimal first injection variable, so detailed description is omitted.

[0098] Referring to FIG. 3, the device 100 can perform step 330 of applying at least one of the first quality attribute, the first influence variable, and the first injection variable to the culture simulation model to obtain a plurality of second quality attributes and a plurality of second influence variables. The culture simulation model can use the prediction results of a plurality of sub-prediction models. The sub-prediction model can be a prediction model that constitutes the culture simulation model. For example, the device 100 can apply the first quality attribute, the first influence variable, and the first injection variable at the first time point to one sub-prediction model included in the culture simulation model to obtain a second sub-influence variable included in the second influence variable at the second time point. The device 100 can obtain all sub-influence variables included in the second influence variable at the second time point using a plurality of sub-prediction models. Also, the device 100 can obtain the second quality attribute at the second time point based on one sub-prediction model included in the culture simulation model.

[0099] The culture simulation model can be a prediction model formed by combining at least one sub-prediction model. The culture simulation model may include at least one of a glutamine change amount prediction model, a glutamate change amount prediction model, a glucose consumption amount prediction model per VCD, a lactate change amount prediction model, an ammonia change amount prediction model, an osmotic pressure prediction model, a cell division rate prediction model, a viability change amount prediction model, and an IgG production amount prediction model per VCD. The process of obtaining a plurality of second quality attributes and a plurality of second influencing variables using the culture simulation model will be described with reference to FIGS. 4 to 6. The sub-prediction models included in the culture simulation model may be performed in parallel. However, without being limited thereto, the sub-prediction models included in the culture simulation model may be performed serially.

[0100] The sub-prediction models included in the culture simulation model can be generated using an ensemble prediction model. The ensemble prediction model may include NGBoost and XGBoost. The machine learning algorithms for generating the sub-prediction models can use decision trees and linear discriminant analysis including the NGBoost model or the XGBoost model. The ensemble prediction model is a method that uses multiple learning algorithms to obtain better prediction performance compared to the case of using learning algorithms separately. That is, in the case of the ensemble prediction model, when only one machine running model is learned and used for one target task (for example, lactate concentration prediction), the resulting value may be unreliable. Therefore, after other machine running models are also learned together for the same target task to derive respective resulting values, the final prediction value is determined based on the various derived resulting values. For example, in the target task of predicting lactate concentration, if the prediction result of model A is 1 g / L and the prediction result of model B is 2 g / L, it can be said that the ensemble prediction model determines 1.5 g / L, which is the average of these two results, as the final prediction value.

[0101] Also, the ensemble prediction model may include NGBoost and XGBoost based on the Gradient Boosting machine running technique. For example, the ensemble prediction model may first include model A that generally learns a target task with a simple structure. Also, the ensemble prediction model may include model B that learns more detailed parts than model A. Also, the ensemble prediction model may include a model that learns more detailed parts than model B. The ensemble prediction model can output prediction information for the target task using model A or model B, etc.

[0102] The sub-prediction models included in the culture simulation model can each be generated using an ensemble prediction model. That is, among the Glutamine change amount prediction model, Glutamate change amount prediction model, Glucose consumption amount per VCD prediction model, Lactate change amount prediction model, Ammonia change amount prediction model, osmotic pressure prediction model, Cell division rate prediction model, Viability change amount prediction model, and IgG production amount per VCD prediction model, one can be generated using an ensemble prediction model.

[0103] Since the apparatus 100 predicts at least one of the influence variables and quality attributes at the second time point based on the "plurality of" first injection variables, it can generate at least one of the "plurality of" second influence variables and the "plurality of" second quality attributes. In the present invention, the "plurality of" second quality attributes may mean that there are a plurality of sets of second quality attributes including the second VCD and the amount of the second IgG. Also, in the present invention, the "plurality of" second influence variables may mean that there are a plurality of sets of second influence variables including at least one of the amount of the second Glutamine, the amount of the second Glutamate, the amount of the second Glucose, the amount of the second Lactate, the amount of the second Ammonia, and the second osmotic pressure.

[0104] The apparatus 100 can perform the step of obtaining the predicted second influence variables based on two or more types of measured first influence variables. The number of types of influence variables included in the second influence variables can be two or more. The types of the first influence variables and the types of the second influence variables may be the same. However, it is not limited thereto, and the types of the first influence variables and the types of the second influence variables may be different from each other. Also, the number of types of the first influence variables and the number of types of the second influence variables may be different from each other.

[0105] The apparatus 100 can perform a step 340 of selecting second influencing variables and second quality attributes that meet predetermined conditions among a plurality of second influencing variables and a plurality of second quality attributes. The predetermined conditions will be described in connection with FIGS. 8 to 11.

[0106] The apparatus 100 can perform a step 350 of selecting a first injection variable corresponding to the selected second quality attribute among the plurality of first injection variables. Also, the apparatus 100 can determine the selected first injection variable as an optimal injection variable. Further, the apparatus 100 can transmit the optimal injection variable to the bioreactor 1810. Also, the bioreactor 1810 can determine the culture conditions for the first time point based on the received optimal injection variable. Further, the bioreactor 1810 compares the measured value of the culture solution with the received optimal injection variable and controls the culture solution to reach the received optimal injection variable, whereby the bioreactor can produce the maximum amount of antibody.

[0107] FIG. 4 is a flowchart for explaining a culture simulation model according to an embodiment of the present invention. FIG. 5 is a flowchart for explaining a culture simulation model according to an embodiment of the present invention. FIG. 6 is a flowchart for explaining a culture simulation model according to an embodiment of the present invention.

[0108] The step 330 of obtaining a plurality of second quality attributes and a plurality of second influence variables may include the processes of FIGS. 4 to 6. Referring to FIGS. 4 to 6, the apparatus 100 may include a configuration that predicts the amount of Glutamine and the amount of Glutamate, and updates influence variables based on the predicted amount of Glutamine and the predicted amount of Glutamate. Further, the apparatus 100 may include a configuration that predicts the consumption amount of Glucose and updates influence variables based on the predicted consumption amount of Glucose. Further, the apparatus 100 may include a configuration that predicts the amount of Lactate and updates influence variables based on the predicted amount of Lactate. Further, the apparatus 100 may include a configuration that predicts the amount of Ammonia and updates influence variables based on the predicted amount of Ammonia. Further, the apparatus 100 can predict the Osmolarity and update influence variables based on the predicted Osmolarity. Further, the apparatus 100 can predict the cell division rate and the cell death rate, and predict the VCD value or the amount of IgG included in the quality attribute based on the predicted cell division rate and the predicted cell death rate. Hereinafter, the above-described configuration will be described in more detail.

[0109] Referring to FIG. 4, the apparatus 100 can perform step 410 of applying the first quality attribute, the first influence variable, and the plurality of first injection variables to the Glutamine change amount prediction model to obtain the plurality of second Glutamine amounts included in the plurality of second influence variables.

[0110] The second influencing variable may include at least one of the amount of the second Glutamine, the amount of the second Glutamate, the amount of the second Glucose, the amount of the second Lactate, the amount of the second Ammonia, and the second Osmolarity. In the present invention, the "amount" may be density. For example, the unit of the amount may be the same as cells / mL, mmol / L, or g / L. However, it is not limited thereto, and in the present invention, the amount may mean the absolute amount of a substance. For example, the unit of the amount may be the number of cells, mol, or g. Also, in the present invention, the amount may mean the ratio of the substance to the total substance. For example, the unit may be %.

[0111] The reason for obtaining a plurality of amounts of the second Glutamine is to use a plurality of first injection variables. That is, the plurality of first injection variables can correspond one-to-one with the plurality of amounts of the second Glutamine. As already explained, the plurality of first injection variables are combinations generated using candidate values of sub-injection variables included in the injection variables. One of the plurality of first injection variables can be used to obtain one of the plurality of amounts of the second Glutamine.

[0112] Also, the apparatus 100 can perform a step of obtaining a first-1 influencing variable based on one of the plurality of amounts of the second Glutamine. According to an embodiment of the present invention, the apparatus 100 can perform step 420 of obtaining a first-1 influencing variable by replacing the amount of the first Glutamine included in the first influencing variable with one of the plurality of amounts of the second Glutamine. That is, the apparatus 100 can use the first influencing variable changed for the steps after step 420, and the changed first influencing variable can be the first-1 influencing variable.

[0113] As already described, the apparatus 100 can generate a plurality of amounts of a second Glutamine using a plurality of first injection variables. Also, the apparatus 100 can update a first influence variable using the plurality of amounts of the second Glutamine. Therefore, the apparatus 100 can obtain a plurality of first-1 influence variables that correspond one-to-one with the plurality of amounts of the second Glutamine. Also, since the plurality of first injection variables correspond one-to-one with the plurality of amounts of the second Glutamine, the plurality of first-1 influence variables can correspond one-to-one with the plurality of first injection variables. The apparatus 100 can generate one amount of the second Glutamine using one of the plurality of first injection variables. Also, the apparatus 100 can obtain a first-1 influence variable corresponding to the one of the plurality of first injection variables. The apparatus 100 can apply one of the plurality of first injection variables and the first-1 influence variable corresponding to the one of the plurality of first injection variables to a sub-prediction model at a stage after stage 420.

[0114] Also, stage 420 can be replaced with the following stages. That is, the apparatus 100 can perform stage 420 of obtaining a first-1 influence variable by replacing the amount of the first Glutamine included in the first influence variable with a weighted average of the amount of the first Glutamine and the amount of the second Glutamine. The weighted average may be the same as w1*(amount of the first Glutamine)+w2*(amount of the second Glutamine). Here, w1 + w2 = 1, and w1 and w2 can be between 0 and 1 inclusive.

[0115] The device 100 can perform step 430 of applying the first quality attribute, the first-1 influencing variable, and the plurality of first injection variables to the Glutamate change amount prediction model to obtain the amounts of the plurality of second Glutamates included in the plurality of second influencing variables. The reason for predicting Glutamine and Glutamate first, like in step 410 and step 430, is that these are the variables most directly affected by the injection of the nutrient medium (Feed). When performing step 430, the device 100 can replace the first-1 influencing variable with the first influencing variable. That is, the device 100 can perform step 430 of applying the first quality attribute, the first influencing variable, and the plurality of first injection variables to the Glutamate change amount prediction model to obtain the amounts of the plurality of second Glutamates included in the plurality of second influencing variables.

[0116] The reason for obtaining the amounts of the plurality of second Glutamates is to use the plurality of first injection variables. That is, the plurality of first injection variables can correspond one-to-one with the amounts of the plurality of second Glutamates. One of the plurality of first injection variables can be used to obtain one of the amounts of the plurality of second Glutamates.

[0117] Also, the device 100 can perform the step of obtaining the first-2 influencing variable based on one of the amounts of the plurality of second Glutamates. The device 100 according to an embodiment of the present invention can perform step 440 of obtaining the first-2 influencing variable by replacing the amount of the first Glutamate included in the first-1 influencing variable with one of the amounts of the plurality of second Glutamates. That is, the device 100 can use the changed first influencing variable for the steps after step 440, and the changed first influencing variable can be the first-2 influencing variable.

[0118] The device 100 can acquire a plurality of first-2 influence variables that correspond one-to-one with the amounts of a plurality of second Glutamates. Also, the plurality of first-2 influence variables can correspond one-to-one with a plurality of first injection variables. The device 100 can generate an amount of one second Glutamate using one of the plurality of first injection variables at stage 430. Also, the device 100 can acquire a first-2 influence variable corresponding to the one first injection variable at stage 440. The device 100 can apply the one first injection variable and one first-2 influence variable corresponding to the one first injection variable among the plurality of first-2 influence variables to a sub-prediction model at stages after stage 440.

[0119] Also, stage 440 can be replaced with the following stages. That is, the device 100 can perform stage 440 of acquiring a first-2 influence variable by replacing the amount of the first Glutamate included in the first-1 influence variable with a weighted average of the amount of the first Glutamate and the amount of the second Glutamate. The weighted average may be the same as w3*(amount of the first Glutamate) + w4*(amount of the second Glutamate). Here, w3 + w4 = 1, and w3 and w4 can be between 0 and 1 inclusive.

[0120] Referring to FIG. 5, the apparatus 100 can perform step 510 of applying the first quality attribute, the first - 2 influence variables, and the plurality of first injection variables to the Glucose consumption prediction model per VCD to obtain a plurality of Glucose consumption amounts. Also, the apparatus 100 can perform step 520 of subtracting the plurality of Glucose consumption amounts from the amount of the first Glucose included in the first influence variable to obtain the amounts of the plurality of second Glucose included in the plurality of second influence variables. When performing step 520, the apparatus 100 can replace the first - 2 influence variables with the first influence variable or the first - 1 influence variable. That is, the apparatus 100 can perform step 510 of applying at least one of the first quality attribute, the first influence variable, the first - 1 influence variable, the first - 2 influence variable, and the plurality of first injection variables to the Glucose consumption prediction model per VCD to obtain a plurality of Glucose consumption amounts.

[0121] The reason for obtaining the plurality of Glucose consumption amounts and the amounts of the plurality of second Glucose is to use the plurality of first injection variables. That is, the plurality of first injection variables can correspond one - to - one with the plurality of Glucose consumption amounts. Also, the plurality of first injection variables can correspond one - to - one with the amounts of the plurality of second Glucose. Also, one of the plurality of first injection variables can be used to obtain one of the plurality of Glucose consumption amounts and one of the amounts of the plurality of second Glucose.

[0122] If one of the plurality of Glucose consumption amounts is greater than the amount of the first Glucose at the first time point, the apparatus 100 can output a warning if Glucose depletion is expected. Or, the apparatus 100 can remove the first injection variable corresponding to the Glucose consumption amount from the plurality of first injection variables. That is, the apparatus 100 may not use the first injection variable predicted to cause the Glucose consumption amount to be greater than the amount of the first Glucose at the first time point to find the optimal injection variable.

[0123] In addition, the apparatus 100 can perform a step of obtaining the first to third influence variables based on one of the plurality of second Glucose amounts. The apparatus 100 according to an embodiment of the present invention can perform a step 530 of obtaining the first to third influence variables by replacing the amount of the first Glucose included in the first to second influence variables with one of the plurality of second Glucose amounts. That is, the apparatus 100 can use the changed first influence variable for the steps after step 530, and the changed first influence variable can be the first to third influence variables.

[0124] The apparatus 100 can obtain a plurality of first to third influence variables that correspond one-to-one with the plurality of second Glucose amounts. In addition, the plurality of first to third influence variables can correspond one-to-one with the plurality of first injection variables. The apparatus 100 can generate one amount of the second Glucose using one of the plurality of first injection variables in step 520. In addition, the apparatus 100 can obtain the first to third influence variables corresponding to the one first injection variable in step 530. The apparatus 100 can apply the one first injection variable and the one first to third influence variable corresponding to the one first injection variable among the plurality of first to third influence variables to a sub-prediction model in the steps after step 530.

[0125] In addition, step 530 can be replaced with the following steps. That is, the apparatus 100 can perform step 530 of obtaining the first to third influence variables by replacing the amount of the first Glucose included in the first to second influence variables with a weighted average of the amount of the first Glucose and the amount of the second Glucose. The weighted average may be the same as w5*(amount of the first Glucose)+w6*(amount of the second Glucose). Here, w5+w6 = 1, and w5 and w6 can be between 0 and 1.

[0126] Referring to FIG. 5, the apparatus 100 can perform a step 540 of applying the first quality attribute, the first to third influence variables, and the plurality of first injection variables to a Lactate change amount prediction model to obtain a plurality of amounts of the second Lactate included in the plurality of second influence variables.

[0127] The reason for obtaining the amounts of a plurality of second Lactates is to use a plurality of first injection variables. That is, the plurality of first injection variables can correspond one-to-one with the amounts of the plurality of second Lactates. Also, one of the amounts of the plurality of second Lactates can be obtained by using one of the plurality of first injection variables.

[0128] When performing step 540, the apparatus 100 can replace the first - third influencing variables with at least one of the first influencing variable, the first - first influencing variable, or the first - second influencing variable. That is, the apparatus 100 can perform step 540 of applying at least one of the first quality attribute, the first influencing variable, the first - first influencing variable, the first - second influencing variable, the first - third influencing variable, and the plurality of first injection variables to the Lactate change amount prediction model to obtain the amounts of the plurality of second Lactates included in the plurality of second influencing variables.

[0129] Also, the apparatus 100 can perform a step of obtaining a first - fourth influencing variable based on one of the amounts of the plurality of second Lactates. The apparatus 100 can perform step 550 of obtaining the first - fourth influencing variable by replacing the amount of the first Lactate included in the first - third influencing variable with one of the amounts of the plurality of second Lactates. That is, the apparatus 100 can use the changed first influencing variable for the steps after step 550, and the changed first influencing variable can be the first - fourth influencing variable.

[0130] The device 100 can obtain a plurality of first to fourth influencing variables that correspond one-to-one with a plurality of amounts of the second Lactate. Also, the plurality of first to fourth influencing variables can correspond one-to-one with a plurality of first injection variables. The device 100 can generate one amount of the second Lactate using one of the plurality of first injection variables at stage 540. Also, the device 100 can obtain the first to fourth influencing variables corresponding to the one first injection variable at stage 550. The device 100 can apply the one first injection variable and one of the first to fourth influencing variables corresponding to the one first injection variable among the plurality of first to fourth influencing variables to a sub-prediction model at stages after stage 550.

[0131] Also, stage 550 can be replaced with the following stages. That is, the device 100 can perform stage 550 of obtaining the first to fourth influencing variables by replacing the amount of the first Lactate included in the first to third influencing variables with a weighted average of the amount of the first Lactate and the amount of the second Lactate. The weighted average may be the same as w7*(amount of the first Lactate)+w8*(amount of the second Lactate). Here, w7 + w8 = 1, and w7 and w8 can be between 0 and 1.

[0132] Referring to FIG. 5, the device 100 can perform stage 560 of applying the first quality attribute, the first to fourth influencing variables, and the plurality of first injection variables to an Ammonia change amount prediction model to obtain a plurality of amounts of the second Ammonia included in the plurality of second influencing variables.

[0133] The reason for obtaining the plurality of amounts of the second Ammonia is to use the plurality of first injection variables. That is, the plurality of first injection variables can correspond one-to-one with the plurality of amounts of the second Ammonia. Also, one of the plurality of amounts of the second Ammonia can be obtained using one of the plurality of first injection variables.

[0134] When performing step 560, the apparatus 100 can replace the first to fourth influencing variables with at least one of the first influencing variable, the first - 1 influencing variable, the first - 2 influencing variable, or the first - 3 influencing variable. That is, the apparatus 100 can perform step 560 of applying at least one of the first quality attribute, the first influencing variable, the first - 1 influencing variable, the first - 2 influencing variable, the first - 3 influencing variable, the first - 4 influencing variable, and the plurality of first injection variables to the Ammonia change amount prediction model to obtain the plurality of second Ammonia amounts included in the plurality of second influencing variables.

[0135] Also, the apparatus 100 can perform a step of obtaining a first - 5 influencing variable based on one of the plurality of second Ammonia amounts. For example, step 570 of replacing the first Ammonia amount included in the first - 4 influencing variable with one of the plurality of second Ammonia amounts to obtain the first - 5 influencing variable can be performed. That is, the apparatus 100 can use the changed first influencing variable for the steps after step 570, and the changed first influencing variable can be the first - 5 influencing variable.

[0136] The apparatus 100 can obtain a plurality of first - 5 influencing variables that correspond one - to - one with the plurality of second Ammonia amounts. Also, the plurality of first - 5 influencing variables can correspond one - to - one with the plurality of first injection variables. The apparatus 100 can generate one second Ammonia amount using one of the plurality of first injection variables in step 560. Also, the apparatus 100 can obtain the first - 5 influencing variable corresponding to the one first injection variable in step 570. The apparatus 100 can apply the one first injection variable and the one first - 5 influencing variable corresponding to the one first injection variable among the plurality of first - 5 influencing variables to the sub - prediction model in the steps after step 570.

[0137] Also, step 570 can be replaced with the following steps. That is, the apparatus 100 can perform step 570 of obtaining the 1st - 5th influence variables by replacing the amount of the 1st Ammonia included in the 1st - 4th influence variables with the weighted average of the amount of the 1st Ammonia and the amount of the 2nd Ammonia. The weighted average may be the same as w9*(the amount of the 1st Ammonia)+w10*(the amount of the 2nd Ammonia). Here, w9 + w10 = 1, and w9 and w10 can be between 0 and 1 inclusive.

[0138] Referring to FIG. 5, the apparatus 100 can perform step 580 of applying the 1st quality attribute, the 1st - 5th influence variables, and the plurality of 1st injection variables to the osmotic pressure prediction model to obtain the plurality of 2nd osmotic pressures included in the plurality of 2nd influence variables.

[0139] The reason for obtaining the plurality of 2nd osmotic pressures is to use the plurality of 1st injection variables. That is, the plurality of 1st injection variables can correspond one - to - one with the plurality of 2nd osmotic pressures. Also, one of the plurality of 1st injection variables can be used to obtain one of the plurality of 2nd osmotic pressures.

[0140] When performing step 580, the device 100 can replace the first to fifth influencing variables with one of the first influencing variable, the first - 1 influencing variable, the first - 2 influencing variable, the first - 3 influencing variable, or the first - 4 influencing variable. That is, the device 100 can perform step 580 of applying at least one of the first quality attribute, the first influencing variable, the first - 1 influencing variable, the first - 2 influencing variable, the first - 3 influencing variable, the first - 4 influencing variable, the first - 5 influencing variable, and a plurality of first injection variables to the osmotic pressure prediction model to obtain a plurality of second osmotic pressures included in the plurality of second influencing variables. Also, the device 100 can perform a step of obtaining a first - 6 influencing variable based on one of the plurality of second osmotic pressures. More specifically, it can perform step 590 of obtaining the first - 6 influencing variable by replacing the first osmotic pressure included in the first - 6 influencing variable with one of the plurality of second osmotic pressures. That is, the device 100 can use the changed first influencing variable for the steps after step 590, and the changed first influencing variable can be the first - 6 influencing variable.

[0141] The device 100 can obtain a plurality of first - 6 influencing variables that correspond one - to - one with the plurality of second osmotic pressures. Also, the plurality of first - 6 influencing variables can correspond one - to - one with the plurality of first injection variables. The device 100 can generate one second osmotic pressure using one of the plurality of first injection variables in step 580. Also, the device 100 can obtain the first - 6 influencing variable corresponding to the one first injection variable in step 590. The device 100 can apply the one first injection variable and the one first - 6 influencing variable corresponding to the one first injection variable among the plurality of first - 6 influencing variables to the sub - prediction model in the steps after step 590.

[0142] Also, step 590 can be replaced with the following steps. That is, the apparatus 100 can perform step 590 of obtaining the first to sixth influencing variables by replacing the first osmotic pressure included in the first to fifth influencing variables with a weighted average of the first osmotic pressure and the second osmotic pressure. The weighted average may be the same as w11*(amount of the first Ammonia)+w12*(amount of the second Ammonia). Here, w11 + w12 = 1, and w11 and w12 can be between 0 and 1.

[0143] FIG. 6 illustrates the process of updating the quality attribute using the influencing variables changed as described above. The changed influencing variables can be the first to sixth influencing variables. That is, FIG. 6 shows the process of predicting the quality attribute at the second time point.

[0144] Referring to FIG. 6, the apparatus 100 can perform step 610 of applying the first quality attribute, the first to sixth influencing variables, and a plurality of first injection variables to the Cell division rate prediction model to obtain the division rates of a plurality of cells. The division rate of the cells can be the ratio at which the cells at the first time point divide. For example, a division rate of 300% may mean that the cells have increased threefold.

[0145] When performing step 610, the apparatus 100 can replace the first to sixth influencing variables with one of the first influencing variable, the first - 1 influencing variable, the first - 2 influencing variable, the first - 3 influencing variable, the first - 4 influencing variable, or the first - 5 influencing variable. The apparatus 100 can perform step 610 of applying at least one of the first quality attribute, the first influencing variable, the first - 1 influencing variable, the first - 2 influencing variable, the first - 3 influencing variable, the first - 4 influencing variable, the first - 5 influencing variable, the first - 6 influencing variable, and a plurality of first injection variables to the Cell division rate prediction model to obtain the division rates of a plurality of cells.

[0146] The reason for obtaining the division rates of multiple cells is to use multiple first injection variables. That is, the multiple first injection variables can correspond one-to-one with the division rates of the multiple cells. Also, one of the multiple first injection variables can be used to obtain one of the division rates of the multiple cells.

[0147] Device 100 can perform step 620 of applying the first quality attribute, the first - 6 influencing variables, and the multiple first injection variables to the Viability change prediction model to obtain the mortality rates of multiple cells. The mortality rate of the cells can be the ratio of the cells that died at the first time point. For example, a cell mortality rate of 30% can mean that 70% survived. For example, the amount of cells at the second time point may be the same as (the amount of cells at the first time point) * the division rate of the cells * (100% - the mortality rate of the cells). The amount of cells at the second time point can be the second VCD value.

[0148] When performing step 620, device 100 can replace the first - 6 influencing variables with one of the first influencing variable, the first - 1 influencing variable, the first - 2 influencing variable, the first - 3 influencing variable, the first - 4 influencing variable, or the first - 5 influencing variable. Device 100 can perform step 620 of applying at least one of the first quality attribute, the first influencing variable, the first - 1 influencing variable, the first - 2 influencing variable, the first - 3 influencing variable, the first - 4 influencing variable, the first - 5 influencing variable, the first - 6 influencing variable, and the multiple first injection variables to the Viability change prediction model to obtain the mortality rates of multiple cells.

[0149] The reason for obtaining the mortality rates of multiple cells is to use multiple first injection variables. That is, the multiple first injection variables can correspond one-to-one with the mortality rates of the multiple cells. Also, one of the multiple first injection variables can be used to obtain one of the mortality rates of the multiple cells.

[0150] The apparatus 100 can perform step 630 of obtaining a plurality of second viable cell density (VCD) values included in a plurality of second quality attributes based on at least one of a first VCD value included in a first quality attribute, a division rate of a plurality of cells, and a mortality rate of a plurality of cells. The second VCD value can be a predicted VCD value at a second time point.

[0151] The reason for obtaining a plurality of second VCD values is to use a plurality of first injection variables. That is, a plurality of first injection variables can correspond one-to-one with a plurality of second VCD values. Also, one of the plurality of first injection variables can be used to obtain one of the plurality of second VCD values. The apparatus 100 can select a second VCD value that meets a preset condition among the plurality of second VCD values. Also, the apparatus 100 can select a first injection variable corresponding to the selected second VCD value as an optimal first injection variable.

[0152] The apparatus 100 can perform step 640 of applying the first quality attribute, the first - 6 influencing variables, and the plurality of first injection variables to an IgG production amount prediction model per VCD value to obtain a plurality of IgG production amounts per VCD.

[0153] When performing step 640, the apparatus 100 can replace the first - 6 influencing variables with one of the first influencing variable, the first - 1 influencing variable, the first - 2 influencing variable, the first - 3 influencing variable, the first - 4 influencing variable, or the first - 5 influencing variable.

[0154] The reason for obtaining a plurality of IgG production amounts per VCD is to use a plurality of first injection variables. That is, a plurality of first injection variables can correspond one-to-one with a plurality of IgG production amounts per VCD. Also, one of the plurality of first injection variables can be used to obtain one of the plurality of IgG production amounts per VCD.

[0155] The apparatus 100 can perform step 650 of multiplying the IgG production amount per VCD in a plurality of second VCDs to obtain the amount of a plurality of second IgGs (Immunoglobulin G) included in a plurality of second quality attributes. The apparatus 100 can select a second IgG that meets a preset condition among the plurality of second IgGs. Further, the apparatus 100 can select a first injection variable corresponding to the selected second IgG as an optimal first injection variable. Further, the apparatus 100 can transmit the optimal injection variable to the bioreactor 1810. Further, the bioreactor 1810 can determine the culture conditions for the first time point based on the received optimal injection variable. Through such a process, the bioreactor can produce the maximum amount of antibody.

[0156] FIG. 7 is a block diagram for explaining a prediction model according to an embodiment of the present invention. The culture simulation model may include at least one of a Glutamine change amount prediction model, a Glutamate change amount prediction model, a Glucose consumption amount prediction model per VCD, a Lactate change amount prediction model, an Ammonia change amount prediction model, an osmotic pressure prediction model, a Cell division rate prediction model, a Viability change amount prediction model, and an IgG production amount prediction model per VCD.

[0157] The prediction model 730 in FIG. 7 can be one of the Glutamine change amount prediction model, Glutamate change amount prediction model, Glucose consumption amount prediction model per VCD, Lactate change amount prediction model, Ammonia change amount prediction model, osmotic pressure prediction model, Cell division rate prediction model, Viability change amount prediction model, and IgG production amount prediction model per VCD. For the Glutamine change amount prediction model, Glutamate change amount prediction model, Glucose consumption amount prediction model per VCD, Lactate change amount prediction model, Ammonia change amount prediction model, osmotic pressure prediction model, Cell division rate prediction model, Viability change amount prediction model, and IgG production amount prediction model per VCD included in the culture simulation model, the same explanation is possible. Therefore, in FIG. 7, they are collectively described as the prediction model 730.

[0158] The device 100 can obtain a prediction model 730 that machine-learns the correlation between the data 710 at the past k-th time point and the sub-variables 720 at the past k+1-th time point. The machine learning algorithm for generating the machine learning prediction model 730 may include latent variable modeling including at least one of a neural network, a support vector machine, or partial least squares analysis. Also, the machine learning prediction model 730 can be generated using an ensemble prediction model, and the machine learning algorithm for generating the ensemble prediction model can use a decision tree including an NGBoost model or an XGBoost model and linear discriminant analysis. The sub-prediction models included in the culture simulation model can be generated using an ensemble prediction model. That is, one of the Glutamine change amount prediction model, Glutamate change amount prediction model, Glucose consumption amount prediction model per VCD, Lactate change amount prediction model, Ammonia change amount prediction model, osmotic pressure prediction model, Cell division rate prediction model, Viability change amount prediction model, and IgG production amount prediction model per VCD can be generated using an ensemble prediction model.

[0159] In addition, the machine learning algorithm can use a reduced order time varying autoregressive exogenous model (ARX model). The process of obtaining the prediction model 730 from the data 710 at the k-th past time point and the sub-variable 720 at the (k + 1)-th past time point may be performed by the data learning unit 110 of the apparatus 100.

[0160] The data 710 at the k-th past time point and the sub-variable 720 at the (k + 1)-th past time point can be obtained based on the data measured in real time using a Raman spectrometer. The (k + 1)-th time point can be after the k-th time point. The data 710 at the k-th past time point may include at least one of the quality attribute 711 at the k-th past time point, the influence variable 712 at the k-th past time point, and the injection variable 713 at the k-th past time point. The quality attribute 711 at the k-th past time point, the influence variable 712 at the k-th past time point, and the injection variable 713 at the k-th past time point can be a data set for machine learning. Since the quality attribute, the influence variable, and the injection variable have been described above, detailed descriptions thereof are omitted.

[0161] Also, the sub-variable at the past time point k+1 can be the label information corresponding to the data set. That is, the sub-variable 720 at the past time point k+1 can be the actual (ground truth) information. The sub-variable 720 at the past time point k+1 may vary depending on the type of the prediction model 730. In the case of the Glutamine change amount prediction model, the sub-variable 720 at the time point k+1 can be the amount of Glutamine at the time point k+1. In the case of the Glutamate change amount prediction model, the sub-variable 720 at the time point k+1 can be the amount of Glutamate at the time point k+1. In the case of the Glucose consumption amount prediction model per VCD, the sub-variable 720 at the time point k+1 can be the Glucose consumption amount at the time point k+1. In the case of the Lactate change amount prediction model, the sub-variable 720 at the time point k+1 can be the amount of Lactate at the time point k+1. In the case of the Ammonia change amount prediction model, the sub-variable 720 at the time point k+1 can be the amount of Ammonia at the time point k+1. In the case of the osmotic pressure prediction model, the sub-variable 720 at the time point k+1 can be the osmotic pressure at the time point k+1. In the case of the Cell division rate prediction model, the sub-variable 720 at the time point k+1 can be the cell division rate at the time point k+1. In the case of the Viability change amount prediction model, the sub-variable 720 at the time point k+1 can be the cell death rate at the time point k+1. Also, in the case of the IgG production amount prediction model per VCD, the sub-variable 720 at the time point k+1 can be the IgG per VCD at the time point k+1.

[0162] The device 100 can store the generated prediction model 730 or transmit it to other devices. The device 100 can receive the prediction model 730 from an external device.

[0163] The device 100 can acquire the second sub-variable using the prediction model 730. The process of applying at least one of the first quality attribute 741, the first influence variable 742, and the first injection variable 743 to the prediction model 730 to acquire the second sub-variable 750 may be performed by the data recognition unit 120 of the device 100.

[0164] The first quality attribute 741, the first influencing variable 742, and the first injection variable 743 can be values at the first point in time. The second sub-variable 750 can be a value at the second point in time. The first point in time can be before the second point in time. The difference between the first point in time and the second point in time can be 10 minutes, 1 hour, or 1 day, etc. The second sub-variable 750 can be a value predicted by the prediction model 730. Therefore, the second sub-variable 750 can be somewhat different from the actual information.

[0165] The second sub-variable 750 may vary depending on the type of the prediction model 730. In the case of a glutamine change amount prediction model, the second sub-variable 750 can be the predicted amount of glutamine at the second point in time. In the case of a glutamate change amount prediction model, the second sub-variable 750 can be the predicted amount of glutamate at the second point in time. In the case of a glucose consumption amount per VCD prediction model, the second sub-variable 750 can be the predicted glucose consumption amount at the second point in time. In the case of a lactate change amount prediction model, the second sub-variable 750 can be the predicted amount of lactate at the second point in time. In the case of an ammonia change amount prediction model, the second sub-variable 750 can be the predicted amount of ammonia at the second point in time. In the case of an osmotic pressure prediction model, the second sub-variable 750 can be the predicted osmotic pressure at the second point in time. In the case of a cell division rate prediction model, the second sub-variable 750 can be the predicted cell division rate at the second point in time. In the case of a viability change amount prediction model, the second sub-variable 750 can be the predicted cell death rate at the second point in time. Also, in the case of an IgG production amount per VCD prediction model, the second sub-variable 750 can be the predicted IgG per VCD at the second point in time. At least one of the above prediction models can be combined to form a culture simulation model.

[0166] FIG. 8 is a diagram for explaining predetermined conditions according to an embodiment of the present invention. FIG. 9 is a diagram for explaining predetermined conditions according to an embodiment of the present invention. FIG. 10 is a diagram for explaining predetermined conditions according to an embodiment of the present invention. FIG. 11 is a diagram for explaining predetermined conditions according to an embodiment of the present invention.

[0167] In the case of the first period, the apparatus 100 can perform a step of changing the first injection variable so that the second VCD value included in the second quality attribute has a maximum value. Further, in the case of the second period, the apparatus 100 can perform a step of changing the first injection variable so that the amount of the second IgG included in the second quality attribute has a maximum value.

[0168] The apparatus 100 can perform tests on a plurality of first injection variables to change the first injection variable. Further, performing the test may mean obtaining a predicted second influence variable and a predicted second quality attribute for each of the plurality of first injection variables. The apparatus 100 can select a first injection variable that meets predetermined conditions among the plurality of first injection variables. The predetermined conditions may mean that the second VCD value included in the predicted second quality attribute has a maximum value or the amount of the second IgG has a maximum value. The selected first injection variable may be the changed first injection variable. Hereinafter, the operation of the apparatus 100 will be described in more detail.

[0169] Referring to FIG. 8, the apparatus 100 can determine the optimal injection variable by selecting the influence variable or quality attribute that meets the predetermined conditions among the plurality of influence variables or quality attributes, and selecting the injection variable corresponding to the selected influence variable or quality attribute. In this way, the apparatus 100 can set an objective function to select the optimal first injection variable among the plurality of first injection variables. The objective function can be a condition that the influence variable or quality attribute must satisfy. For example, the objective function can be defined such that the influence variable does not exceed a critical value, or the quality attribute has a maximum value. For example, the objective function can be a function that selects the maximum among a plurality of second VCD values. Also, the objective function can be a function that selects the maximum among a plurality of second IgG amounts. Further, the objective function can be a function that selects the maximum among a plurality of second VCD values during the first culture period, and a function that selects the maximum among a plurality of second IgG amounts during the second period. The first period can be before the second period. As an example, the first period can be from the 4th day to less than the 8th day, and the second period can be 8th day or more. Special control of the medium may not be performed before the first period. That is, the objective function can be shown as in FIG. 8. In FIG. 8, VCD predicted can correspond to a plurality of second VCD values. Also, in FIG. 8, IgG predicted can correspond to a plurality of second IgG amounts.

[0170] When the objective function is as shown in FIG. 8, the apparatus 100 can further perform the following steps to perform step 340 of selecting the second influence variable and the second quality attribute. When within the first period, the apparatus 100 can perform the step of selecting the maximum value among the plurality of second VCD values included in the plurality of second quality attributes as the second quality attribute. Also, when within the second period, the apparatus 100 can perform the step of selecting the maximum value among the plurality of second IgG amounts included in the plurality of second quality attributes as the second quality attribute. The first period can be a predetermined time range during which cells are cultured in the cell culture medium. The second period can be after the first period. The second period may not overlap with the first period.

[0171] Also, the first time point may mean a time point having a measurement value. Also, the second time point is a time point without a measurement value, but is a time point at which a predicted value can be obtained based on the measurement value at the first time point. The second time point may be after a predetermined time has elapsed from the first time point. The predetermined time may be 10 minutes, 1 hour, or 1 day. Also, since the first time point is a time point at which a measurement value can be measured, it can change as time passes. For example, the first time point may be the present. The second time point is after a predetermined time has elapsed from the first time point, and as the first time point changes, the second time point can also change.

[0172] Both the first period and the second period may include the first time point and the second time point. That is, in the first period, after the apparatus 100 obtains a measurement value (the first time point), it can obtain a predicted value for the second time point after a predetermined time has elapsed at the first time point based on the measurement value. Also, in the second period, after the apparatus 100 obtains a measurement value (the first time point), it can obtain a predicted value for the second time point after a predetermined time has elapsed at the first time point based on the measurement value.

[0173] The apparatus 100 can obtain a predicted value for the second time point (future time point) based on the measurement value measured at the first time point (present) included in the first period. The measurement value may be a first influencing variable and a first quality attribute. Also, the predicted value may be a plurality of second influencing variables and a plurality of second quality attributes. For the predicted value in the first period, the apparatus 100 can select the maximum value among the plurality of second VCD values. Also, for the predicted value in the second period, the apparatus 100 can select the maximum value among the plurality of second IgG amounts.

[0174] In addition, the apparatus 100 can perform step 350 of selecting a first injection variable corresponding to the selected second quality attribute among a plurality of first injection variables. Further, the apparatus 100 can determine the selected first injection variable as an optimal injection variable and transmit it to the bioreactor. Also, the bioreactor can determine the culture conditions for the first time point based on the received optimal injection variable. Through such a process, the bioreactor can maximize the production of antibodies.

[0175] Referring to FIG. 9, the user may want to minimize the amount of Ammonia while satisfying the conditions of FIG. 8. The apparatus 100 can perform the step of changing the first injection variable such that the value obtained by multiplying the second VCD value included in the second quality attribute by e^(-k * the amount of the second Ammonia) has a maximum value during the first period. Also, the apparatus 100 can perform the step of changing the first injection variable such that the value obtained by multiplying the amount of the second IgG included in the second quality attribute by e^(-k * the amount of the second Ammonia) has a maximum value during the second period. The apparatus 100 can perform tests on a plurality of first injection variables in order to change the first injection variable. Also, performing a test may mean obtaining a predicted second influence variable and a predicted second quality attribute for each of the plurality of first injection variables. The apparatus 100 can select a first injection variable that meets predetermined conditions among the plurality of first injection variables. The predetermined conditions may mean that the value obtained by multiplying the second VCD value included in the predicted second quality attribute by e^(-k * the amount of the second Ammonia) has a maximum value, or the value obtained by multiplying the amount of the second IgG by e^(-k * the amount of the second Ammonia) has a maximum value. The selected first injection variable can be the changed first injection variable. Below, the operation of the apparatus 100 will be described in more detail.

[0176] Referring to FIG. 9, the user may want to minimize the amount of Ammonia while satisfying the conditions of FIG. 8. In this case, the objective function can be shown as in FIG. 9. In FIG. 9, Ammonia predictedcan correspond to a plurality of second Ammonia amounts. The apparatus 100 can maintain the VCD at its maximum or the IgG at its maximum while minimizing the second Ammonia amount in order to maximize the VCD value or the amount of IgG predicted xe -0.1Ammonia_predicted while maintaining it at its maximum, or the IgG predicted xe -0.1Ammonia_predicted while maintaining it at its maximum.

[0177] The plurality of first injection variables can correspond one-to-one to a plurality of second VCD values. Also, the plurality of first injection variables can correspond one-to-one to a plurality of second IgG amounts. Also, the plurality of first injection variables can correspond one-to-one to a plurality of second Ammonia amounts. The apparatus 100 can substitute one of the plurality of second VCD values among the plurality of second VCD values and the amount of second Ammonia corresponding to the one second VCD value into the equation of FIG. 9. Also, the apparatus 100 can substitute one of the plurality of second IgG amounts among the plurality of second IgG amounts and the amount of second Ammonia corresponding to the one second IgG amount into the equation of FIG. 9.

[0178] The apparatus 100 can further perform the following process to perform step 340 of selecting the second influencing variable and the second quality attribute. Referring to FIG. 9, when in the first period, the apparatus 100 can perform the step of selecting, as the second quality attribute, the second VCD value when the value obtained by multiplying a plurality of second VCD values included in the plurality of second quality attributes by e^(-k * amount of second Ammonia) is the maximum value. For example, the first period can be 4 days or more and less than 8 days. Also, k can be a positive number. For example, k can have a value greater than 0 and less than or equal to 1. For example, k can be 0.1. Also, when in the second period, the apparatus 100 can perform the step of selecting, as the second quality attribute, the second IgG amount when the value obtained by multiplying a plurality of second IgG amounts included in the plurality of second quality attributes by e^(-k * amount of second Ammonia) is the maximum value. Also, k can be a positive number. For example, k can have a value greater than 0 and less than or equal to 1. For example, k can be 0.1.

[0179] Since e^(-k * the amount of the second Ammonia) decreases as the amount of the second Ammonia increases and increases as the amount of the second Ammonia decreases, in order for the value obtained by multiplying e^(-k * the amount of the second Ammonia) by a plurality of the second VCDs or the value obtained by multiplying e^(-k * the amount of the second Ammonia) by a plurality of the amounts of the second IgG to reach the maximum value, the second VCD value or the amount of the second IgG must be the maximum, and the amount of the second Ammonia must be the minimum.

[0180] In addition, the apparatus 100 can perform step 350 of selecting a first injection variable corresponding to the selected second quality attribute among a plurality of first injection variables. Also, the apparatus 100 can determine the selected first injection variable as an optimal injection variable and transmit it to the bioreactor 1810. Further, the bioreactor 1810 can determine the culture conditions for the first time point based on the received optimal injection variable. Through such a process, the bioreactor 1810 can produce the maximum amount of antibodies.

[0181] Referring to FIG. 10, the user may want to limit the amount of Lactate to a first critical amount while still satisfying the conditions of FIG. 8 or FIG. 9. The apparatus 100 can perform a step of determining whether the amount of the second Lactate included in the second influencing variable is less than or equal to the first critical amount. When the amount of the second Lactate is less than or equal to the first critical amount and within the first period, the apparatus 100 can perform a step of changing the first injection variable so that the second VCD value included in the second quality attribute has a maximum value. Also, when the amount of the second Lactate is less than or equal to the first critical amount and within the second period, the apparatus 100 can perform a step of changing the first injection variable so that the amount of the second IgG included in the second quality attribute has a maximum value. The apparatus 100 can perform tests on a plurality of first injection variables to change the first injection variable. Also, performing the tests may mean obtaining the predicted second influencing variable and the predicted second quality attribute for each of the plurality of first injection variables. The apparatus 100 can select a first injection variable that meets predetermined conditions among the plurality of first injection variables. The predetermined conditions may mean that the amount of the second Lactate included in the predicted second influencing variable is less than or equal to the first critical amount and the second VCD value included in the predicted second quality attribute has a maximum value, or the amount of the second Lactate included in the predicted second influencing variable is less than or equal to the first critical amount and the amount of the second IgG included in the predicted second quality attribute has a maximum value. The selected first injection variable can be the changed first injection variable. Below, the operation of the apparatus 100 will be described in more detail.

[0182] Referring to FIG. 10, there may be a case where the user wants to limit the amount of Lactate to a first critical amount or less while satisfying the conditions of FIG. 8 or FIG. 9. In this case, the objective function can be shown as in FIG. 10. The first critical amount can be 1.2 g / L. When the objective function is set in this way, the apparatus 100 can derive an optimal result at a level where the amount of Lactate included in the predicted influencing variables does not exceed the first critical amount. More specifically, when the predicted amount of Lactate exceeds the first critical amount, the apparatus 100 can set the predicted second VCD value or the amount of the second IgG corresponding to the predicted amount of Lactate to 0. Therefore, when the predicted amount of Lactate exceeds the first critical amount, it is possible to prevent the predicted second VCD value or the amount of the second IgG from being selected as the maximum value.

[0183] In FIG. 10, when the amount of Lactate included in the predicted influencing variables exceeds the first critical amount, the apparatus 100 discloses a configuration in which the predicted second VCD value or the amount of the second IgG is set to 0, but is not limited thereto. When the amount of Lactate included in the predicted influencing variables exceeds the first critical amount, the apparatus 100 can determine the predicted second VCD value or the amount of the second IgG as a value obtained by multiplying the predicted second VCD value or the amount of the second IgG by a positive number less than or equal to 0.01. Therefore, it is possible to prevent the predicted second VCD value or the amount of the second IgG from being selected as the maximum value.

[0184] More specifically, the apparatus 100 can remove the injection variable in which the amount of Lactate exceeds the first critical amount for the first day to the third day of the culture. Alternatively, the apparatus 100 may not control the injection variable for the first day to the third day of the culture. Also, measurements may not be performed on the influencing variables or quality attributes. That is, the apparatus 100 can perform control from the first period.

[0185] Lactate predictedIt can correspond to a plurality of amounts of the second Lactate. The apparatus 100 can remove the second quality attribute, the second influencing variable, and the first injection variable corresponding to the amount of the second Lactate exceeding the first critical amount. Further, the apparatus 100 can determine the optimal first injection variable using some of the second quality attributes, some of the second influencing variables, and some of the first injection variables corresponding to the amount of the second Lactate equal to or less than the first critical amount.

[0186] More specifically, when performing the step 340 of selecting the second influencing variable and the second quality attribute, the apparatus 100 can further perform the following process. The apparatus 100 can perform a step of obtaining some of the second influencing variables among the plurality of second influencing variables, where the amounts of the plurality of second Lactates are equal to or less than the first critical amount. Further, the apparatus 100 can perform a step of obtaining some of the second quality attributes corresponding to some of the second influencing variables. Further, when in the first period, the apparatus 100 can perform a step of selecting the maximum value among the plurality of second VCD values included in some of the second quality attributes as the second quality attribute. Further, when in the second period, the apparatus 100 can perform a step of selecting the maximum value among the plurality of amounts of the second IgG included in some of the second quality attributes as the second quality attribute.

[0187] The apparatus 100 can perform a step 350 of selecting the first injection variable corresponding to the selected second quality attribute among the plurality of first injection variables. The apparatus 100 can perform a step of determining the first injection variable selected in step 350 as the optimal injection variable at the first time point. Further, the apparatus 100 can determine the selected first injection variable as the optimal injection variable and transmit it to the bioreactor 1810. Further, the bioreactor 1810 can determine the culture conditions for the first time point based on the received optimal injection variable. Through such a process, the bioreactor 1810 can produce the maximum amount of antibody.

[0188] Referring to FIG. 11, the user may sometimes want to limit the amount of Glutamate to be not less than the second critical amount and not more than the third critical amount while satisfying one of the conditions in FIGS. 8 to 10. The apparatus 100 can perform a step of determining whether the amount of the second Glutamate included in the second influence variable is not less than the second critical amount and not more than the third critical amount. When the amount of the second Glutamate is not less than the second critical amount and not more than the third critical amount and within the first period, the apparatus 100 can perform a step of changing the first injection variable so that the second VCD value included in the second quality attribute has a maximum value. When the amount of the second Glutamate is not less than the second critical amount and not more than the third critical amount and within the second period, the apparatus 100 can perform a step of changing the first injection variable so that the amount of the second IgG included in the second quality attribute has a maximum value. The apparatus 100 can perform tests on a plurality of first injection variables in order to change the first injection variable. Also, performing a test may mean obtaining the predicted second influence variable and the predicted second quality attribute for each of the plurality of first injection variables. The apparatus 100 can select a first injection variable that meets a predetermined condition among the plurality of first injection variables. The predetermined condition may mean that the amount of the second Glutamate included in the predicted second influence information is not less than the second critical amount and not more than the third critical amount, and the second VCD value included in the predicted second quality attribute has a maximum value, or the amount of the second Glutamate included in the predicted second influence variable is not less than the second critical amount and not more than the third critical amount, and the amount of the second IgG included in the second quality attribute has a maximum value. The selected first injection variable may be the changed first injection variable. The changed first injection variable may be the optimal first injection variable. The operation of the apparatus 100 will be described in more detail below.

[0189] Referring to Fig. 11, there may be a case where the user wants to limit the amount of Glutamate to be not less than the second critical amount and not more than the third critical amount while satisfying one of the conditions in Figs. 8 to 10. For example, the second critical amount may be 0.45 g / L. Also, the third critical amount may be 0.65 g / L. In this case, the objective function can be shown as in Fig. 11. When the objective function is set in this way, the apparatus 100 can derive an optimal result at a level where the amount of Glutamate is less than the second critical amount or does not exceed the third critical amount. More specifically, when the predicted amount of Glutamate is less than the second critical amount and exceeds the third critical amount, the apparatus 100 can set the predicted second VCD value or the amount of the second IgG corresponding to the predicted amount of Glutamate to 0. Therefore, when the predicted amount of Glutamate is less than the second critical amount and exceeds the third critical amount, it is possible to prevent the predicted second VCD value or the amount of the second IgG from being selected as the maximum value.

[0190] In Fig. 11, when the amount of Glutamate included in the predicted influence variable is less than the second critical amount and exceeds the third critical amount, the apparatus 100 is disclosed to have a configuration of setting the predicted second VCD value or the amount of the second IgG to 0, but it is not limited thereto. When the amount of Glutamate included in the predicted influence variable is less than the second critical amount and exceeds the third critical amount, the apparatus 100 can determine the predicted second VCD value or the predicted amount of the second IgG as a value obtained by multiplying the predicted second VCD value or the predicted amount of the second IgG by a positive number less than or equal to 0.01. Therefore, it is possible to prevent the predicted second VCD value or the amount of the second IgG from being selected as the maximum value.

[0191] More specifically, the apparatus 100 can remove injection variables where the amount of Glutamate is less than the second critical amount or more than the third critical amount for the first to third days of culture. Or, the apparatus 100 may not control the injection variables for the first to third days of culture. Also, measurements may not be performed on the influencing variables or quality attributes. That is, the apparatus 100 can perform control from the first period.

[0192] Glutamate predicted can correspond to a plurality of second amounts of Glutamate. The apparatus 100 can remove the second quality attributes, second influencing variables, and first injection variables corresponding to the second amount of Glutamate that is less than the second critical amount or more than the third critical amount. Also, the apparatus 100 can determine the optimal first injection variable using some of the second quality attributes, some of the second influencing variables, and some of the first injection variables corresponding to the second amount of Glutamate that is between the second critical amount and the third critical amount.

[0193] More specifically, when performing step 340 of selecting the second influencing variables and the second quality attributes, the apparatus 100 can further perform the following process. The apparatus 100 can perform the step of obtaining some of the second influencing variables among the plurality of second influencing variables where the plurality of second Glutamate concentrations are between the second critical amount and the third critical amount. Also, the apparatus 100 can perform the step of obtaining some of the second quality attributes corresponding to some of the second influencing variables. Also, when within the first period, the apparatus 100 can perform the step of selecting the maximum value among the plurality of second VCD values included in some of the second quality attributes as the second quality attribute. Also, when within the second period, the apparatus 100 can perform the step of selecting the maximum value among the plurality of second amounts of IgG included in some of the second quality attributes as the second quality attribute.

[0194] The device 100 can perform step 350 of selecting a first injection variable corresponding to the selected second quality attribute among a plurality of first injection variables. The device 100 can perform the step of determining the first injection variable selected in step 350 as the optimal injection variable at the first time point. Also, the device 100 can determine the selected first injection variable as the optimal injection variable and transmit it to the bioreactor 1810. Further, the bioreactor 1810 can determine the culture conditions for the first time point based on the received optimal injection variable. Through such a process, the bioreactor 1810 can produce the maximum amount of antibody.

[0195] The device 100 can select the optimal first injection variable for the second time point in step 350 of FIG. 3. Also, the device 100 can predict a third influencing variable and a third quality attribute based on the second influencing variable and the second quality attribute at the second time point. The device 100 can predict the third influencing variable and the third quality attribute using the second influencing variable and the second quality attribute at the second time point predicted in FIG. 3. The third influencing variable and the third quality attribute can be the predicted information at the third time point. The third time point can be in the future from the second time point. The interval between the first time point and the second time point may be the same as the interval between the second time point and the third time point.

[0196] However, it is not limited to predicting the third influencing variable and the third quality attribute using the second influencing variable and the second quality attribute at the second time point predicted in FIG. 3. The device 100 can measure the actual second influencing variable and the actual second quality attribute at the second time point and predict the third influencing variable and the third quality attribute based on the actual second influencing variable and the actual second quality attribute. This is because actual measurement is possible when the second time point is the current time point.

[0197] More specifically, the device 100 can perform the step of obtaining a plurality of second injection variables included within a predetermined range from the selected first injection variable. The step of obtaining a plurality of second injection variables can correspond identically to the step 320 of obtaining a plurality of first injection variables.

[0198] The device 100 can perform the step of applying the selected second quality attribute, the selected second influencing variable, and a plurality of second injection variables to a culture simulation model to obtain a plurality of third quality attributes and a plurality of third influencing variables. The step of obtaining the plurality of third quality attributes and the plurality of third influencing variables can correspond identically to step 330. The selected second quality attribute and the selected second influencing variable can be the second quality attribute or the second influencing variable predicted and selected in FIG. 3. Also, the selected second quality attribute and the selected second influencing variable can be the second quality attribute or the second influencing variable actually measured at the second time point.

[0199] The device 100 can perform the step of selecting third quality attributes and third influencing variables that meet predetermined conditions among the plurality of third quality attributes and the plurality of third influencing variables. The step of selecting the third quality attributes and the third influencing variables can correspond identically to step 340.

[0200] The device 100 can perform the step of selecting a second injection variable corresponding to the selected third quality attribute among the plurality of second injection variables. The step of selecting the second injection variable can correspond identically to step 350 in FIG. 3. The device 100 can repeat such a process to obtain further predicted influencing variables and quality attributes in the future. More specifically, the device 100 can obtain a predicted fourth quality attribute and a predicted fourth influencing variable based on the selected third quality attribute, the selected third influencing variable, and the third injection variable. As described above, the time between the first time point and the second time point and the time between the second time point and the third time point can be 10 minutes, 1 hour, or 1 day. The device 100 can make predictions at intervals of such a unit time to predict further future quality attributes or influencing variables. For example, the device 100 can predict the quality attribute or the influencing variable 4 days later or 1 week later.

[0201] FIG. 12 is a table showing simulation conditions according to an embodiment of the present invention. Also, FIGS. 13 to 17 can be simulation results according to an embodiment of the present invention.

[0202] Six samples of the embodiments of FIGS. 13 to 17 are all the results of simulation tests having the control targets shown in FIG. 12 from day 0 to day 13 for virtual CHO cell culture and antibody production experiments.

[0203] Glucose after feeding (g / L) in the embodiments of FIGS. 13 to 17 means the Glucose concentration (g / L) measured immediately after Glucose is added to the culture medium. That is, the apparatus 100 can inject Glucose into the current culture medium based on the amount of Glucose included in the injection variable selected in step 350, and Glucose after feeding (g / L) can be the amount of Glucose in the culture medium after injecting Glucose into the current culture medium based on the amount of Glucose included in the selected injection variable.

[0204] Agitator Speed (RPM) in the embodiments of FIGS. 13 to 17 means how many times per minute the agitator in the Bioreactor rotates.

[0205] Vessel temp (° C) in the embodiments of FIGS. 13 to 17 means the temperature inside the Bioreactor.

[0206] Feed X in the embodiments of FIGS. 13 to 17 means how much of a certain nutrient medium (Feed) corresponding to a nitrogen source or the like in CHO cells is added to the culture solution in the Bioreactor in the previous time unit.

[0207] For the Lactate added in yesterday (g / L) in the embodiments of FIGS. 13 to 17, it means how many grams of Lactate were added to 1 L of the culture solution per previous time unit. When adding Lactate, it is added after neutralizing the pH in advance using a basic substance. When 1 g of Lactate neutralized with a basic substance in terms of pH is added per 1 L of the culture solution, the increase in osmolarity was arbitrarily set to 22 mOsmol / kg on the culture simulation system.

[0208] For the Dissolved Oxygen (%) in the embodiments of FIGS. 13 to 17, it means the percentage of the dissolved oxygen amount in the culture solution in the Bioreactor.

[0209] FIG. 13 is a graph showing the simulation test results of each of Sample 1 and Sample 2.

[0210] The injection variable-related graph 1310 in FIG. 13 shows the injection variables selected over time among a plurality of injection variables. In graph 1310, the first time point can be the time point immediately before the second time point. For example, in graph 1310, if the first day is the first time point, the second day can be the second time point; if the second day is the first time point, the third day can be the second time point.

[0211] The influence variable-related graph 1320 in FIG. 13 can show the influence variables selected over time among a plurality of influence variables. Graph 1320 can be related to the predicted second influence variable by the selected first injection variable. For example, if the amount of the first Glucose contained in the first injection variable selected on the fourth day is 10 g / L (refer to reference number 1311), the amount of the second Glucose contained in the second influence variable predicted on the fifth day may be slightly less than about 9 g / L (refer to reference number 1321).

[0212] The quality attribute related graph 1330 in FIG. 13 can show the quality attributes selected over time among a plurality of quality attributes. The graph 1330 can be related to the predicted second quality attribute based on the selected first injection variable. For example, if the amount of the first Glucose included in the first injection variable selected on the 4th day is 10 g / L (see reference number 1311), the amount of the second IgG included in the second quality attribute predicted on the 5th day may be slightly less than about 1 g / L (see reference number 1331).

[0213] Sample 1 is the result of simulation testing after fixing other injection variables (excluding Feed X) at a constant value in actual CHO cell culture experiment data instead of simulation. This is almost the same as the culture data of the actual experiment and serves as a control group in this embodiment.

[0214] Sample 2 is the result of culturing with the combination of injection variable values set daily by a simulation-based control system such that only the VCD value of each day is maximized until the 8th day (the first period) after the 4th day of culture, and only the IgG concentration of each day is maximized from the 8th day of culture (the second period). That is, the conditions may be the same as those in FIG. 8. Note that Sample 2 may be the result of setting the combination of injection variable values by maximizing only the objective function at the second time point (future time point), which is the next time point after the first time point (for example, the current time point).

[0215] The simulation-based control system adjusted the injection variables daily during the culture days to maximize the objective function, and as a result, calculated a final IgG concentration higher than that of Sample 1, which was the control group. Specifically, in Sample 2, less Feed X was introduced overall from Sample 1, which was the control group, and the Osmolarity, Glutamine concentration, and Glutamate concentration of the culture broth were generally maintained at values lower than those of the control group. From the 7th day of culture, the pH, Agitator Speed (RPM), and Vessel temp (°C) were intentionally set lower than those of the control group. However, the simulation-based control system of the present invention showed a specific control tendency to intentionally reduce the Dissolved Oxygen (%) from the 7th day of culture to 20% and amplify the Lactate concentration to about 3.4 g / L. Since a high Lactate concentration tends to suppress the increase in Ammonia concentration, the simulation-based control system of the present invention calculated that even if the Dissolved Oxygen (%) was set to a low value, suppressing the generation of Ammonia, which has a negative impact on CHO cells due to the Lactate generated through this, was more beneficial from the perspective of the objective function from the 7th day of culture than when the Dissolved Oxygen (%) was set to a high value.

[0216] However, since the prediction model used in this embodiment has never learned similar culture data, it is uncertain whether this control strategy will produce positive results like those of Sample 2 in actual culture rather than in simulation. Also, in Sample 2, less Feed X was introduced than in Sample 1, but it is not known exactly what results will be brought about by reducing the Feed X input amount in actual culture. Nevertheless, if such an untested strategy is applied to actual culture and the result is more negative than that of Sample 2 in Fig. 13, when the culture simulation system is further trained with the data of this negative result, the subsequent simulation-based control system can also be familiar with this result and perform subsequent control, and as a result, improve its own control performance and accuracy.

[0217] FIG. 14 may include an injection variable-related graph, an influence variable-related graph, and a quality attribute-related graph, which are the same as those in FIG. 13.

[0218] FIG. 14 is a graph representing the simulation test results of Sample 1, Sample 2, and Sample 3 respectively. Sample 1 and Sample 2 are the same as the above description. Sample 3 is basically the same as the control target of Sample 2, but the objective function is set to minimize the amount of Ammonia while maintaining the amount of Lactate at 1.2 g / L or less. The objective function of Sample 3 can be shown as in FIG. 10. The objective function of Sample 3 specifies that when the predicted Lactate concentration in the next time unit exceeds 1.2 g / L, the objective function value is fixed at 0. However, if it is impossible to maintain the Lactate concentration at 1.2 g / L or less on the culture simulation system, all the objective function values that need to be compared with each other will become 0, making it impossible to derive a substantial optimal result. Therefore, to prevent this, when the Lactate concentration exceeds 1.2 g / L, alternatives such as using a value obtained by multiplying a small positive number of about 0.01 or less by the predicted VCD value or the predicted IgG amount as the objective function instead of fixing the objective function at 0 can be considered. From the results of Sample 3, by intentionally injecting Lactate into the culture solution by the apparatus 100 of the present invention, the Dissolved Oxygen (%) is set higher than that of Sample 2. Even if the amount of IgG on the final day of culture is lower than that of Sample 2, the concentration of Lactate is maintained within the range of 1.2 g / L or less throughout the entire culture period, and at the same time, the Ammonia concentration is also maintained lower than that of Sample 2 from the 5th to the 11th day of culture.

[0219] FIG. 15 may include an injection variable-related graph, an influence variable-related graph, and a quality attribute-related graph, which are the same as those in FIG. 13.

[0220] Figure 15 is a graph showing the simulation test results of Sample 1, Sample 3, and Sample 4 respectively. Sample 1 and Sample 3 are the same as the above description. Sample 4 is basically the same as the control target of Sample 3, but the objective function is set to control while maintaining the Glutamate concentration in the range of 0.45 g / L or more and 0.65 g / L or less on the 4th day or later of cultivation. The objective function of Figure 15 can be shown as in Figure 11.

[0221] The apparatus 100 of the present invention achieved this goal within the simulation by adjusting the Feed X input amount and the like. Also, the IgG concentration on the final day of cultivation was calculated to be higher for Sample 4 than for Sample 3. This can show that even if further constraints are added to the objective function of the control system of the present invention, it can produce more optimal results from a long-term perspective.

[0222] Figure 16 may include an injection variable-related graph, an influence variable-related graph, and a quality attribute-related graph, which are the same as those in Figure 13.

[0223] Figure 16 is a graph showing the simulation test results of Sample 1 and Sample 5 respectively. Sample 1 is the same as the above description. Sample 5 is basically the same as the control target of Sample 2, but it is the result of controlling while keeping other injection variables, except for the Feed X input amount and Lactate added in yesterday (g / L) among the injection variables during cultivation, constant in the same manner as Sample 1. The said Sample 5 showed a positive result with a higher IgG concentration on the last day of cultivation than Sample 1. Sample 5 maintained the Lactate concentration at around 1.5 g / L high in the middle of cultivation, suppressed the increase in Ammonia concentration, also fed a relatively small amount of Feed X as a whole, and maintained the Osmolarity, Glutamine concentration, and Glutamate concentration at values lower than those of Sample 1, which is the control group. In this way, the user or the device 100 can obtain optimal quality attributes by selectively modifying at least one of the amount of the first nutrient medium (feed) included in the first injection variable, the amount of Glucose in the first nutrient medium, the amount of the first oxygen molecule, the amount of the first carbon dioxide molecule, the amount of the first nitrogen molecule, the amount of the first air, the first pH value, the first Agitator Speed, the first Vessel temperature, the first Media injection schedule, and the amount of the first Antifoam.

[0224] Figure 17 may include an injection variable-related graph, an influence variable-related graph, and a quality attribute-related graph, which are the same as those in Figure 13.

[0225] Figure 17 is a graph showing the simulation test results of Sample 1, Sample 5, and Sample 6 respectively. Sample 1 and Sample 5 are the same as the above description. Sample 6 is basically the same as the control target of Sample 5, but the objective function is set to control while maintaining the Lactate concentration at 1.2 g / L or less, minimizing the Ammonia concentration, and maintaining the Glutamate concentration in the range of 0.45 g / L or more and 0.65 g / L or less on the 4th day or more of the culture. The Lactate concentration and the Glutamate concentration achieved the control target, but the Ammonia concentration was lower than that of Sample 1 but higher than that of Sample 5, and the IgG concentration on the last day of the culture was also higher than that of Sample 1 but lower than that of Sample 5. Nevertheless, Figure 17 is significant in that it is theoretically possible for the simulation-based control system of the present invention to maximize or minimize the influence variables and quality variables at the user's discretion, or maintain them within a desired range, only by adjusting the injection variables.

[0226] The apparatus 100 of the present invention predicts at least one of the influence variables and quality attributes at the second time point based on the injection variables, influence variables, and quality attributes at the first time point. Here, the first time point can be the time immediately before the second time point. Therefore, if the apparatus 100 pursues short-term results, it may have an adverse effect from a long-term perspective. Therefore, the apparatus 100 can be determined by studying the "optimal range" of the injection variables or influence variables that can derive long-term optimal quality attributes through various experiments or simulations. Once the optimal range of the injection variables or influence variables is determined, it is quite possible to reset the objective function and control while adjusting the optimal range using the simulation-based control system. In addition, the apparatus 100 can learn an artificial intelligence reinforcement learning model such as a Deep Q Learning Network and actively determine how to set the range of the values of the influence variables for each future time point. Through this, the apparatus 100 can determine the injection variables for obtaining optimal quality attributes from a long-term perspective.

[0227] The above has been described mainly with respect to various embodiments. A person having ordinary skill in the technical field to which the present invention pertains can understand that the present invention can be embodied in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered from an illustrative perspective rather than a limiting perspective. The scope of the present invention is shown not in the above description but in the claims, and all differences within the scope equivalent thereto should be construed as being included in the present invention.

[0228] In addition, the embodiments of the present invention as described above can be created by a program executable by a computer, and may be embodied in a general-purpose digital computer that operates the program using a computer-readable recording medium. The computer-readable recording medium includes storage media such as magnetic storage media (e.g., ROM, floppy (registered trademark) disk, hard disk, etc.) and optical reading media (e.g., CD-ROM, DVD, etc.).

Claims

1. A method for operating an apparatus for determining cell culture conditions, comprising: obtaining a first quality attribute and a first influencing variable at a first time point from a cell culture medium; obtaining a first injection variable; applying at least one of the first quality attribute, the first influencing variable, and the first injection variable to a culture simulation model to obtain a second quality attribute and a second influencing variable at a second time point; selectively varying a combination and amount of the first injection variable to be applied to the culture simulation model next time so that the second quality attribute meets predetermined conditions; wherein the second time point is in the future of the first time point; wherein the first quality attribute includes a first viable cell density (VCD) value and a first Immunoglobulin G (IgG) amount; wherein the second quality attribute includes a second viable cell density (VCD) value and a second Immunoglobulin G (IgG) amount; wherein the first influencing variable includes at least one of an amount of first Glutamine, an amount of first Glutamate, an amount of first Glucose, an amount of first Lactate, an amount of first Ammonia, and a first Osmolarity; wherein the second influencing variable includes at least one of an amount of second Glutamine, an amount of second Glutamate, an amount of second Glucose, an amount of second Lactate, an amount of second Ammonia, and a second Osmolarity; wherein the first injection variable includes at least one of an amount of first feed, an amount of Glucose in the first feed, an amount of first oxygen molecules, an amount of first carbon dioxide molecules, an amount of first nitrogen molecules, an amount of first air, a first pH value, a first Agitator Speed, a first Vessel temperature, a first Media injection schedule, and an amount of first Antifoam; selectively varying the combination and amount of the first injection variable comprises: selectively varying the combination and amount of the first injection variable in a first period and in a second period; wherein the second period is after the first period; Selectively changing the combination and amount of the first injection variable means that in the case within the first period, changing the combination and amount of the first injection variable so that the second VCD value included in the second quality attribute has a maximum value, in the case within the second period, changing the combination and amount of the first injection variable so that the amount of the second IgG included in the second quality attribute has a maximum value, A method for operating a cell culture condition determination device.

2. The step of obtaining the first quality attribute and the first influence variable at the first time point from the cell culture solution includes the step of obtaining a plurality of first influence variables measured at the same time point, The step of obtaining the second quality attribute and the second influence variable at the second time point includes the step of obtaining a second influence variable predicted based on the plurality of measured first influence variables. The method for operating a cell culture condition determination device according to claim 1.

3. Selectively changing the combination and amount of the first injection variable means that in the case within the first period, changing the combination and amount of the first injection variable so that the value obtained by multiplying the second VCD value included in the second quality attribute by e^(-k * the amount of the second Ammonia) has a maximum value, in the case within the second period, changing the combination and amount of the first injection variable so that the value obtained by multiplying the amount of the second IgG included in the second quality attribute by e^(-k * the amount of the second Ammonia) has a maximum value, where k is a positive number. The method for operating a cell culture condition determination device according to claim 1.

4. Selectively changing the combination and amount of the first injection variable means that determining whether the amount of the second Lactate included in the second influence variable is less than or equal to a first critical amount, when the amount of the second Lactate is less than or equal to the first critical amount and within the first period, changing the combination and amount of the first injection variable so that the second VCD value included in the second quality attribute has a maximum value, when the amount of the second Lactate is less than or equal to the first critical amount and within the second period, changing the first injection variable so that the amount of the second IgG included in the second quality attribute has a maximum value, The method for operating a cell culture condition determination device according to claim 1.

5. Selectively changing the combination and amount of the first injection variable means that determining whether the amount of the second Glutamate included in the second influence variable is greater than or equal to a second critical amount and less than or equal to a third critical amount, When the amount of the second Glutamate is equal to or greater than the second critical amount and equal to or less than the third critical amount, and within the first period, the combination and amount of the first injection variables are changed such that the second VCD value included in the second quality attribute has a maximum value. When the amount of the second Glutamate is equal to or greater than the second critical amount and equal to or less than the third critical amount, and within the second period, the combination and amount of the first injection variables are changed such that the amount of the second IgG included in the second quality attribute has a maximum value. The operation method of the cell culture condition determination device according to claim 1.

6. The operation method of the cell culture condition determination device according to claim 1, including the step of transmitting the changed combination and amount of the first injection variables to the bioreactor.

7. An apparatus for determining cell culture conditions, including a processor and a memory, wherein the processor, based on the instruction words stored in the memory, acquires a first quality attribute and a first influencing variable at a first time point from a cell culture medium, acquires a first injection variable, applies at least one of the first quality attribute, the first influencing variable, and the first injection variable to a culture simulation model, acquires a second quality attribute and a second influencing variable at a second time point, and selectively changes the combination and amount of the first injection variables to be applied to the culture simulation model next time so that the second quality attribute meets predetermined conditions. The second time point is in the future of the first time point. The first quality attribute includes a first VCD (viable cell density) value and a first IgG (Immunoglobulin G) amount. The second quality attribute includes a second VCD (viable cell density) value and a second IgG (Immunoglobulin G) amount. The first influencing variable includes at least one of the amount of the first Glutamine, the amount of the first Glutamate, the amount of the first Glucose, the amount of the first Lactate, the amount of the first Ammonia, and the first Osmolarity. The second influencing variable includes at least one of the amount of second glutamine, the amount of second glutamate, the amount of second glucose, the amount of second lactate, the amount of second ammonia, and the second osmolarity. The first injection variable includes at least one of the amount of the first feed, the amount of glucose in the first feed, the amount of first oxygen molecules, the amount of first carbon dioxide molecules, the amount of first nitrogen molecules, the amount of first air, the first pH value, the first Agitator Speed, the first Vessel temperature, the first Media injection schedule, and the amount of first Antifoam. Selectively changing the combination and amount of the first injection variable Selectively change the combination and amount of the first injection variable between the case within the first period and the case within the second period. The second period is after the first period. Selectively changing the combination and amount of the first injection variable In the case within the first period, change the combination and amount of the first injection variable so that the second VCD value included in the second quality attribute has a maximum value. In the case within the second period, change the combination and amount of the first injection variable so that the amount of the second IgG included in the second quality attribute has a maximum value. Cell culture condition determination device.

Images