Cell culture multi-parameter collaborative AI dynamic optimization system

Abstract

The application discloses a cell culture multi-parameter collaborative AI dynamic optimization system, and relates to the technical field of cell culture. The system comprises a collector configured to acquire growth cycle data of multiple groups of historical cell cultures; and a processor configured to acquire a data feature group corresponding to each group of cells based on physiological parameters of each group of historical cell cultures. The application breaks through the limitation of traditional static optimization methods that cannot adapt to the dynamic time-varying process of cells by precisely anchoring cell metabolic stress, co-stimulation coupling and functional balance state. At the same time, based on eigenvalue multidimensional clustering, the law of high-quality cell populations is mined, and the particle swarm algorithm is improved to obtain a fitness value capable of representing the pros and cons of the cell culture state, realize dynamic coupling analysis of culture parameters and cell metabolism and phenotype state, and overcome the regulation lag problem caused by the fact that traditional optimization methods cannot analyze the dynamic coupling of metabolism and culture parameters and are difficult to track the cell demand within the window.

Description

Technical Field

[0001] This application relates to the field of cell culture technology, specifically a multi-parameter collaborative AI dynamic optimization system for cell culture. Background Technology

[0002] The in vitro expansion of immune cells is a typical dynamic, time-varying process. As culture progresses, cells undergo a latent phase, a logarithmic growth phase, a plateau phase, and even a death phase. Throughout these phases, the metabolic patterns, nutritional requirements, and responses to external stimuli of immune cells constantly change. This means that optimal parameter combinations in the early stages of culture may become limiting or even harmful conditions for immune cells in the later stages. Therefore, an ideal control system must possess dynamic optimization capabilities, continuously adjusting strategies to adapt to the ever-changing state of immune cells.

[0003] Traditional optimization methods (e.g., static culture medium adjustment methods) have limited effectiveness in dynamic environments, while AI-driven dynamic optimization can adapt to changes through continuous learning. For example, evolutionary algorithms or particle swarm optimization algorithms can be used to handle time-varying problems. Particle swarm optimization models can effectively track multiple optimization peaks in dynamic environments, such as the optimal point of immune cells at different growth stages. However, they cannot analyze the dynamic coupling relationship between cell metabolic state and culture parameters in real time, leading to the loss of critical regulatory windows. Summary of the Invention

[0004] The purpose of this application is to provide a multi-parameter collaborative AI dynamic optimization system for cell culture, in order to solve the technical problem that existing technologies cannot analyze the dynamic coupling relationship between cell metabolic state and culture parameters in real time, which leads to the easy loss of key regulatory windows.

[0005] To achieve the above objectives, this application provides the following technical solution:

[0006] A multi-parameter collaborative AI dynamic optimization system for cell culture includes:

[0007] A data acquisition device is used to acquire growth cycle data from multiple historical cell cultures; the growth cycle includes at least a latent phase, a growth phase, and a plateau phase; the growth cycle data includes at least the first culture parameter and physiological parameter for each historical cell culture corresponding to the growth cycle.

[0008] The processor is used to acquire a data feature set corresponding to each group of cells based on the physiological parameters of each group of historical cell cultures; the data feature set includes feature values ​​corresponding to each growth cycle; the feature values ​​are used to characterize the quality of historical cell cultures at least for the corresponding growth cycle.

[0009] Furthermore, based on each data feature group, the range of high-quality physiological parameters corresponding to each growth cycle is obtained;

[0010] Furthermore, based on each data feature group, multiple correlation coefficients are obtained; the correlation coefficients are the Pearson correlation coefficients between the first culture parameter and the feature value corresponding to each growth cycle.

[0011] Furthermore, based on each Pearson correlation coefficient, multiple second culture parameters are obtained; the second culture parameters are the culture parameters whose corresponding Pearson correlation coefficients are greater than or equal to a first preset value among the various first culture parameters.

[0012] Additionally, each second culture parameter of each group of historical cell cultures is treated as a particle to obtain the current particle swarm;

[0013] Furthermore, based on the range of superior physiological parameters, the current fitness value of the current particle swarm is obtained; the current fitness value is used at least to characterize the probability that the physiological parameters corresponding to each particle in the current particle swarm fall within the range of superior physiological parameters.

[0014] Furthermore, the current particle swarm is iteratively updated based on the particle swarm algorithm so that the updated current fitness value is greater than or equal to the second preset value, thereby obtaining the globally optimal parameter combination for cell culture;

[0015] The controller is used to dynamically adjust the second culture parameter of the current cell culture growth cycle based on the globally optimal parameter combination.

[0016] As a specific embodiment of the technical solution in this application, the collector includes a lactate analyzer, a dissolved oxygen electrode, and a flow cytometer; the physiological parameters of the latency period include lactate concentration, basal respiratory rate, and fluorescence intensity of CD3 molecules;

[0017] The lactate analyzer is used to obtain the lactate concentration of each group of historical cell cultures;

[0018] The dissolved oxygen electrode is used to obtain the baseline respiratory rate of each group of historical cell cultures; the baseline respiratory rate is the dissolved oxygen consumption per unit cell per hour.

[0019] The flow cytometer is used to obtain the fluorescence intensity of CD3 molecules from each group of historical cell cultures.

[0020] As a specific solution in this application, the processor is further configured to obtain the clustering distribution entropy based on the fluorescence intensity of the CD3 molecule;

[0021] Furthermore, based on the baseline respiration rate, a first time interval for each group of historical cell cultures is obtained; the start time of the first time interval is the start time of the historical cell culture, and the end time is any time point where the baseline respiration rate is still increasing and the difference between the first cluster distribution entropy and the second cluster distribution entropy is less than a third preset value; the first cluster distribution entropy is the cluster distribution entropy corresponding to the start time of the first time interval; the second cluster distribution entropy is any cluster distribution entropy whose time sequence is after the first cluster distribution entropy;

[0022] And, based on the first time interval, a first value and a second value are obtained; the first value is the average value of the increase in lactic acid concentration per unit time within the first time interval; the second value is the average rate at which the respiratory entropy of cells recovers from the initial value to the fourth preset value within the first time interval.

[0023] Furthermore, based on the first clustering distribution entropy, the second clustering distribution entropy, the first value, and the second value, a first feature value for each group of historical cell cultures is obtained; the first feature value is the feature value corresponding to the incubation period of each group of historical cell cultures.

[0024] As a specific embodiment of the technical solution in this application, the collector further includes a fluorescence cell counter, a high-performance liquid chromatograph, and a PCR amplification instrument; the physiological parameters of the growth phase include the cell division index, the rate of change in the concentration of nucleotide synthesis precursors, and the CD28 transcription level;

[0025] The fluorescent cell counter is used to obtain the cell division index of each group of historical cell cultures;

[0026] The high-performance liquid chromatograph is used to obtain the rate of change in the concentration of nucleotide synthesis precursors in each group of historical cell cultures.

[0027] The PCR amplification instrument is used to obtain the CD28 transcription level of each group of historical cell cultures.

[0028] As a specific solution in this application, the processor is further configured to obtain a second time interval based on the cell division index, the concentration change rate, and the CD28 transcription level; the start time of the second time interval is the end time of the first time interval, and the end time of the second time interval is the time point when the growth rates of the concentration change rate and the CD28 transcription level are both less than 0.5 times their highest growth rate, and the absolute value of the second difference of the cell division index is less than or equal to a fifth preset value.

[0029] Furthermore, based on the second time interval, a third, fourth, fifth, and sixth value are obtained; the third value is the average growth rate of CD28 transcription level within the second time interval; the fourth value is the CD3 expression level corresponding to the end time of the second time interval; the fifth value is the nucleotide consumption entropy within the second time interval; and the sixth value is the cell division index within the second time interval.

[0030] Furthermore, based on the third, fourth, fifth, and sixth values, a second characteristic value is obtained for each group of historical cell cultures; the second characteristic value is the characteristic value corresponding to the growth period of each group of historical cell cultures.

[0031] As a specific solution in this application, the collector further includes a confocal microscope; the physiological parameters of the plateau phase include IL-2 secretion concentration, IFN-γ secretion concentration, and the uniformity of CD62L localization and distribution on the cell membrane;

[0032] The flow cytometer is also used to obtain time-series curves of IL-2 secretion concentration and IFN-γ secretion concentration of each group of historical cell cultures.

[0033] The confocal microscope is used to obtain the uniformity of CD62L localization and distribution on the cell membrane of each group of historical cell cultures.

[0034] As a specific solution in this application, the processor is further configured to obtain a third time interval based on the uniformity of the positioning distribution; the start time of the third time interval is the end time of the second time interval, and the end time of the third time interval is the time point when the uniformity of the positioning distribution of temporally adjacent units meets the first preset condition.

[0035] Furthermore, based on the third time interval, a seventh, eighth, and ninth value are obtained; the seventh value is the uniformity of CD62L localization distribution corresponding to the end time of the third time interval; the eighth value is the average value of IL-2 secretion concentration within the third time interval; and the ninth value is the average slope of the time-series curve of IFN-γ secretion concentration within the third time interval.

[0036] Furthermore, based on the seventh value, the eighth value, and the ninth value, a third characteristic value is obtained for each group of historical cell cultures; the third characteristic value is the characteristic value corresponding to the plateau phase of each group of historical cell cultures.

[0037] As a specific solution in the technical solution of this application, the processor is further configured to project each data feature group into a multi-dimensional space to form multiple spatial coordinate points;

[0038] Furthermore, based on the density clustering algorithm, multiple clusters are obtained from each spatial coordinate point;

[0039] Furthermore, based on each cluster, the optimal cluster is obtained; the optimal cluster is any one cluster or a combination of multiple clusters that meet the second preset condition.

[0040] Furthermore, based on the optimal cluster, the range of high-quality physiological parameters is obtained; the lower limit of the range of high-quality physiological parameters is the minimum value in the optimal cluster, and the upper limit of the range of high-quality physiological parameters is the maximum value in the optimal cluster.

[0041] As a specific solution in the technical solution of this application, the processor is further configured to obtain the optimal probability corresponding to each particle based on the current particle swarm; the optimal probability is the probability that the physiological parameters corresponding to each growth cycle of the corresponding particle fall within the range of the high-quality physiological parameters.

[0042] Furthermore, based on each optimal probability, a probability fluctuation value and a probability average value are obtained; the probability fluctuation value is used at least to characterize the degree of difference between each optimal probability; the probability average value is the average of each optimal probability.

[0043] Furthermore, the current fitness value is obtained based on the probability fluctuation value and the probability average value.

[0044] As a specific solution in this application, the processor is further configured to obtain the current particle based on the current particle swarm; the current particle is any particle in the current particle swarm that currently needs to have its optimal probability calculated according to the calculation order;

[0045] Based on the current particle, multiple fall-in probabilities are obtained; the fall-in probabilities are at least used to characterize the probability that the culture parameters of the current particle are adjusted so that the physiological parameters corresponding to the current particle are more likely to fall into the range of high-quality physiological parameters.

[0046] Based on each fall probability, the optimal probability corresponding to the current particle is obtained.

[0047] Compared with the prior art, the beneficial effects of this application are:

[0048] This application overcomes the limitations of traditional static optimization methods in adapting to dynamic, time-varying cellular processes by precisely anchoring cellular metabolic stress, co-stimulatory coupling, and functional equilibrium. Simultaneously, it improves the particle swarm optimization algorithm by mining patterns in high-quality cell populations through eigenvalue multidimensional clustering to obtain fitness values ​​that characterize the quality of cell culture. This enables dynamic coupling analysis of culture parameters with cell metabolism and phenotypic status, overcoming the regulatory lag problems caused by the inability of traditional optimization methods to analyze the dynamic coupling between metabolism and culture parameters and to track cell demand within a window. Ultimately, it achieves synergistic dynamic optimization of culture parameters throughout the entire growth cycle, ensuring timely regulation of cells at each stage and improving the stability of terminal cell quality. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating a multi-parameter collaborative AI dynamic optimization method for cell culture proposed in an embodiment of this application.

[0050] Figure 2 This is a schematic diagram of the structure of a cell culture multi-parameter collaborative AI dynamic optimization system proposed in an embodiment of this application. Detailed Implementation

[0051] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0052] The terms "first," "second," etc., in the specification and accompanying drawings of the embodiments of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. For example, the first culture parameter and the second culture parameter mentioned below are different culture parameters. It should be understood that such names can be used interchangeably where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or modules is not necessarily limited to those steps or modules explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. The division of modules in the embodiments of this application is merely a logical division. In actual applications, there may be other division methods. For example, multiple modules may be combined into or integrated into another system, or some features may be ignored or not performed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interface, and the indirect coupling or communication connection between modules may be electrical or other similar forms. None of these are limited in the embodiments of this application. Furthermore, the modules or sub-modules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed among multiple circuit modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the embodiments of this application.

[0053] To address the technical problem mentioned in the background art—that existing technologies cannot analyze the dynamic coupling relationship between cell metabolic state and culture parameters in real time, leading to the easy loss of critical regulatory windows—this application proposes an embodiment of a multi-parameter collaborative AI dynamic optimization method for cell culture. Specifically, as shown in the example... Figure 1 As shown, the cell culture multi-parameter collaborative AI dynamic optimization method includes steps S100 to S900.

[0054] Step S100: Obtain growth cycle data from multiple historical cell cultures.

[0055] In this embodiment, the growth cycle includes at least a latent phase, a growth phase, and a plateau phase. The growth cycle data includes at least the first culture parameters and physiological parameters corresponding to the growth cycle of each group of historical cell cultures.

[0056] In this embodiment, the growth cycle is divided into three stages (i.e., the latent phase, the growth phase, and the plateau phase) merely to illustrate the multi-parameter collaborative AI dynamic optimization method for cell culture proposed in this application. This does not imply that the growth cycle can only be divided into three stages in this application. It should be understood that in other embodiments of this application, the growth cycle can be divided into any number of stages (e.g., two stages, four stages, or five stages, etc.), and further examples will not be provided hereafter.

[0057] In this embodiment, the first culture parameter refers to the culture conditions set during cell culture, such as temperature, pH, glucose concentration, dissolved oxygen, glutamine concentration, and ammonia ion concentration. In this embodiment, the physiological parameter refers to the cell culture state during cell culture, such as lactate concentration, CD3 molecule fluorescence intensity, and basal respiration rate (described below).

[0058] In the embodiments of this application, growth cycle data can be obtained in any reasonable manner. For example, growth cycle data can be read directly from a cell culture database (the database can be self-built or shared by the industry), or, as described in steps S201 to S210 below, culture parameters and physiological parameters can be tested and recorded in real time during cell culture to form growth cycle data.

[0059] Step S200: Based on the physiological parameters of each group of historical cell cultures, obtain data feature groups that correspond one-to-one with each group of cells.

[0060] In this embodiment, the data feature set includes feature values ​​corresponding to each growth cycle. These feature values ​​are used at least to characterize the quality of historical cell culture at the corresponding growth cycle.

[0061] In this embodiment, the characteristic value can be any numerical value that reflects the quality of the corresponding growth cycle. For example, since cells in the latent phase have a low respiration rate, the rate of increase in respiration rate can be used as a characteristic value of the latent phase. That is, if the rate of increase in respiration rate is small, the latent phase is better; if the rate of increase in respiration rate is large, the latent phase is worse. Since cells in the growth phase undergo significant division and require a large amount of nucleotide synthesis precursors, the rate of change in the concentration of nucleotide synthesis precursors can be used as a characteristic value of the growth phase. That is, if the rate of change in the concentration of nucleotide synthesis precursors is more uniform, the growth phase is better; if the rate of change in the concentration of nucleotide synthesis precursors is more uneven, the growth phase is worse. Since the uniformity of CD62L localization and distribution on the cell membrane gradually increases in cells in the plateau phase, the uniformity of CD62L localization and distribution can be used as a characteristic value of the plateau phase. That is, if the uniformity of CD62L localization and distribution is high, the plateau phase is better; if the uniformity of CD62L localization and distribution is low, the plateau phase is worse.

[0062] In order to accurately obtain the characteristic values ​​that can characterize the quality of each growth cycle, the characteristic values ​​corresponding to the latent period can be obtained as shown in steps S201 to S204; the characteristic values ​​corresponding to the growth period can be obtained as shown in steps S205 to S207; and the characteristic values ​​corresponding to the plateau period can be obtained as shown in steps S208 to S210.

[0063] Specifically, in this embodiment, the physiological parameters of the historical cell culture are pre-acquired by the collector.

[0064] In one embodiment of this application, the data acquisition device includes a lactate analyzer, a dissolved oxygen electrode, and a flow cytometer. The physiological parameters of the latency period include lactate concentration, basal respiration rate, and the fluorescence intensity of CD3 molecules. In this embodiment, the lactate analyzer is used to obtain the lactate concentration during the culture of each group of historical cells. The dissolved oxygen electrode is used to obtain the basal respiration rate during the culture of each group of historical cells. The basal respiration rate is the dissolved oxygen consumption per unit cell per hour. The flow cytometer is used to obtain the fluorescence intensity of CD3 molecules during the culture of each group of historical cells.

[0065] It is important to note that after cell seeding, membrane structure repair and energy metabolism restart will be initiated first, and mitochondrial function will begin to rebuild, with continuous metabolic energy supply supporting volume expansion. Specifically, the cell volume expansion rate will increase from 0 (i.e., the slope of the cell volume expansion rate curve is greater than 0), reflecting the restoration of cell membrane integrity. Simultaneously, the basal respiration rate will also increase from a low level (i.e., the slope of the basal respiration rate curve is greater than 0), marking the beginning of the latency period. Based on metabolic reconstruction, T cell receptor (CD3) begins to be expressed in an orderly manner. In the early stage of the latency period, the average fluorescence intensity of CD3 molecules is low and the cluster distribution entropy is high, indicating a disordered molecular distribution. As the membrane potential recovers, CD3 molecules gradually aggregate, and the corresponding cluster distribution entropy decreases. Respiration provides adenosine triphosphate (ATP) for protein synthesis, so the basal respiration rate of cells will continue to increase. When the rate of increase in cell volume expansion decreases, the cell respiration rate remains high, and the rate of decrease in the cluster distribution entropy corresponding to CD3 molecules slows down, the latency period ends. Based on this, the acquisition of the characteristic values ​​corresponding to the latency period includes steps S201 to S204.

[0066] Step S201: Obtain the cluster distribution entropy based on the fluorescence intensity of the CD3 molecule.

[0067] Specifically, the cluster distribution entropy is obtained as follows: The single-cell fluorescence image acquired by flow cytometry is divided into several 100×100 pixel grids. The ratio of the fluorescence intensity in each grid to the total fluorescence intensity is calculated. Based on these ratios, the information entropy is calculated, and the resulting information entropy is the cluster distribution entropy of the CD3 molecules. It should be noted that calculating information entropy for multiple values ​​is a mature technique and will not be elaborated upon here.

[0068] Step S202: Based on the baseline respiration rate, obtain the first time interval of each group of historical cell cultures.

[0069] In this embodiment, the start time of the first time interval is the start time of the historical cell culture, and the end time is any time point where the basal respiration rate is still increasing and the difference between the first cluster distribution entropy and the second cluster distribution entropy is less than a third preset value. The first cluster distribution entropy is the cluster distribution entropy corresponding to the start time of the first time interval; the second cluster distribution entropy is any cluster distribution entropy whose time sequence follows the first cluster distribution entropy.

[0070] In this embodiment, the start time of cell culture can be selected according to the needs. For example, the start time of cell culture can be 10 to 60 minutes after cell seeding, or 1 to 24 hours after cell seeding, etc.

[0071] In this embodiment, the third preset value can be set according to requirements. For example, the third preset value can be 0, or any value close to 0 (e.g., 0.1 or 0.01).

[0072] Step S203: Based on the first time interval, obtain the first value and the second value.

[0073] In this embodiment, the first value is the average increase in lactic acid concentration per unit time within the first time interval. For example, in one embodiment of this application, if the length of the first time interval is 6 hours, and the lactic acid concentration increases from 1 mM to 13 mM within the first time interval, then the average increase in lactic acid concentration per unit time within the first time interval is (13 mM - 1 mM) / 6 hours = 2 mM / hour.

[0074] In this embodiment, the second value is the average rate at which the respiratory entropy (RQ) of cells recovers from its initial value to a fourth preset value within the first time interval. For example, in one embodiment of this application, if the time it takes for the respiratory entropy of cells to recover from its initial value to the fourth preset value within the first time interval is 4 hours, then the value corresponding to the average rate can be 1 / 4; if the time it takes for the respiratory entropy of cells to recover from its initial value to the fourth preset value within the first time interval is 6 hours, then the value corresponding to the average rate can be 1 / 6, and so on. Cellular respiratory entropy refers to the molar ratio of carbon dioxide to oxygen produced by cell metabolism. It is important to understand that the core function of respiratory entropy is as a quantitative indicator of cellular metabolic state. That is, respiratory entropy can indirectly reflect the type of energy substrate, metabolic pathway selection, and overall metabolic activity of cells under specific conditions. If, after cell inoculation, the respiratory entropy recovers to a normal level (i.e., the fourth preset value) faster, then the cell activity is stronger, and that is, the cell state is better.

[0075] In this embodiment, the fourth preset value is set according to the requirements. For example, the fourth preset value can be 0.7, 0.8 or 0.9, etc.

[0076] Step S204: Based on the first cluster distribution entropy, the second cluster distribution entropy, the first value, and the second value, obtain the first feature value of each group of historical cell cultures.

[0077] In this embodiment, the first feature value is the feature value corresponding to the incubation period of each group of historical cell cultures. The first feature value of each group of historical cell cultures can be obtained in any reasonable way based on the first cluster distribution entropy, the second cluster distribution entropy, the first value, and the second value. For example, in one embodiment of this application, step S204, the calculation formula for obtaining the first feature value of each group of historical cell cultures based on the first cluster distribution entropy, the second cluster distribution entropy, the first value, and the second value is as follows:

[0078]

[0079] in, Represents the first eigenvalue; Indicates the first numerical value; Indicates the second numerical value; Represents the entropy of the first cluster distribution; This represents the entropy of the second cluster distribution.

[0080] In this embodiment, a larger first characteristic value indicates that cells in the latent phase are under high metabolic stress during cell culture, resulting in rapid lactate accumulation, slow recovery of mitochondrial function, and disordered CD3 molecule distribution. This may be due to severe cell damage during inoculation, such as membrane structure disruption after cryopreservation and thawing, or excessive fluctuations in the initial culture environment's pH and temperature, leading to impaired energy metabolism switching and an inability to efficiently transition from glycolysis to oxidative phosphorylation. Conversely, a smaller first characteristic value indicates a high degree of adaptability of the initial culture environment to cell culture.

[0081] In one embodiment of this application, the collector further includes a fluorescent cell counter, a high-performance liquid chromatograph (HPLC), and a PCR amplification instrument. The physiological parameters of the growth phase include the cell division index, the rate of change in the concentration of nucleotide synthesis precursors, and the CD28 transcription level. In this embodiment, the fluorescent cell counter is used to obtain the cell division index of each group of historical cell cultures. The HPLC is used to obtain the rate of change in the concentration of nucleotide synthesis precursors in each group of historical cell cultures. The PCR amplification instrument is used to obtain the CD28 transcription level of each group of historical cell cultures.

[0082] It is important to note that after the cell culture latency period ends, CD28 transcription levels are the first to be upregulated, lagging behind CD3 expression by approximately 2 to 4 hours. This aligns with the temporal logic of co-stimulatory signals. The increase in CD28 transcription levels is synergistic with the increase in CD3 molecule expression, indicating receptor-co-stimulatory molecule expression coupling. After CD28 signaling activation, cells enter a rapid division cycle, and the rate of increase in the cell division index is positively correlated with the rate of increase in CD28 transcription levels, representing that co-stimulatory signals drive the cell cycle process. Since the consumption of nucleotide synthesis precursors supports DNA replication, the consumption rate of nucleotide synthesis precursors is linearly related to the rate of increase in the cell division index. In this embodiment, the negative value of the rate of change in the concentration of nucleotide synthesis precursors is used as the consumption rate of nucleotide synthesis precursors. If the second derivative of the cell division index, i.e., the rate of change of the cell division index, changes from positive to negative, it indicates that the acceleration of division has slowed down, the consumption rate of nucleotide synthesis precursors has dropped to below half the peak value, indicating a reduction in raw material demand, and when the rate of increase of CD28 transcription level and cell division index decreases by more than half of their respective peak rates, it indicates that the driving force of co-stimulatory signals on proliferation has weakened, suggesting that cell proliferation activity has begun to decline and the cell has entered a plateau phase. Based on this, in one embodiment of this application, the acquisition of the characteristic values ​​corresponding to the growth phase includes steps S205 to S207.

[0083] Step S205: Obtain a second time interval based on the cell division index, the concentration change rate, and the CD28 transcription level.

[0084] In this embodiment, the start time of the second time interval is the end time of the first time interval. The end time of the second time interval is the time point when the rate of change in concentration and the rate of increase in CD28 transcription level are both less than 0.5 times their highest rate of increase, and the absolute value of the second difference of the cell division index is less than or equal to a fifth preset value.

[0085] In this embodiment, a fifth preset value can be set according to requirements. For example, the fifth preset value can be 0, or any value close to 0 (e.g., 0.01 or 0.1, etc.).

[0086] Step S206: Based on the second time interval, obtain the third, fourth, fifth and sixth values.

[0087] In this embodiment, the third value is the average growth rate of CD28 transcription levels within the second time interval. The fourth value is the CD3 expression level corresponding to the end time of the second time interval. The fifth value is the nucleotide consumption entropy within the second time interval. The sixth value is the cell division index within the second time interval.

[0088] Step S207: Based on the third value, the fourth value, the fifth value, and the sixth value, obtain the second characteristic value for each group of historical cell cultures.

[0089] In this embodiment, the second feature value is the feature value corresponding to the growth period of each group of historical cell cultures. The second feature value of each group of historical cell cultures can be obtained in any reasonable manner based on the third, fourth, fifth, and sixth values. For example, in one embodiment of this application, step S207: the calculation formula for obtaining the second feature value of each group of historical cell cultures based on the third, fourth, fifth, and sixth values ​​can be as follows:

[0090]

[0091] in, Indicates the second eigenvalue; Indicates the third value; Indicates the fourth value; Indicates the fifth numerical value; This represents the sixth numerical value.

[0092] In this embodiment, a larger second eigenvalue indicates an excessively strong co-stimulatory signal and an imbalanced proliferation during cell culture. Specifically, the rate of increase in CD28 transcription is much higher than that of CD3 expression, nucleotide consumption is disordered, and the mitotic index does not increase synchronously. This may be due to excessive nutrient supply, such as high glucose concentration or dissolved oxygen, leading to overactivation of the co-stimulatory pathway; or, disordered utilization of DNA replication raw materials (e.g., nucleotide synthesis precursors), such as excessively rapid consumption of some nucleotide synthesis precursors, resulting in local shortages, which in turn leads to chromosomal instability and a faster rate of apoptosis during the subsequent plateau phase.

[0093] In one embodiment of this application, the data acquisition device further includes a confocal microscope. Physiological parameters during the plateau phase include IL-2 secretion concentration, IFN-γ secretion concentration, and the uniformity of CD62L distribution on the cell membrane. The flow cytometer is also used to acquire time-series curves of IL-2 and IFN-γ secretion concentrations from each group of historical cell cultures. The confocal microscope is used to acquire the uniformity of CD62L distribution on the cell membrane from each group of historical cell cultures.

[0094] It is important to note that while cell proliferation slows down during the growth phase, IL-2 secretion concentration rises first, marking an early sign of immune function activation. The rate of increase in IL-2 secretion concentration is negatively correlated with the rate of decline in CD28 transcription levels, indicating that weakened proliferation signals lead to enhanced functional signals. Driven by IL-2 signaling, the membrane localization uniformity of the effector marker CD62L gradually increases, resulting in an ordered molecular distribution. The rate of increase in CD62L localization uniformity is lagging behind the rate of increase in IFN-γ secretion, forming a cascade effect of IL-2→CD62L→IFN-γ. Meanwhile, the apoptosis rate remains at a low level and its rate of increase is stable, indicating a dynamic balance between proliferation and apoptosis. When the rate of increase in both IL-2 and IFN-γ secretion concentrations is less than 10% of their initial values, it indicates that cell culture is gradually entering a plateau phase. Based on this, in one embodiment of this application, the acquisition of the characteristic values ​​corresponding to the plateau phase includes steps S208 to S210.

[0095] Step S208: Based on the uniformity of the positioning distribution, obtain the third time interval.

[0096] In this embodiment, the start time of the third time interval is the end time of the second time interval, and the end time of the third time interval is the time point when the uniformity of the temporally adjacent positioning distribution meets the first preset condition.

[0097] In this embodiment, the first preset condition can be any condition that can determine the end of the plateau phase of cells. For example, if the uniformity of the localization distribution (i.e., the uniformity of the localization distribution of CD62L) at a certain time point is equal to a sixth preset value, then that time point can be considered to meet the first preset condition. The sixth preset value can be set according to needs; for example, the sixth preset value can be 70% or 80%, etc.

[0098] To accurately determine the end of the plateau phase, in one embodiment of this application, the ratio of the standard deviation to the mean of five adjacent detection points can be calculated, centered on each detection point in the time series data corresponding to the uniformity of CD62L's localization distribution (in the computer field, calculating the ratio of the standard deviation to the mean of multiple values ​​is a mature technique, which will not be elaborated here), to obtain the coefficient of variation for each detection point. The average of the coefficients of variation of three consecutive detection points is obtained as the fluctuation range of the uniformity of CD62L's localization distribution. This is compared with the average of the coefficients of variation of the previous detection window (either the latter minus the former, or the former minus the latter). If the change exceeds 15%, it indicates a decrease in the stability of the functional marker and an increase in the rate of apoptosis of more than double compared to the previous period, indicating the accumulation of senescent cells and marking the peak of cell functional maturation, i.e., the plateau phase structure of the cell, and the cell entering the decline phase. In other words, in this embodiment, the time point corresponding to a change exceeding 15% can be considered as the time point that meets the first preset condition.

[0099] Step S209: Based on the third time interval, obtain the seventh, eighth, and ninth values.

[0100] In this embodiment, the seventh value is the uniformity of CD62L localization distribution corresponding to the end time of the third time interval. The eighth value is the average IL-2 secretion concentration within the third time interval. The ninth value is the average slope of the time-series curve of IFN-γ secretion concentration within the third time interval.

[0101] Step S210: Based on the seventh value, the eighth value, and the ninth value, obtain the third characteristic value of each group of historical cell cultures.

[0102] In this embodiment, the third feature value is the feature value corresponding to the plateau phase of each group of historical cell cultures. The third feature value of each group of historical cell cultures can be obtained in any reasonable manner based on the seventh, eighth, and ninth values. For example, in one embodiment of this application, step S210, the calculation formula for obtaining the third feature value of each group of historical cell cultures based on the seventh, eighth, and ninth values, can be as follows:

[0103]

[0104] in, Indicates the third eigenvalue; Indicates the seventh value; Indicates the eighth value; Indicates the ninth value; This represents a logarithmic function with the natural constant as its base.

[0105] In this embodiment, a larger third characteristic value indicates more stable cell function and a greater ability to delay apoptosis, meaning mature and balanced cell function, uniform CD62L distribution (strong homing ability), moderate IL-2 secretion (not excessively driving proliferation), and rapid IFN-γ secretion (strong effector function). Conversely, a smaller third characteristic value indicates disordered or immature cell function, which may be due to the accumulation of metabolic byproducts during the plateau phase or excessively high cell density, leading to cell stress and hindering functional differentiation, resulting in terminal cell quality that does not meet clinical application standards.

[0106] Step S300: Based on each data feature group, obtain the range of high-quality physiological parameters corresponding to each growth cycle.

[0107] As mentioned above, for each group of cultured cells, the smaller the first characteristic value, the moderate the second characteristic value, and the larger the third characteristic value, the more successful the cell culture is. In other words, in this embodiment, the physiological parameters corresponding to historical cell cultures that meet the above characteristics can be considered as high-quality physiological parameters. Based on this, in one embodiment of this application, step S300, based on each data characteristic group, obtains the range of high-quality physiological parameters corresponding to each growth cycle, including steps S310 to S340.

[0108] Step S310: Project each data feature group into a multidimensional space to form multiple spatial coordinate points.

[0109] It is important to understand that in this embodiment, if a data feature group has several feature values, then each data feature group is projected into a space of that dimension. For example, as seen in the previous embodiments, since each data feature group has three feature values ​​(i.e., a first feature value, a second feature value, and a third feature value), each data feature group can be projected into a three-dimensional space. That is, each data feature group forms a corresponding spatial coordinate point in the three-dimensional space.

[0110] Step S320: Based on the density clustering algorithm, obtain multiple clusters from each spatial coordinate point.

[0111] In the embodiments of this application, any reasonable density clustering algorithm can be used to obtain multiple clusters for each spatial coordinate point. For example, the density clustering algorithm can be the DBSCAN algorithm, the OPTICS algorithm, or the DENCLUE algorithm, etc.

[0112] Step S330: Based on each cluster, obtain the optimal cluster.

[0113] In this embodiment, the optimal cluster is any one cluster or a combination of multiple clusters that meet the second preset condition.

[0114] In this embodiment, the second preset condition can be set according to requirements. For example, if in a certain cluster, the corresponding first feature value is small (e.g., the average of each first feature value in the cluster is less than or equal to 0.5 times the average of all first feature values); the corresponding second feature value is moderate (e.g., the average of each second feature value in the cluster is greater than or equal to 0.9 times the average of all second feature values, and less than or equal to 1.1 times the average of all second feature values); and the corresponding third feature value is large (e.g., the average of each third feature value in the cluster is greater than or equal to 1.5 times the average of all third feature values). The above multiples (i.e., 0.5 times, 0.9 times, 1.1 times, and 1.5 times, etc.) can be set according to requirements. This embodiment is only an example and does not represent a specific limitation of this application.

[0115] Step S340: Based on the optimal cluster, obtain the range of high-quality physiological parameters.

[0116] In this embodiment, the lower limit of the range of high-quality physiological parameters is the minimum value in the optimal cluster, and the upper limit of the range of high-quality physiological parameters is the maximum value in the optimal cluster.

[0117] Step S400: Obtain multiple correlation coefficients based on each data feature group.

[0118] In this embodiment, the correlation coefficient is the Pearson correlation coefficient between the first culture parameter and the characteristic value corresponding to each growth cycle. The first culture parameter is the actual culture parameter for each growth cycle in the historical cell culture. If a certain first culture parameter has a significant impact on the characteristic value corresponding to a certain growth cycle (i.e., the latent phase, growth phase, or plateau phase), it indicates that the Pearson correlation coefficient between the first culture parameter and the corresponding characteristic value is larger.

[0119] It's important to understand that the Pearson Correlation Coefficient (PCC) is a statistic that measures the degree of linear correlation between two continuous variables. Its core function is to quantify whether a change in the first variable (i.e., the eigenvalue) is accompanied by a linear change in the second variable (i.e., the first culture parameter), directly reflecting the direction and strength of the linear correlation between the first and second variables. In other words, obtaining the Pearson correlation coefficient between a given first culture parameter and its corresponding eigenvalue is a well-established technique, and will not be elaborated upon here.

[0120] Step S500: Based on each Pearson correlation coefficient, obtain multiple second culture parameters.

[0121] In this embodiment, the second culture parameter is the culture parameter whose Pearson correlation coefficient is greater than or equal to the first preset value among the various first culture parameters.

[0122] In this embodiment, the first preset value can be set according to requirements. For example, the first preset value can be 0.5 or 0.6. It is easy to understand that the larger the first preset value, the greater the influence of the obtained second cultivation parameters on the feature values ​​(e.g., the first feature value, the second feature value, or the third feature value mentioned above).

[0123] Step S600: Treat each second culture parameter of each group of historical cell cultures as a particle and obtain the current particle swarm.

[0124] It's important to understand that Particle Swarm Optimization (PSO) is a heuristic optimization algorithm based on "swarm intelligence." Its core idea is to mimic the collective cooperative behavior of biological groups such as flocks of birds foraging and schools of fish migrating. It gradually searches for the optimal solution to a problem through information sharing and dynamic adjustments among individuals (particles) within the group. It does not rely on the mathematical gradient information of the problem and is suitable for complex nonlinear and multi-extremum optimization problems, with wide applications in function optimization, neural network training, and engineering design. In other words, in this embodiment, each second culture parameter in each group of historical cell cultures represents a particle in the PSO algorithm.

[0125] Step S700: Based on the range of superior physiological parameters, obtain the current fitness value of the current particle swarm.

[0126] In this embodiment, the current fitness value is used at least to characterize the probability that the physiological parameters corresponding to each particle in the current particle swarm fall within the range of high-quality physiological parameters.

[0127] It is important to understand that the higher the probability that the physiological parameters corresponding to a certain particle fall within the range of optimal physiological parameters, the better the culture parameters corresponding to that particle are. In other words, the culture parameters corresponding to that particle should be referenced more frequently during subsequent adjustments to the culture parameters. Based on this, step S700, based on the range of optimal physiological parameters, obtains the current fitness value of the current particle swarm, including steps S710 to S730.

[0128] Step S710: Based on the current particle swarm, obtain the optimal probability corresponding to each particle.

[0129] In this embodiment, the optimal probability is the probability that the physiological parameters corresponding to each growth cycle of the particle all fall within the range of optimal physiological parameters. A reasonable method can be used to obtain the optimal probability for each particle based on the current particle swarm. For example, in one embodiment of this application, step S710, obtaining the optimal probability for each particle based on the current particle swarm, includes steps S711 to S713.

[0130] Step S711: Based on the current particle swarm, obtain the current particle.

[0131] In this embodiment, the current particle is any particle in the current particle swarm that needs to have its optimal probability calculated according to the calculation order.

[0132] Step S712: Based on the current particle, obtain multiple falling probabilities.

[0133] In this embodiment, the probability of falling into the optimal range is at least used to characterize the probability that adjusting the culture parameters of the current particle will make it easier for the physiological parameters corresponding to the current particle to fall into the optimal range. It is important to understand that the closer the physiological parameters of a particle are to the optimal range, the easier it is for the culture parameters of that particle to fall into the optimal range when adjusted; that is, the higher the probability of falling into the optimal range for that particle.

[0134] In this embodiment, multiple fall-in probabilities can be obtained based on the current particle using any reasonable method. For example, if the first feature value corresponding to the current particle is a; the second feature value is b; and the third feature value is c; wherein the range of optimal physiological parameters corresponding to the latency period is [ , The optimal physiological parameter range corresponding to the growth period is [ ]; , The optimal physiological parameter range corresponding to the plateau phase is [ ]; , The formula for calculating the probability of each particle falling into the air can be as follows:

[0135]

[0136]

[0137]

[0138] in, This indicates the probability of falling in during the incubation period; This represents the probability of falling into the air during the growth period; This indicates the probability of falling into the plateau phase. This indicates the absolute value; This represents the zero-prevention coefficient, which can be any value close to 0. For example: zero-prevention coefficient It can be 0.01 or 0.001, etc.; This represents a normalization function used to normalize the values ​​within the parentheses to the range [0, 1].

[0139] Alternatively, the formula for calculating the probability of each particle falling into the air can be as follows:

[0140]

[0141]

[0142]

[0143] in, This indicates the probability of falling in during the incubation period; This represents the probability of falling into the air during the growth period; This indicates the probability of falling into the plateau phase. This indicates the absolute value; This represents the zero-prevention coefficient, which can be any value close to 0. For example: zero-prevention coefficient It can be 0.01 or 0.001, etc.; This represents a normalization function used to normalize the values ​​within the parentheses to the range [0, 1].

[0144] Step S713: Based on each fall probability, obtain the optimal probability corresponding to the current particle.

[0145] In this embodiment, the optimal probability can be the product of the various fall probabilities corresponding to the current particle; the higher the optimal probability, the better the cultivation parameters corresponding to the current particle.

[0146] Step S720: Based on each optimal probability, obtain the probability fluctuation value and the probability average value.

[0147] In this embodiment, the probability fluctuation value is used at least to characterize the degree of difference between the various optimal probabilities. The probability average value is the average of the various optimal probabilities.

[0148] Step S730: Obtain the current fitness value based on the probability fluctuation value and the probability average value.

[0149] It is important to understand that if the differences between the optimal probabilities are smaller and the average value is larger, it indicates that the optimal training parameters have been reached, meaning that updates to the current particle swarm can be stopped. In other words, in the embodiments of this application, the current fitness value can be positively correlated with the average probability value and negatively correlated with the probability fluctuation value. Based on this, step S730: the formula for calculating the current fitness value based on the probability fluctuation value and the average probability value can be as follows:

[0150]

[0151] in, This indicates the current fitness value; This represents the average probability. Indicates the probability fluctuation value; This represents the zero-prevention coefficient, which can be any value close to 0. For example: zero-prevention coefficient It can be 0.01 or 0.001, etc.

[0152] Step S800: Iteratively update the current particle swarm based on the particle swarm algorithm so that the updated current fitness value is greater than or equal to the second preset value, thereby obtaining the globally optimal parameter combination for cell culture.

[0153] It should be noted that iteratively updating the current particle swarm based on the particle swarm algorithm is a mature technology, which will not be elaborated here.

[0154] In this embodiment, the second preset value can be set according to requirements, for example, the second preset value can be 5 or 10, etc.

[0155] Step S900: Based on the globally optimal parameter combination, dynamically adjust the second culture parameter of the current cell culture growth cycle.

[0156] It is important to understand that the globally optimal parameter combination obtained in step S800 is the combination of culture parameters that ensures the characteristic values ​​for each cell culture cycle fall within the range of optimal physiological parameters. In other words, the globally optimal parameter combination allows for the optimal state of cell culture. It is also important to understand that adjusting each culture parameter (i.e., the second culture parameter) to match the culture parameters in the globally optimal parameter combination is a mature technique, and will not be elaborated upon here.

[0157] The embodiments of the multi-parameter collaborative AI dynamic optimization method for cell culture proposed in this application overcome the limitations of traditional static optimization methods that cannot adapt to the dynamic time-varying processes of cells by accurately anchoring cellular metabolic pressure, co-stimulatory coupling, and functional equilibrium. Simultaneously, based on feature value multidimensional clustering to mine the patterns of high-quality cell populations, the particle swarm optimization algorithm is improved to obtain fitness values ​​that characterize the quality of cell culture. This achieves dynamic coupling analysis between culture parameters and cell metabolism and phenotypic state, overcoming the regulatory lag problem caused by the inability of traditional optimization methods to analyze the dynamic coupling between metabolism and culture parameters and the difficulty in tracking cell demand within a window. Ultimately, it achieves collaborative dynamic optimization of culture parameters throughout the entire growth cycle, ensuring timely regulation of cells at each stage and improving the stability of terminal cell quality.

[0158] Having introduced the cell culture multi-parameter collaborative AI dynamic optimization method proposed in the embodiments of this application, the following describes a cell culture multi-parameter collaborative AI dynamic optimization system proposed in this application. For example... Figure 2 As shown, the cell culture multi-parameter collaborative AI dynamic optimization system 10 includes:

[0159] The data acquisition device 11 is used to acquire growth cycle data of multiple historical cell cultures; the growth cycle includes at least a latent phase, a growth phase, and a plateau phase; the growth cycle data includes at least the first culture parameter and physiological parameter of each historical cell culture corresponding to the growth cycle.

[0160] The processor 12 is used to acquire a data feature set corresponding to each group of cells based on the physiological parameters of each group of historical cell cultures; the data feature set includes feature values ​​corresponding to each growth cycle; the feature values ​​are used at least to characterize the quality of historical cell cultures for the corresponding growth cycle.

[0161] Furthermore, based on each data feature group, the range of high-quality physiological parameters corresponding to each growth cycle is obtained;

[0162] Furthermore, based on each data feature group, multiple correlation coefficients are obtained; the correlation coefficients are the Pearson correlation coefficients between the first culture parameter and the feature value corresponding to each growth cycle.

[0163] Furthermore, based on each Pearson correlation coefficient, multiple second culture parameters are obtained; the second culture parameters are the culture parameters whose corresponding Pearson correlation coefficients are greater than or equal to a first preset value among the various first culture parameters.

[0164] Additionally, each second culture parameter of each group of historical cell cultures is treated as a particle to obtain the current particle swarm;

[0165] Furthermore, based on the range of superior physiological parameters, the current fitness value of the current particle swarm is obtained; the current fitness value is used at least to characterize the probability that the physiological parameters corresponding to each particle in the current particle swarm fall within the range of superior physiological parameters.

[0166] Furthermore, the current particle swarm is iteratively updated based on the particle swarm algorithm so that the updated current fitness value is greater than or equal to the second preset value, thereby obtaining the globally optimal parameter combination for cell culture;

[0167] The controller 13 is used to dynamically adjust the second culture parameter of the current cell culture growth cycle based on the globally optimal parameter combination.

[0168] As a specific embodiment of this application, the collector 11 includes a lactate analyzer, a dissolved oxygen electrode, and a flow cytometer; the physiological parameters of the latency period include lactate concentration, basal respiratory rate, and fluorescence intensity of CD3 molecules;

[0169] The lactate analyzer is used to obtain the lactate concentration of each group of historical cell cultures;

[0170] The dissolved oxygen electrode is used to obtain the baseline respiratory rate of each group of historical cell cultures; the baseline respiratory rate is the dissolved oxygen consumption per unit cell per hour.

[0171] The flow cytometer is used to obtain the fluorescence intensity of CD3 molecules from each group of historical cell cultures.

[0172] As a specific embodiment of this application, the processor 12 is further configured to obtain the cluster distribution entropy based on the fluorescence intensity of the CD3 molecules;

[0173] Furthermore, based on the baseline respiration rate, a first time interval for each group of historical cell cultures is obtained; the start time of the first time interval is the start time of the historical cell culture, and the end time is any time point where the baseline respiration rate is still increasing and the difference between the first cluster distribution entropy and the second cluster distribution entropy is less than a third preset value; the first cluster distribution entropy is the cluster distribution entropy corresponding to the start time of the first time interval; the second cluster distribution entropy is any cluster distribution entropy whose time sequence is after the first cluster distribution entropy;

[0174] And, based on the first time interval, a first value and a second value are obtained; the first value is the average value of the increase in lactic acid concentration per unit time within the first time interval; the second value is the average rate at which the respiratory entropy of cells recovers from the initial value to the fourth preset value within the first time interval.

[0175] Furthermore, based on the first clustering distribution entropy, the second clustering distribution entropy, the first value, and the second value, a first feature value for each group of historical cell cultures is obtained; the first feature value is the feature value corresponding to the incubation period of each group of historical cell cultures.

[0176] As a specific embodiment of this application, the collector 11 further includes a fluorescence cell counter, a high-performance liquid chromatograph, and a PCR amplification instrument; the physiological parameters of the growth phase include the cell division index, the rate of change in the concentration of nucleotide synthesis precursors, and the CD28 transcription level;

[0177] The fluorescent cell counter is used to obtain the cell division index of each group of historical cell cultures;

[0178] The high-performance liquid chromatograph is used to obtain the rate of change in the concentration of nucleotide synthesis precursors in each group of historical cell cultures.

[0179] The PCR amplification instrument is used to obtain the CD28 transcription level of each group of historical cell cultures.

[0180] As a specific embodiment of this application, the processor 12 is further configured to obtain a second time interval based on the cell division index, the concentration change rate, and the CD28 transcription level; the start time of the second time interval is the end time of the first time interval, and the end time of the second time interval is the time point when the growth rates of the concentration change rate and the CD28 transcription level are both less than 0.5 times their highest growth rate, and the absolute value of the second difference of the cell division index is less than or equal to a fifth preset value;

[0181] Furthermore, based on the second time interval, a third, fourth, fifth, and sixth value are obtained; the third value is the average growth rate of CD28 transcription level within the second time interval; the fourth value is the CD3 expression level corresponding to the end time of the second time interval; the fifth value is the nucleotide consumption entropy within the second time interval; and the sixth value is the cell division index within the second time interval.

[0182] Furthermore, based on the third, fourth, fifth, and sixth values, a second characteristic value is obtained for each group of historical cell cultures; the second characteristic value is the characteristic value corresponding to the growth period of each group of historical cell cultures.

[0183] As a specific embodiment of this application, the collector 11 also includes a confocal microscope; the physiological parameters of the plateau phase include IL-2 secretion concentration, IFN-γ secretion concentration and the uniformity of CD62L localization and distribution on the cell membrane;

[0184] The flow cytometer is also used to obtain time-series curves of IL-2 secretion concentration and IFN-γ secretion concentration of each group of historical cell cultures.

[0185] The confocal microscope is used to obtain the uniformity of CD62L localization and distribution on the cell membrane of each group of historical cell cultures.

[0186] As a specific embodiment of this application, the processor 12 is further configured to obtain a third time interval based on the uniformity of the positioning distribution; the start time of the third time interval is the end time of the second time interval, and the end time of the third time interval is the time point when the uniformity of the positioning distribution of temporally adjacent units meets the first preset condition.

[0187] Furthermore, based on the third time interval, a seventh, eighth, and ninth value are obtained; the seventh value is the uniformity of CD62L localization distribution corresponding to the end time of the third time interval; the eighth value is the average value of IL-2 secretion concentration within the third time interval; and the ninth value is the average slope of the time-series curve of IFN-γ secretion concentration within the third time interval.

[0188] Furthermore, based on the seventh value, the eighth value, and the ninth value, a third characteristic value is obtained for each group of historical cell cultures; the third characteristic value is the characteristic value corresponding to the plateau phase of each group of historical cell cultures.

[0189] As a specific embodiment of this application, the processor 12 is further configured to project each data feature group into a multi-dimensional space to form multiple spatial coordinate points;

[0190] Furthermore, based on the density clustering algorithm, multiple clusters are obtained from each spatial coordinate point;

[0191] Furthermore, based on each cluster, the optimal cluster is obtained; the optimal cluster is any one cluster or a combination of multiple clusters that meet the second preset condition.

[0192] Furthermore, based on the optimal cluster, the range of high-quality physiological parameters is obtained; the lower limit of the range of high-quality physiological parameters is the minimum value in the optimal cluster, and the upper limit of the range of high-quality physiological parameters is the maximum value in the optimal cluster.

[0193] As a specific embodiment of this application, the processor 12 is further configured to obtain the optimal probability corresponding to each particle based on the current particle swarm; the optimal probability is the probability that the physiological parameters corresponding to each growth cycle of the corresponding particle fall within the range of the high-quality physiological parameters.

[0194] Furthermore, based on each optimal probability, a probability fluctuation value and a probability average value are obtained; the probability fluctuation value is used at least to characterize the degree of difference between each optimal probability; the probability average value is the average of each optimal probability.

[0195] Furthermore, the current fitness value is obtained based on the probability fluctuation value and the probability average value.

[0196] As a specific embodiment of this application, the processor 12 is further configured to obtain the current particle based on the current particle swarm; the current particle is any particle in the current particle swarm that currently needs to have its optimal probability calculated according to the calculation order;

[0197] Based on the current particle, multiple fall-in probabilities are obtained; the fall-in probabilities are at least used to characterize the probability that the culture parameters of the current particle are adjusted so that the physiological parameters corresponding to the current particle are more likely to fall into the range of high-quality physiological parameters.

[0198] Based on each fall probability, the optimal probability corresponding to the current particle is obtained.

[0199] The embodiments of the cell culture multi-parameter collaborative AI dynamic optimization system proposed in this application overcome the limitations of traditional static optimization methods that cannot adapt to the dynamic time-varying processes of cells by accurately anchoring cellular metabolic pressure, co-stimulatory coupling, and functional balance. Simultaneously, based on feature value multidimensional clustering to mine the patterns of high-quality cell populations and improving the particle swarm optimization algorithm, it obtains fitness values ​​that characterize the quality of cell culture, achieving dynamic coupling analysis between culture parameters and cell metabolism and phenotypic state. This overcomes the regulatory lag problem caused by the inability of traditional optimization methods to analyze the dynamic coupling between metabolism and culture parameters and the difficulty in tracking cell demand within a window. Ultimately, it achieves collaborative dynamic optimization of culture parameters throughout the entire growth cycle, ensuring timely regulation of cells at each stage and improving the stability of terminal cell quality.

[0200] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0201] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the methods, apparatuses, and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0202] In the several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between devices or modules may be electrical, mechanical, or other forms.

[0203] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0204] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0205] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0206] The computer program product includes one or more computer instructions. When the computer program is loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video optical disc), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0207] Although embodiments of this application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles of this application.

Claims

1. A multi-parameter collaborative AI dynamic optimization system for cell culture, characterized in that, include: A data acquisition device is used to acquire growth cycle data from multiple historical cell cultures; the growth cycle includes at least a latent phase, a growth phase, and a plateau phase; the growth cycle data includes at least the first culture parameter and physiological parameter for each historical cell culture corresponding to the growth cycle. The processor is used to acquire a data feature set corresponding to each group of cells based on the physiological parameters of each group of historical cell cultures; the data feature set includes feature values ​​corresponding to each growth cycle; the feature values ​​are used to characterize the quality of historical cell cultures at least for the corresponding growth cycle. Furthermore, based on each data feature group, the range of high-quality physiological parameters corresponding to each growth cycle is obtained; Furthermore, multiple correlation coefficients are obtained based on each data feature group; The correlation coefficient is the Pearson correlation coefficient between the first culture parameter and the characteristic value corresponding to each growth cycle; Furthermore, based on each Pearson correlation coefficient, multiple second culture parameters are obtained; the second culture parameters are the culture parameters whose corresponding Pearson correlation coefficients are greater than or equal to a first preset value among the various first culture parameters. Additionally, each second culture parameter of each group of historical cell cultures is treated as a particle to obtain the current particle swarm; And, based on the range of superior physiological parameters, obtain the current fitness value of the current particle swarm; The current fitness value is used to characterize at least the probability that the physiological parameters corresponding to each particle in the current particle swarm fall within the range of high-quality physiological parameters; Furthermore, the current particle swarm is iteratively updated based on the particle swarm algorithm so that the updated current fitness value is greater than or equal to the second preset value, thereby obtaining the globally optimal parameter combination for cell culture; The controller is used to dynamically adjust the second culture parameter of the current cell culture growth cycle based on the globally optimal parameter combination.

2. The cell culture multi-parameter collaborative AI dynamic optimization system according to claim 1, characterized in that, The data acquisition device includes a lactate analyzer, a dissolved oxygen electrode, and a flow cytometer; the physiological parameters of the latency period include lactate concentration, basal respiratory rate, and fluorescence intensity of CD3 molecules; The lactate analyzer is used to obtain the lactate concentration of each group of historical cell cultures; The dissolved oxygen electrode is used to obtain the baseline respiration rate of each group of historical cell cultures; The basal respiratory rate is the dissolved oxygen consumption per unit cell per hour. The flow cytometer is used to obtain the fluorescence intensity of CD3 molecules from each group of historical cell cultures.

3. The cell culture multi-parameter collaborative AI dynamic optimization system according to claim 2, characterized in that, The processor is also used to obtain the cluster distribution entropy based on the fluorescence intensity of the CD3 molecules; Furthermore, based on the baseline respiration rate, a first time interval for each group of historical cell cultures is obtained; the start time of the first time interval is the start time of the historical cell culture, and the end time is any time point where the baseline respiration rate is still increasing and the difference between the first cluster distribution entropy and the second cluster distribution entropy is less than a third preset value; the first cluster distribution entropy is the cluster distribution entropy corresponding to the start time of the first time interval; the second cluster distribution entropy is any cluster distribution entropy whose time sequence is after the first cluster distribution entropy; And, based on the first time interval, a first value and a second value are obtained; the first value is the average value of the increase in lactic acid concentration per unit time within the first time interval; The second value is the average rate at which the respiratory entropy of cells recovers from its initial value to the fourth preset value within the first time interval; Furthermore, based on the first clustering distribution entropy, the second clustering distribution entropy, the first value, and the second value, a first feature value for each group of historical cell cultures is obtained; the first feature value is the feature value corresponding to the incubation period of each group of historical cell cultures.

4. The cell culture multi-parameter collaborative AI dynamic optimization system according to claim 3, characterized in that, The collector also includes a fluorescence cell counter, a high-performance liquid chromatograph, and a PCR amplification instrument; the physiological parameters of the growth phase include the cell division index, the rate of change in the concentration of nucleotide synthesis precursors, and the CD28 transcription level; The fluorescent cell counter is used to obtain the cell division index of each group of historical cell cultures; The high-performance liquid chromatograph is used to obtain the rate of change in the concentration of nucleotide synthesis precursors in each group of historical cell cultures. The PCR amplification instrument is used to obtain the CD28 transcription level of each group of historical cell cultures.

5. The cell culture multi-parameter collaborative AI dynamic optimization system according to claim 4, characterized in that, The processor is further configured to obtain a second time interval based on the cell division index, the concentration change rate, and the CD28 transcription level; the start time of the second time interval is the end time of the first time interval, and the end time of the second time interval is the time point when the growth rates of the concentration change rate and the CD28 transcription level are both less than 0.5 times their highest growth rate, and the absolute value of the second difference of the cell division index is less than or equal to a fifth preset value. Furthermore, based on the second time interval, a third, fourth, fifth, and sixth value are obtained; the third value is the average growth rate of CD28 transcription level within the second time interval; the fourth value is the CD3 expression level corresponding to the end time of the second time interval; the fifth value is the nucleotide consumption entropy within the second time interval; and the sixth value is the cell division index within the second time interval. Furthermore, based on the third, fourth, fifth, and sixth values, a second characteristic value is obtained for each group of historical cell cultures; the second characteristic value is the characteristic value corresponding to the growth period of each group of historical cell cultures.

6. The cell culture multi-parameter collaborative AI dynamic optimization system according to claim 5, characterized in that, The collector also includes a confocal microscope; the physiological parameters of the plateau phase include IL-2 secretion concentration, IFN-γ secretion concentration, and the uniformity of CD62L localization and distribution on the cell membrane. The flow cytometer is also used to obtain time-series curves of IL-2 secretion concentration and IFN-γ secretion concentration of each group of historical cell cultures. The confocal microscope is used to obtain the uniformity of CD62L localization and distribution on the cell membrane of each group of historical cell cultures.

7. The cell culture multi-parameter collaborative AI dynamic optimization system according to claim 6, characterized in that, The processor is further configured to obtain a third time interval based on the uniformity of the positioning distribution; the start time of the third time interval is the end time of the second time interval, and the end time of the third time interval is the time point when the uniformity of the positioning distribution of temporally adjacent units meets the first preset condition. Furthermore, based on the third time interval, a seventh, eighth, and ninth value are obtained; the seventh value is the uniformity of CD62L localization distribution corresponding to the end time of the third time interval; the eighth value is the average value of IL-2 secretion concentration within the third time interval; and the ninth value is the average slope of the time-series curve of IFN-γ secretion concentration within the third time interval. Furthermore, based on the seventh value, the eighth value, and the ninth value, a third characteristic value is obtained for each group of historical cell cultures; the third characteristic value is the characteristic value corresponding to the plateau phase of each group of historical cell cultures.

8. The cell culture multi-parameter collaborative AI dynamic optimization system according to claim 7, characterized in that, The processor is also used to project each data feature group into a multidimensional space to form multiple spatial coordinate points; Furthermore, based on the density clustering algorithm, multiple clusters are obtained from each spatial coordinate point; Furthermore, based on each cluster, the optimal cluster is obtained; the optimal cluster is any one cluster or a combination of multiple clusters that meet the second preset condition. Furthermore, based on the optimal cluster, a high-quality range of physiological parameters is obtained; The lower limit of the range of high-quality physiological parameters is the minimum value in the optimal cluster, and the upper limit of the range of high-quality physiological parameters is the maximum value in the optimal cluster.

9. The cell culture multi-parameter collaborative AI dynamic optimization system according to any one of claims 1 to 8, characterized in that, The processor is further configured to obtain the optimal probability for each particle based on the current particle swarm; the optimal probability is the probability that the physiological parameters corresponding to each growth cycle of the corresponding particle fall within the range of the optimal physiological parameters. Furthermore, based on each optimal probability, a probability fluctuation value and a probability average value are obtained; the probability fluctuation value is used at least to characterize the degree of difference between each optimal probability; the probability average value is the average of each optimal probability. Furthermore, the current fitness value is obtained based on the probability fluctuation value and the probability average value.

10. The cell culture multi-parameter collaborative AI dynamic optimization system according to claim 9, characterized in that, The processor is further configured to, based on the current particle swarm, obtain the current particle; the current particle is any particle in the current particle swarm that currently needs to have its optimal probability calculated according to the calculation order; Based on the current particle, multiple fall-in probabilities are obtained; the fall-in probabilities are at least used to characterize the probability that the culture parameters of the current particle are adjusted so that the physiological parameters corresponding to the current particle are more likely to fall into the range of high-quality physiological parameters. Based on each fall probability, the optimal probability corresponding to the current particle is obtained.

Images