Cell activity and purity detection method and system in chimeric antigen receptor T cell therapy

Abstract

The invention discloses a method and a system for detecting cell viability and purity in chimeric antigen receptor T cell therapy. The cell viability and purity detection method in the chimeric antigen receptor T cell therapy comprises the following steps: obtaining a to-be-detected CAR-T cell sample; dyeing a CAR-T cell sample to be detected so as to obtain a dyed cell suspension; detecting the dyed cell suspension by using a flow cytometer so as to obtain a fluorescence signal; and carrying out cell activity detection and cell purity analysis according to the fluorescence signal so as to obtain cell activity information and cell purity information. The cell activity and purity detection method in the chimeric antigen receptor T cell therapy comprehensively considers multiple aspects of a cell metabolic network, including metabolic flux, metabolic pathways, metabolic network structures and the like, and can more comprehensively reflect the cell activity and purity.

Description

Technical Field

[0001] The present application relates to the technical field of chimeric antigen receptor T cells, and in particular to a method for detecting cell viability and purity in chimeric antigen receptor T cell therapy and a system for detecting cell viability and purity in chimeric antigen receptor T cell therapy. Background Art

[0002] In the field of chimeric antigen receptor T-cell (CAR-T) therapy, viability and purity testing of cell products is a critical component of quality control, directly impacting clinical efficacy and patient safety. Traditional detection methods primarily rely on trypan blue staining, flow cytometry, and fluorescent-labeled metabolic activity assays. While these methods can reflect cell status to a certain extent, they also have numerous limitations and are unable to meet growing clinical needs.

[0003] Trypan blue staining, a classic cell viability assay, distinguishes live from dead cells by measuring cell membrane integrity. However, this method is a terminal assay and cannot reflect changes in cellular metabolic status in real time. Furthermore, trypan blue staining cannot distinguish between cells with different causes of death or provide detailed information about cellular metabolic activity.

[0004] Therefore, it is desired to have a technical solution to overcome or at least alleviate at least one of the above-mentioned deficiencies of the prior art.

[0005] Application Contents

[0006] The purpose of the present application is to provide a method for detecting cell viability and purity in chimeric antigen receptor T cell therapy to overcome or at least alleviate at least one of the above-mentioned defects of the prior art.

[0007] To achieve the above objectives, the present application provides a method for detecting cell viability and purity in chimeric antigen receptor T cell therapy, wherein the method comprises:

[0008] Obtain CAR-T cell samples to be tested;

[0009] Staining the CAR-T cell sample to be tested to obtain a stained cell suspension;

[0010] The stained cell suspension is detected using a flow cytometer to obtain a fluorescent signal;

[0011] Cell viability detection and cell purity analysis are performed based on the fluorescence signal to obtain cell viability information and cell purity information.

[0012] Optionally, performing cell viability detection and cell purity analysis based on the fluorescence signal to obtain cell viability information and cell purity information includes:

[0013] performing enzyme-substrate coupling kinetics-inspired nonlinear fluorescence signal enhancement processing on the fluorescence signal, thereby obtaining an enhanced fluorescence signal matrix;

[0014] A domain graph is constructed based on the enhanced fluorescence signal matrix to obtain a weighted neighborhood graph guided by the Laplacian matrix of the metabolic network.

[0015] Obtaining an activity index and an activity distribution map according to a weighted neighborhood graph guided by the Laplacian matrix of the metabolic network;

[0016] A purity index and a purity distribution graph are obtained according to a weighted neighborhood graph guided by the Laplace matrix of the metabolic network.

[0017] Optionally, performing enzyme-substrate coupling kinetics-inspired nonlinear fluorescence signal enhancement processing on the fluorescence signal to obtain an enhanced fluorescence signal matrix includes:

[0018] The final signal intensity of each cell i was calculated using the following formula:

[0019]

[0020] in, represents the enhanced fluorescence signal, representing the final signal intensity of cell i in the c fluorescence channel; x i,c is the original fluorescence signal, representing the initial signal intensity of cell i in the c fluorescence channel; V max,c is the maximum enhancement rate of fluorescence channel c, indicating the theoretical maximum efficiency of the enzyme-catalyzed reaction; K M,c is the Michaelis constant of fluorescence channel c, reflecting the affinity between enzyme and substrate, K M,c The smaller the value, the stronger the binding ability between enzyme and substrate; inh,c is the inhibitory factor of fluorescence channel c, representing the substance that inhibits enzyme activity in the cell, estimated by inversion of CellROX signal; K i,c is the inhibition constant of fluorescence channel c, which indicates the inhibitory intensity of the inhibitory factor on enzyme activity; K i,c The smaller the value, the stronger the inhibitory effect; c is the weight coefficient of the time domain convolution term of the fluorescence channel c, which controls the contribution ratio of the historical signal to the current enhancement; β c is the time domain attenuation coefficient of the fluorescence channel c, which indicates the decay rate of the historical signal over time; γ c is the nonlinear suppression coefficient of the fluorescence channel c, which controls the dynamic suppression strength of the high signal area; c is the second-order kinetic coefficient of fluorescence channel c, which indicates the intensity of the cascade amplification effect of the enzymatic reaction;

[0021] The final signal intensity of each of the cells constitutes the enhanced fluorescence signal matrix.

[0022] Optionally, constructing a domain graph based on the enhanced fluorescence signal matrix to obtain a weighted neighborhood graph guided by a Laplacian matrix of the metabolic network includes:

[0023] Constructing a metabolic network based on a domain graph of the enhanced fluorescence signal matrix, wherein the metabolic network includes nodes and edge information between each node, wherein the nodes represent metabolites and the edges represent metabolic fluxes;

[0024] Obtaining a Laplacian matrix and a modularity matrix according to the metabolic network;

[0025] Obtaining a metabolic flux similarity matrix according to the enhanced fluorescence signal matrix, the Laplace matrix, and the modular matrix;

[0026] A weighted neighborhood graph guided by a Laplacian matrix of a metabolic network is generated according to the metabolic flux similarity matrix.

[0027] Optionally, obtaining a Laplace matrix and a modular matrix according to the metabolic network includes:

[0028] Determine the metabolite or metabolic pathway corresponding to each fluorescence channel based on the enhanced fluorescence signal matrix;

[0029] The changes in fluorescence signals were used to inversely estimate metabolic fluxes;

[0030] Constructing the weight matrix of the metabolic network based on the estimated results of metabolic flux;

[0031] Get the Laplacian matrix based on the weight matrix;

[0032] A modularity matrix is ​​obtained according to the metabolic network.

[0033] Optionally, obtaining a modular matrix according to the metabolic network includes:

[0034] The metabolic network is divided into multiple functional modules using a community detection algorithm. Each module contains a group of metabolite nodes, and the node connectivity within the module is higher than the node connectivity between modules.

[0035] According to the functional module division results, construct the initial modular matrix;

[0036] The constructed initial modularity matrix is ​​normalized to obtain the modularity matrix.

[0037] Optionally, obtaining a metabolic flux similarity matrix according to the enhanced fluorescence signal matrix, the Laplace matrix, and the modular matrix includes:

[0038] Gaussian similarity is calculated for every two cells to obtain Gaussian similarity;

[0039] The Laplace regularization term is calculated based on the metabolic state vectors of every two cells to obtain the Laplace regularization term;

[0040] Obtain the block constraint term based on the modular membership vector of every two cells;

[0041] The Gaussian similarity, the Laplace regularization term, and the modularity constraint term are multiplied to obtain metabolic flux similarity, wherein the metabolic flux similarities of each cell constitute the metabolic flux similarity matrix.

[0042] Optionally, obtaining the vitality index and the vitality distribution map according to the weighted neighborhood graph guided by the metabolic network Laplace matrix includes:

[0043] Obtain metabolic flux dynamic data of each cell and metabolic pathway database;

[0044] Align the metabolic flux dynamic data of each cell to eliminate time sampling bias;

[0045] The time series similarity of metabolic flux between every two cells is calculated based on the metabolic flux dynamic data of each cell after alignment;

[0046] The time series similarity of metabolic fluxes between cells is incorporated into the neighborhood matrix to obtain a weighted domain matrix;

[0047] Obtain the weighted centrality of each cell according to the weighted neighborhood matrix;

[0048] Each weighted centrality was normalized separately to obtain the vitality index of each cell;

[0049] The vitality index of each cell is embedded in the multi-scale metabolic space, and the multi-scale metabolic space embedding coordinates are obtained by combining local linear embedding and global manifold learning;

[0050] The activity distribution map was drawn using data visualization tools based on the multi-scale metabolic space embedding coordinates.

[0051] Optionally, obtaining a purity index and a purity distribution graph based on a weighted neighborhood graph guided by the metabolic network Laplace matrix includes:

[0052] Obtaining the flux of each cell in each metabolic pathway according to the weighted neighborhood graph guided by the Laplacian matrix of the metabolic network;

[0053] The total metabolic flux of each cell was obtained based on the flux of each cell in each metabolic pathway;

[0054] Calculate the flux proportion of each cell in each metabolic pathway based on the total metabolic flux of each cell;

[0055] The initial purity index of each cell was calculated based on the flux proportion of each metabolic pathway;

[0056] Normalizing the initial purity index to obtain the purity index of the cell;

[0057] The purity index of each cell is embedded in the multi-scale metabolic space, and the multi-scale purity space embedding coordinates are obtained by combining local linear embedding and global manifold learning;

[0058] The purity distribution map was plotted using data visualization tools based on the multi-scale purity space embedding coordinates.

[0059] The present application also provides a system for detecting cell viability and purity in chimeric antigen receptor T cell therapy, which comprises:

[0060] A flow cytometer, wherein the flow cytometer is used to detect the stained cell suspension using a flow cytometer to obtain a fluorescent signal;

[0061] The viability and purity analysis module is used to perform cell viability detection and cell purity analysis based on the fluorescence signal, thereby obtaining cell viability information and cell purity information.

[0062] The cell viability and purity detection method in chimeric antigen receptor T cell therapy of the present application comprehensively considers multiple aspects of the cell metabolic network, including metabolic flux, metabolic pathway, metabolic network structure, etc., and can more comprehensively reflect cell viability and purity. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 4 is a flow chart of a method for detecting cell viability and purity in chimeric antigen receptor T cell therapy according to one embodiment of the present application.

[0064] Figure 2 Schematic diagram showing the concentration of each metabolite.

[0065] Figure 3 Schematic representation of the mean and standard deviation for each metabolite.

[0066] Figure 4 A schematic diagram of the normalized data. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solutions and advantages of the implementation of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below in conjunction with the drawings in the embodiments of this application. In the drawings, the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The described embodiments are part of the embodiments of this application, not all of the embodiments. The embodiments described below with reference to the drawings are exemplary and are intended to be used to explain this application, and should not be understood as limitations on this application. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. The embodiments of this application are described in detail below in conjunction with the drawings.

[0068] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be understood as limiting the scope of protection of this application.

[0069] like Figure 1 The cell viability and purity assays shown for chimeric antigen receptor T cell therapy include:

[0070] Obtain CAR-T cell samples to be tested;

[0071] Staining the CAR-T cell sample to be tested to obtain a stained cell suspension;

[0072] The stained cell suspension is detected using a flow cytometer to obtain a fluorescent signal;

[0073] Cell viability detection and cell purity analysis are performed based on the fluorescence signal to obtain cell viability information and cell purity information.

[0074] In this embodiment, the CAR-T cell sample to be tested is stained to obtain a stained cell suspension using the following method:

[0075] The reagents used are as follows: Viability dye: 7-AAD (a membrane-impermeable nucleic acid dye, final concentration 5 μg / mL) or Fixable Viability Dye (e.g., eFluor 780). Purity-labeled antibodies: CD3-APC (for T cell recognition), CD4-PerCP-Cy5.5 (for helper T cells), and CD8-BV421 (for cytotoxic T cells). CAR detection reagent: PE-labeled target antigen protein (e.g., PE-CD19) or anti-idiotypic antibody (e.g., R19M-PE).

[0076] The processing process is as follows:

[0077] Cell staining: add 100 μL cell suspension (containing 2×10 5 cells), add:

[0078] 7-AAD (5 μL / tube, incubate at 4°C in the dark for 15 minutes), CD3-APC, CD4-PerCP-Cy5.5, CD8-BV421 (diluted 1:50, incubated at 4°C in the dark for 30 minutes), PE-CD19 or R19M-PE (diluted 1:20, incubated at 4°C in the dark for 30 minutes), and the cells were washed twice to remove unbound antibodies and dyes.

[0079] Finally, the stained cell suspension is obtained: containing viability, purity and CAR-marked cell populations.

[0080] In this embodiment, the cell viability detection and cell purity analysis based on the fluorescence signal to obtain cell viability information and cell purity information includes:

[0081] performing enzyme-substrate coupling kinetics-inspired nonlinear fluorescence signal enhancement processing on the fluorescence signal, thereby obtaining an enhanced fluorescence signal matrix;

[0082] A domain graph is constructed based on the enhanced fluorescence signal matrix to obtain a weighted neighborhood graph guided by the Laplacian matrix of the metabolic network.

[0083] Obtaining an activity index and an activity distribution map according to a weighted neighborhood graph guided by the Laplacian matrix of the metabolic network;

[0084] A purity index and a purity distribution graph are obtained according to a weighted neighborhood graph guided by the Laplace matrix of the metabolic network.

[0085] In this embodiment, performing enzyme-substrate coupling kinetics-inspired nonlinear fluorescence signal enhancement processing on the fluorescence signal to obtain an enhanced fluorescence signal matrix includes:

[0086] The final signal intensity of each cell i was calculated using the following formula:

[0087]

[0088] in, represents the enhanced fluorescence signal, representing the final signal intensity of cell i in the c fluorescence channel; x i,c is the original fluorescence signal, representing the initial signal intensity of cell i in the c fluorescence channel; V max,c K is the maximum enhancement rate of fluorescence channel c, which represents the theoretical maximum efficiency of the enzyme-catalyzed reaction (unit: fluorescence intensity / unit time); M,c K is the Michaelis constant of fluorescence channel c, reflecting the affinity between enzyme and substrate (unit: fluorescence intensity). M,c The smaller the value, the stronger the binding ability between enzyme and substrate; inh,c is the inhibitory factor of fluorescence channel c, representing substances that inhibit enzyme activity in cells (such as antioxidants), estimated by inversion of CellROX signal (unit: fluorescence intensity); K i,c is the inhibition constant of the fluorescence channel c, which indicates the inhibitory intensity of the inhibitory factor on the enzyme activity (unit: fluorescence intensity); K i,c The smaller the value, the stronger the inhibitory effect; c is the weight coefficient of the time domain convolution term of the fluorescence channel c, which controls the contribution ratio of the historical signal to the current enhancement (dimensionless); β c is the time domain attenuation coefficient of the fluorescence channel c, which indicates the decay rate of the historical signal over time (unit: 1 / time); γ c is the nonlinear suppression coefficient of the fluorescence channel c, which controls the dynamic suppression strength of the high signal area (unit: fluorescence intensity -2 );δ c is the second-order kinetic coefficient of fluorescence channel c, which indicates the strength of the cascade amplification effect of the enzymatic reaction (unit: fluorescence intensity / time2);

[0089] The final signal intensity of each of the cells constitutes the enhanced fluorescence signal matrix.

[0090] In this embodiment, the original fluorescence signal matrix has a dimension of (number of cells × number of fluorescence channels), for example (10 6 ×3) matrix containing the raw fluorescence signals of PI, CellROX, and CD3-FITC.

[0091] In this embodiment, V max,c , K M,c It can be calibrated by enzyme kinetics experiments. For example, under conditions of known substrate concentration, the fluorescence signal enhancement rate is measured and fitted to obtain V max,c , K M,c .

[0092] In this embodiment, S inh,c , Ki,c Calibration is performed through cell experiments. For example, when adding a known concentration of an inhibitory factor (such as an antioxidant), the change in the fluorescence signal enhancement amplitude is measured and fitted to obtain S inh,c , K i,c .

[0093] α, β, γ, and δ can be set as needed, for example, by cross-validation optimization. For example, while retaining a portion of the experimental data as a validation set, the parameters are adjusted to maximize the signal-to-noise ratio of the enhanced signal on the validation set.

[0094] The present application simulates the saturation characteristics of enzyme catalysis reaction by the above formula. i ) is low, the reaction rate increases linearly with the substrate concentration; when the substrate concentration is high, the reaction rate approaches V max , showing a saturation effect. And introduce a non-competitive inhibition mechanism, when the inhibitory factor (S inh ) concentration increases, the enzyme activity is inhibited, and the fluorescence signal enhancement amplitude decreases. And the historical dependence of the simulated signal over time is increased, and the historical signal (x i The current enhancement effect is influenced by an exponential decay factor and a nonlinear inhibitory factor. High signal regions experience stronger inhibitory effects. An acceleration term is introduced to simulate the cascade amplification effect in enzymatic reactions. As the rate of signal change increases, the enhancement effect is further amplified.

[0095] The formula in this application enables the acquisition of an enhanced fluorescence signal matrix with dimensions (number of cells × number of fluorescence channels), a signal-to-noise ratio improvement of ≥45%, and a signal dynamic range extended to [0, 3.5], providing higher-quality input for subsequent analysis.

[0096] In this embodiment, constructing a domain graph based on the enhanced fluorescence signal matrix to obtain a weighted neighborhood graph guided by the Laplacian matrix of the metabolic network includes:

[0097] A metabolic network is constructed based on a domain graph constructed based on the enhanced fluorescence signal matrix. The metabolic network includes nodes and edge information between each node, wherein the nodes represent metabolites and the edges represent metabolic fluxes. Specifically, the nodes of the metabolic network represent metabolites (such as glucose, ATP, NADH, etc.), which are key molecules in the cellular metabolic process and are closely related to cell viability, purity, and function. The edges of the metabolic network represent metabolic flux, that is, the conversion rate between metabolites. The metabolic flux is estimated by inversion of the fluorescence signal, establishing a bridge between the fluorescence signal and the metabolic network.

[0098] In this example, the correlation between the fluorescence signal and the metabolic network is as follows:

[0099] PI signal: reflects the integrity of the cell membrane and is related to the membrane metabolic flux. Cell membrane damage may lead to metabolite leakage and affect the metabolic network.

[0100] CellROX signaling: reflects oxidative stress and is related to antioxidant metabolic flux. Oxidative stress changes the redox state of metabolites and affects the metabolic network.

[0101] CD3-FITC signal: reflects the purity of T cells and is associated with immune-related metabolic flux. The immune function of T cells depends on specific metabolic pathways.

[0102] Obtaining a Laplacian matrix and a modularity matrix according to the metabolic network;

[0103] Obtaining a metabolic flux similarity matrix according to the enhanced fluorescence signal matrix, the Laplace matrix, and the modular matrix;

[0104] A weighted neighborhood graph guided by a Laplacian matrix of a metabolic network is generated according to the metabolic flux similarity matrix.

[0105] In this embodiment, obtaining a Laplacian matrix and a modular matrix according to the metabolic network includes:

[0106] Determine the metabolite or metabolic pathway corresponding to each fluorescence channel based on the enhanced fluorescence signal matrix;

[0107] For example, PI signaling is related to membrane metabolism and may involve the phospholipid metabolism pathway.

[0108] CellROX signaling: related to oxidative stress and may involve antioxidant metabolic pathways (such as glutathione metabolism).

[0109] CD3-FITC signal: related to T cell immune metabolism, and may involve immune-related metabolic pathways such as glycolysis and TCA cycle.

[0110] The metabolic flux is estimated by inversely estimating the change in fluorescence signal; this can be obtained by the following methods:

[0111] Kinetic model: Establish a kinetic model between fluorescence signal and metabolic flux. For example, if the fluorescence signal intensity is proportional to the metabolic flux, the metabolic flux can be estimated by the rate of change of the fluorescence signal.

[0112] Machine Learning Model: Train a machine learning model (e.g., linear regression, support vector machine, etc.) with fluorescence signals as input and metabolic flux as output. This requires training data with known metabolic flux to calibrate the model.

[0113] The weight matrix of the metabolic network is constructed based on the estimated results of metabolic flux; the specific construction method is as follows:

[0114] According to the metabolic flux estimation results, the weight matrix W of the metabolic network is constructed. If there is a metabolic flux from the i-th metabolite to the j-th metabolite, then W ij is the intensity of the flux, each W ij constitutes the weight matrix W; otherwise it is 0.

[0115] Obtain the Laplace matrix based on the weight matrix; specifically, first calculate the degree matrix D, as follows:

[0116] D ij =∑ j W ij Among them, W ij is the intensity of the metabolic flux from the i-th metabolite to the j-th metabolite;

[0117] And the Laplace matrix is ​​obtained according to the degree matrix, as follows:

[0118] L = DW; where L is the Laplace matrix, D is the degree matrix, and W is the weight matrix;

[0119] A modularity matrix is ​​obtained according to the metabolic network.

[0120] The obtaining of a modular matrix according to the metabolic network comprises:

[0121] The metabolic network is divided into multiple functional modules by community detection algorithms (such as the Louvain algorithm). Specifically, the Louvain algorithm divides the network modules by optimizing the modularity Q, which is defined as:

[0122] Among them, A ij is the adjacency matrix of the metabolic network. If there is a metabolic flux edge between the i-th metabolite and the j-th metabolite, then A ij =1, otherwise A ij =0;k i represents the degree of the i-th metabolite (i.e., the number of edges connected to metabolite i); m represents the total number of edges in the metabolic network; c i Indicates the functional module to which the i-th metabolite belongs; δ(c i ,c j ) is the Kronecker function, if c i =c j , then δ(c i ,c j )=1, otherwise δ(c i ,c j )=0.

[0123] Each module contains a set of metabolite nodes, and the node connectivity within the module is higher than the node connectivity between modules;

[0124] Specific process:

[0125] Initialization: Treat each metabolite node as an independent module;

[0126] Iterative optimization: randomly select a metabolite node and try to move it to an adjacent module, calculate the change in modularity Q, and if Q increases, keep the move; otherwise, undo the move;

[0127] Termination condition: When the modularity Q cannot be further increased, the algorithm terminates and outputs the final module partitioning result.

[0128] According to the functional module division results, an initial modular matrix is ​​constructed; in this embodiment, the element B of the modular matrix B kl represents the connectivity between the kth module and the lth (L)th module, which is defined as:

[0129] B kl =∑ i∈k,j∈l A ij ;

[0130] In this embodiment, the diagonal element B kk represents the connectivity within the kth module (i.e., the number of edges within the module).

[0131] Off-diagonal elements B kl (k≠l) represents the connectivity between the kth module and the lth (L)th module (i.e., the number of edges between modules).

[0132] The constructed initial modularity matrix is ​​normalized to obtain the modularity matrix.

[0133] In this embodiment, obtaining a metabolic flux similarity matrix according to the enhanced fluorescence signal matrix, the Laplace matrix, and the modular matrix includes:

[0134] Gaussian similarity is calculated for every two cells to obtain Gaussian similarity;

[0135] In this embodiment, the Gaussian similarity is obtained by the following formula:

[0136] in, represents the enhanced fluorescence signal, which represents the final signal intensity of the i-th cell in the c fluorescence channel; represents the final signal intensity of the jth cell in the c fluorescence channel; represents the Gaussian similarity between the i-th cell and the j-th cell; C represents the fluorescence channel; σ c represents the noise standard deviation of fluorescence channel c, obtained through experimental calibration.

[0137] Calculate the Laplacian regularization term according to the metabolic state vectors of every two cells to obtain the Laplacian regularization term;

[0138] In this embodiment, the Laplacian regularization term is obtained through the following formula:

[0139] Wherein, represents the Laplacian regularization term between the i-th cell and the j-th cell; λ is the Laplacian regularization coefficient, which is an adjustable parameter used to control the influence of the Laplacian matrix on the similarity; u i represents the metabolic state vector of the i-th cell, obtained through PCA dimensionality reduction; u j represents the metabolic state vector of the j-th cell, obtained through PCA dimensionality reduction; L represents the Laplacian matrix; represents the transpose of the vector u[[ID=1 6]] i ;

[0140] In this embodiment, the metabolic state vector of the i-th cell, obtained through PCA dimensionality reduction, can be obtained through the following manner:

[0141] Suppose we have n cells, and the concentration of m metabolites is measured for each cell, forming an n×m metabolite concentration matrix X. Through PCA dimensionality reduction, the data is reduced to p dimensions (p < m) to obtain the metabolic state vector of each cell.

[0142] Perform standardization processing on the metabolite concentration matrix X so that the mean of each metabolite is 0 and the standard deviation is 1.

[0143] Refer to Figure 2 and assume that we have 3 cells (n = 3) and measure the concentration of 4 metabolites (m = 4). Calculate the mean and standard deviation of each metabolite (as shown in Figure 3 ), and the standardized data is as shown in Figure 4 ; [[ID= 34]]

[0144] Calculate the covariance matrix C of the standardized metabolite concentration matrix. The formula is as follows:

[0145] Wherein, X std is the standardized metabolite concentration matrix; is the transpose matrix of X std ;

[0146] Perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues λ1, λ2,..., λ m and the corresponding eigenvectors v1, v2,..., v m ;

[0147] According to the size of the eigenvalue, the eigenvectors corresponding to the first p largest eigenvalues ​​are selected as the principal components, and the standardized metabolite concentration matrix X std Projecting onto the selected principal components, we obtain the metabolic state vector after dimensionality reduction.

[0148] The specific formula is as follows:

[0149] U=X std V p ; Where U is the metabolic state vector matrix after dimensionality reduction, each row represents a cell, and each column represents a principal component; V p is a matrix consisting of the first p principal components (eigenvectors) selected.

[0150] Get the modularity constraint term based on the modularity membership vector of every two cells;

[0151] In this embodiment, the modular constraint term is obtained by the following formula:

[0152] in, represents the modular constraint term between the i-th cell and the j-th cell; k represents the modular constraint coefficient, which is an adjustable parameter used to control the impact of the modular matrix on similarity; V i represents the modular membership vector of the i-th cell, describing the cell's membership in the metabolic network module; V j represents the modular membership vector of the jth cell, describing the cell's membership in the metabolic network module; B is the modular matrix; V i T Represents vector V i The transpose of .

[0153] The Gaussian similarity, the Laplace regularization term, and the modularity constraint term are multiplied to obtain metabolic flux similarity, wherein the metabolic flux similarities of each cell constitute the metabolic flux similarity matrix.

[0154] For example, the similarity of the comprehensive metabolic flux between the i-th cell and the j-th cell is obtained by the following formula:

[0155] Among them, Φ ij is the similarity of the comprehensive metabolic flux between the i-th cell and the j-th cell; represents the modularity constraint between the i-th cell and the j-th cell; represents the Laplace regularization term between the i-th cell and the j-th cell; represents the Gaussian similarity between the i-th cell and the j-th cell;

[0156] In this embodiment, generating a metabolic network Laplacian matrix-guided weighted neighborhood graph according to the metabolic flux similarity matrix includes:

[0157] For each cell, the K cells with the highest similarity to its metabolic flux are selected as neighbors through the metabolic flux similarity matrix:

[0158] in,

[0159] The similarity is converted into neighborhood weight through the metabolic weight function. The specific formula is as follows:

[0160]

[0161] in, is the neighborhood weight between the i-th cell and the j-th cell; Φ ij is the similarity of the comprehensive metabolic flux between the i-th cell and the j-th cell; μ is the amplitude regulation coefficient; I ext is the intensity of external stimulus (such as drug concentration); K I is the inhibition constant; is the metabolic network neighborhood set of the i-th cell.

[0162] In this embodiment, obtaining the vitality index and vitality distribution map according to the weighted neighborhood graph guided by the metabolic network Laplacian matrix includes:

[0163] Obtain metabolic flux dynamic data of each cell and a metabolic pathway database; in this embodiment, the metabolic pathway database (such as KEGG) provides us with prior knowledge of metabolites and metabolic reactions, which guides us to understand the conversion relationship between metabolites.

[0164] In this embodiment, metabolic flux dynamic data can be obtained by:

[0165] Through experimental design, fluorescence signals can be associated with specific metabolic fluxes. For example, PI signals are associated with membrane metabolic fluxes, and CellROX signals are associated with antioxidant metabolic fluxes.

[0166] Metabolic flux dynamic data acquisition:

[0167] Experimental Design: During the experiment, we collected fluorescence signals of cells at multiple time points (such as 0 hours, 2 hours, 4 hours, 6 hours, etc.). To ensure the reliability of the data, the experiment was repeated at each time point.

[0168] Data acquisition: Using a flow cytometer or high-content imaging system, measure the signal intensity of cells in specific fluorescent channels at each time point. This signal intensity data will be used for subsequent analysis.

[0169] Data processing:

[0170] Denoising: Denoising is performed on the original fluorescence signal to eliminate random errors in the experimental process.

[0171] Normalization: The denoised fluorescence signal is normalized to eliminate the systematic errors between different experimental batches.

[0172] Time alignment: Due to experimental manipulation or other factors, the time points between different cells may not be completely aligned. We use the dynamic time warping (DTW) algorithm to align the cell metabolic flux time series data to ensure the consistency of the time dimension.

[0173] Metabolic flux inversion estimation: Using the processed fluorescence signals, we inversely estimate metabolic flux using kinetic or machine learning models. These models take the fluorescence signals as input and output estimated metabolic fluxes. Ultimately, we obtain metabolic flux dynamics for all cells (number of cells × number of time points × number of metabolites), recording changes in metabolite concentrations at different time points.

[0174] The time series similarity of metabolic flux between every two cells is calculated based on the metabolic flux dynamic data of each cell after alignment;

[0175] It can be obtained using the following formula:

[0176] in, represents the dynamic time warping similarity between the i-th cell and the j-th cell based on the metabolic flux time series; f i (t) is the metabolic flux time series of the i-th cell, which records the changes in metabolite concentrations of the cell at different time points; f j (t) is the metabolic flux time series of the jth cell, which records the changes in metabolite concentrations of the cell at different time points. i (t),f j (t)) is the dynamic time warping distance, which is used to quantify the similarity between two time series, even if they are stretched or offset on the time axis; σ DTW is the width of the dynamic similarity kernel, which controls the decay rate of similarity with the dynamic time-warped distance.

[0177] The time series similarity of metabolic fluxes between cells is incorporated into the neighborhood matrix to obtain a weighted domain matrix;

[0178] In this embodiment, the dynamic time warping similarity Incorporate the neighborhood matrix W meta , construct a dynamic weighted neighborhood matrix W dynamic :

[0179] In this embodiment, the vitality index V i Defined as the neighborhood matrix W meta The weighted centrality of is:

[0180]

[0181] Among them, V i It represents the vitality index of the i-th cell, which comprehensively reflects the local metabolic activity and global regulatory ability of the cell. represents the neighborhood weight between the i-th cell and the j-th cell; N represents the total number of cells; λ is the neighborhood matrix W meta The maximum eigenvalue of is used to normalize the eigenvector centrality to ensure that the value of the eigenvector centrality is within a reasonable range; V j is the neighborhood matrix W meta The element in the eigenvector corresponding to the maximum eigenvalue λ represents the eigenvector centrality of cell j; β is the weight coefficient of betweenness centrality, which is used to control the contribution of global regulatory ability (betweenness centrality) to the vitality index. By adjusting the value of β, the effects of local metabolic activity and global regulatory ability on the vitality index can be balanced; σ st is the number of the shortest metabolic pathways from the sth cell to the tth cell in the metabolic network, indicating the number of the shortest paths from cell s to cell t, which reflects the closeness of the connection between cells s and t in the metabolic network; st (i) is the number of shortest metabolic pathways from cell s to cell t that pass through cell i in the metabolic network, indicating the importance of cell i in the shortest path from cell s to cell t. If cell i is located on multiple shortest paths from s to t, then σ st The value of (i) will be larger, indicating that cell i has a higher global regulatory ability in the metabolic network.

[0182] In this embodiment, the domain matrix is ​​constructed by using the metabolic flux similarity matrix and the number of neighbors. This is a prior art and will not be described in detail here.

[0183] Obtain the weighted centrality of each cell according to the weighted neighborhood matrix;

[0184] The step of obtaining the weighted centrality of each cell according to the weighted community matrix includes:

[0185] Dynamic weighted neighborhood matrix W dynamic Perform row normalization to obtain the transition probability matrix P dynamic .

[0186] Solve the transition probability matrix P by power iteration method dynamic The maximum eigenvalue λ dynamic and its corresponding eigenvector V dynamic .

[0187] Calculate betweenness centrality B i , and fuse it with the eigenvector centrality through the weight coefficient β to obtain the weighted centrality V i .

[0188] In this example, this part of the weighted centrality formula is the betweenness centrality B i The calculation method of weighted centrality formula is λ·W meta V is the eigenvector centrality.

[0189] Normalize each weighted centrality separately to obtain the vitality index V of each cell i norm ;

[0190] The vitality index of each cell is embedded in the multi-scale metabolic space, and the multi-scale metabolic space embedding coordinates are obtained by combining local linear embedding and global manifold learning;

[0191] In this embodiment, the multi-scale metabolic space embedding is performed on the vitality index of each cell, and the multi-scale metabolic space embedding coordinates are obtained by combining local linear embedding and global manifold learning, including:

[0192] Neighbor selection: For each cell, select its K nearest neighbor cells (based on the vitality index V norm similarity).

[0193] Weight calculation: Calculate the reconstruction weight between each cell i and its neighbor cells so that the cell can be represented as a linear combination of its neighbor cells.

[0194] Low-dimensional embedding: By preserving the reconstruction weights, cells are mapped to a low-dimensional space (such as 2D or 3D) while preserving the local structure.

[0195] Global Structure Preservation (Global Manifold Learning, ISOMAP), ISOMAP is a global manifold learning method that aims to preserve the global structure in the dataset. It captures the global geometric properties of the data by constructing an adjacency graph of the data and calculating the shortest paths between nodes.

[0196] The specific steps are as follows:

[0197] Adjacency graph construction: Based on the similarity between cells (such as the vitality index V norm ), and construct the adjacency graph of the cells.

[0198] Shortest path calculation: Calculate the shortest path between each pair of cells in the adjacency graph, which reflects the distance between cells in the global metabolic space.

[0199] Multidimensional scaling: Through the multidimensional scaling (MDS) algorithm, cells are mapped to a lower-dimensional space (such as 2D or 3D) while preserving the global structure.

[0200] The results of LLE and ISOMAP are fused to preserve both the local and global structures of the metabolic network.

[0201] The specific steps are as follows:

[0202] Result alignment: Align the low-dimensional embedding results of LLE and ISOMAP to ensure that they are in the same low-dimensional space.

[0203] Weighted Fusion: The results of LLE and ISOMAP are fused into a comprehensive low-dimensional embedding representation through weighted averaging. The weighting coefficient can be adjusted according to specific needs to balance the importance of local and global structures.

[0204] The final output is as follows: Multi-scale metabolic space embedding coordinates (number of cells × 2 or number of cells × 3): Contains the coordinates of each cell in the low-dimensional metabolic space, which simultaneously preserves the local and global structures of the metabolic network.

[0205] The activity distribution map was drawn using data visualization tools based on the multi-scale metabolic space embedding coordinates.

[0206] In this embodiment, drawing a vitality distribution map using a data visualization tool based on the multi-scale metabolic space embedding coordinates includes:

[0207] According to the normalized vitality index vector V norm Assign a color to each cell based on the value of . Typically, you might choose red to represent high viability, blue to represent low viability, and a color gradient to represent intermediate viability values.

[0208] Use a data visualization tool (such as Matplotlib) to plot the vitality distribution:

[0209] If two-dimensional embedding coordinates are used, a scatter plot is drawn with the horizontal and vertical axes representing the first and second principal components of the multi-scale metabolic spatial embedding, respectively.

[0210] If three-dimensional embedding coordinates are used, a three-dimensional scatter plot is drawn, with the horizontal axis, vertical axis, and vertical axis representing the first principal component, second principal component, and third principal component of the multi-scale metabolic spatial embedding, respectively.

[0211] In the scatter plot, each dot represents a cell, and its color is mapped according to the value of the normalized viability index vector.

[0212] In this embodiment, obtaining the purity index and the purity distribution graph according to the weighted neighborhood graph guided by the metabolic network Laplacian matrix includes:

[0213] Obtaining the flux of each cell in each metabolic pathway according to the weighted neighborhood graph guided by the Laplacian matrix of the metabolic network;

[0214] For example, assuming there are N metabolic pathways in the metabolic network, for the i-th cell, the flux of its participation in metabolic pathway k is J ik .

[0215] The total metabolic flux of each cell is obtained based on the flux of each cell in each metabolic pathway; specifically, the total metabolic flux of the i-th cell is

[0216] Specifically, the flux share of each cell in each metabolic pathway is calculated based on the total metabolic flux of each cell; specifically, the flux share of the i-th cell in metabolic pathway k is

[0217] The initial purity index of each cell is calculated based on the flux proportion of each metabolic pathway; specifically, the purity index P of the i-th cell is i Defined as the flux proportion p ik The negative value of the entropy, that is The smaller the entropy value, the more concentrated the distribution of cells in the metabolic pathway and the higher the purity index.

[0218] Normalizing the initial purity index to obtain the purity index of the cell;

[0219] The purity index of each cell is embedded in a multi-scale metabolic space, and the multi-scale purity space embedding coordinates are obtained by combining local linear embedding and global manifold learning (this step is similar to the vitality index visualization method mentioned above and will not be repeated here);

[0220] According to the multi-scale purity space embedding coordinates, a data visualization tool is used to draw a purity distribution map (this step is similar to the vitality index visualization method mentioned above and will not be repeated here).

[0221] The present application also provides a cell viability and purity detection system for chimeric antigen receptor T cell therapy, wherein the cell viability and purity detection system for chimeric antigen receptor T cell therapy includes a flow cytometer and a viability and purity analysis module, wherein:

[0222] The flow cytometer is used to detect the stained cell suspension using a flow cytometer to obtain a fluorescent signal;

[0223] The viability and purity analysis module is used to perform cell viability detection and cell purity analysis based on fluorescence signals, thereby obtaining cell viability information and cell purity information.

[0224] This application has the following advantages:

[0225] Through enzyme-substrate coupling kinetics-inspired nonlinear fluorescence signal enhancement processing, the signal-to-noise ratio is improved by ≥45%, making the fluorescence signal clearer, reducing the interference of background noise, and being able to capture a wider range of fluorescence signal changes, providing higher-quality input data for subsequent analysis.

[0226] By designing a metabolic network Laplacian matrix, we can capture the global structural information of the metabolic network and provide a foundation for the construction of a neighborhood graph. The Laplacian matrix can quantify the strength of associations between metabolites and reflect the overall topological properties of the metabolic network. In metabolic network analysis, the Laplacian matrix helps understand the interactions between metabolites and provides a structural foundation for subsequent neighborhood graph construction and vitality index calculation.

[0227] Metabolic flux similarity calculations comprehensively consider fluorescence signals, metabolic network structure, and functional module information to more accurately reflect metabolic flux similarity between cells. Multiplying Gaussian similarity, Laplace regularization, and modularity constraints achieves multi-dimensional information fusion. The Laplace regularization term incorporates the structural information of the metabolic network, making the similarity calculation more accurate to the actual metabolic process. The modularity constraint considers the functional module information of the metabolic network, helping to identify cell populations with similar metabolic functions.

[0228] By designing the application of Dynamic Time Warping (DTW), we can address temporal sampling bias in time series data and improve the accuracy of data alignment. DTW similarity captures the dynamic changes in metabolic fluxes, enabling the neighborhood matrix to reflect the temporal dimension of the metabolic network. DTW similarity can quantify the dynamic similarity of metabolic fluxes between cells, providing a basis for dynamic updates of the neighborhood matrix.

[0229] The neighborhood matrix can reflect the temporal dimension of the metabolic network, more closely reproducing the actual metabolic process. By incorporating DTW similarity, the neighborhood matrix not only considers the structural information of the metabolic network but also the dynamic changes in metabolic flux. By combining the structural information of the metabolic network with the dynamic information of metabolic flux, the neighborhood matrix can more comprehensively reflect the metabolic state of the cell.

[0230] This application uses weighted centrality as a vitality index, and simultaneously quantifies the local influence and global regulatory ability of cells, providing a more comprehensive assessment of cell metabolic vitality. By fusing eigenvector centrality and betweenness centrality, the vitality index can comprehensively reflect the importance of cells in the metabolic network. Eigenvector centrality reflects the local influence of cells in the metabolic network, that is, the influence of cells directly connected to it on it. Betweenness centrality reflects the global regulatory ability of cells in the metabolic network, that is, its key role in information transmission and material transport. Through weighted centrality, the local and global importance of cells in the metabolic network can be comprehensively evaluated, providing an important basis for cell screening and functional analysis.

[0231] This application combines local linear embedding (LLE) and global manifold learning (ISOMAP) through multi-scale metabolic space embedding to retain the local and global structures of the metabolic network. Through multi-scale embedding, the complex patterns of the metabolic network can be revealed, providing richer information for the generation of vitality distribution maps. LLE can retain the local structural information of the data and is suitable for describing the local metabolic environment of the metabolic network. ISOMAP can reveal the global manifold structure of the data and help understand the overall topological characteristics of the metabolic network. By combining LLE and ISOMAP, multi-scale metabolic space embedding can simultaneously retain the local and global structural information of the metabolic network, providing a more comprehensive perspective for the generation of vitality distribution maps.

[0232] Calculating the purity index based on the entropy of metabolic pathway flux ratios quantifies the concentration of cells along metabolic pathways. A higher purity index indicates a more concentrated distribution of cells along metabolic pathways and a higher cell purity. This metric objectively reflects cell purity and, by calculating the purity index, allows for objective assessment of cell purity, providing an important basis for quality control of CAR-T cell therapies.

[0233] Finally, it should be pointed out that the above embodiments are intended only to illustrate the technical solutions of this application and are not intended to limit them. Although this application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they may modify the technical solutions described in the above embodiments or replace some of the technical features therein with equivalents; and such modifications or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for detecting cell viability and purity in chimeric antigen receptor T cell therapy, characterized in that: The method for detecting cell viability and purity in the chimeric antigen receptor T cell therapy includes: Obtain CAR-T cell samples to be tested; Staining the CAR-T cell sample to be tested to obtain a stained cell suspension; The stained cell suspension is detected using a flow cytometer to obtain a fluorescent signal; Cell viability detection and cell purity analysis are performed based on the fluorescence signal to obtain cell viability information and cell purity information.

2. The method for detecting cell viability and purity in chimeric antigen receptor T cell therapy according to claim 1, wherein: The cell viability detection and cell purity analysis based on the fluorescence signal to obtain cell viability information and cell purity information includes: performing enzyme-substrate coupling kinetics-inspired nonlinear fluorescence signal enhancement processing on the fluorescence signal, thereby obtaining an enhanced fluorescence signal matrix; A domain graph is constructed based on the enhanced fluorescence signal matrix to obtain a weighted neighborhood graph guided by the Laplacian matrix of the metabolic network. Obtaining an activity index and an activity distribution map according to a weighted neighborhood graph guided by the Laplace matrix of the metabolic network; A purity index and a purity distribution graph are obtained according to a weighted neighborhood graph guided by the Laplace matrix of the metabolic network.

3. The method for detecting cell viability and purity in chimeric antigen receptor T cell therapy according to claim 2, wherein: The performing enzyme-substrate coupling kinetics heuristic nonlinear fluorescence signal enhancement processing on the fluorescence signal to obtain an enhanced fluorescence signal matrix includes: The final signal intensity of each cell i was calculated using the following formula: in, represents the enhanced fluorescence signal, representing the final signal intensity of cell i in the c fluorescence channel; x i,c is the original fluorescence signal, representing the initial signal intensity of cell i in the c fluorescence channel; V max,c is the maximum enhancement rate of fluorescence channel c, indicating the theoretical maximum efficiency of the enzyme-catalyzed reaction; K M,c is the Michaelis constant of fluorescence channel c, reflecting the affinity between enzyme and substrate, K M,c The smaller the value, the stronger the binding ability between enzyme and substrate; inh,c is the inhibitory factor of fluorescence channel c, representing the substance that inhibits enzyme activity in the cell, estimated by inversion of CellROX signal; K i,c is the inhibition constant of fluorescence channel c, which indicates the inhibitory intensity of the inhibitory factor on enzyme activity; K i,c The smaller the value, the stronger the inhibitory effect; c is the weight coefficient of the time domain convolution term of the fluorescence channel c, which controls the contribution ratio of the historical signal to the current enhancement; β c is the time domain attenuation coefficient of the fluorescence channel c, which indicates the decay rate of the historical signal over time; γ c is the nonlinear suppression coefficient of the fluorescence channel c, which controls the dynamic suppression strength of the high signal area; c is the second-order kinetic coefficient of fluorescence channel c, which indicates the strength of the cascade amplification effect of the enzymatic reaction; The final signal intensity of each of the cells constitutes the enhanced fluorescence signal matrix.

4. The method for detecting cell viability and purity in chimeric antigen receptor T cell therapy according to claim 3, wherein: The constructing of a domain graph based on the enhanced fluorescence signal matrix to obtain a weighted neighborhood graph guided by the Laplacian matrix of the metabolic network includes: Constructing a metabolic network based on a domain graph of the enhanced fluorescence signal matrix, wherein the metabolic network includes nodes and edge information between each node, wherein the nodes represent metabolites and the edges represent metabolic fluxes; Obtaining a Laplacian matrix and a modularity matrix according to the metabolic network; Obtaining a metabolic flux similarity matrix according to the enhanced fluorescence signal matrix, the Laplace matrix, and the modular matrix; A weighted neighborhood graph guided by a Laplacian matrix of a metabolic network is generated according to the metabolic flux similarity matrix.

5. The method for detecting cell viability and purity in chimeric antigen receptor T cell therapy according to claim 4, wherein: The obtaining of the Laplace matrix and the modular matrix according to the metabolic network includes: Determine the metabolite or metabolic pathway corresponding to each fluorescence channel based on the enhanced fluorescence signal matrix; The changes in fluorescence signals were used to inversely estimate metabolic fluxes; Constructing the weight matrix of the metabolic network based on the estimated results of metabolic flux; Get the Laplacian matrix based on the weight matrix; A modularity matrix is ​​obtained according to the metabolic network.

6. The method for detecting cell viability and purity in chimeric antigen receptor T cell therapy according to claim 5, wherein: The obtaining of a modular matrix according to the metabolic network comprises: The metabolic network is divided into multiple functional modules using a community detection algorithm. Each module contains a group of metabolite nodes, and the node connectivity within the module is higher than the node connectivity between modules. According to the functional module division results, construct the initial modular matrix; The constructed initial modularity matrix is ​​normalized to obtain the modularity matrix.

7. The method for detecting cell viability and purity in chimeric antigen receptor T cell therapy according to claim 6, wherein: The step of obtaining a metabolic flux similarity matrix according to the enhanced fluorescence signal matrix, the Laplace matrix, and the modular matrix includes: Gaussian similarity is calculated for every two cells to obtain Gaussian similarity; The Laplace regularization term is calculated based on the metabolic state vectors of every two cells to obtain the Laplace regularization term; Get the modularity constraint term based on the modularity membership vector of every two cells; The Gaussian similarity, the Laplace regularization term, and the modularity constraint term are multiplied to obtain metabolic flux similarity, wherein the metabolic flux similarities of each cell constitute the metabolic flux similarity matrix.

8. The method for detecting cell viability and purity in chimeric antigen receptor T cell therapy according to claim 7, wherein: The step of obtaining the vitality index and the vitality distribution graph based on the weighted neighborhood graph guided by the metabolic network Laplace matrix includes: Obtain metabolic flux dynamic data of each cell and metabolic pathway database; Align the metabolic flux dynamic data of each cell to eliminate time sampling bias; The time series similarity of metabolic flux between every two cells is calculated based on the metabolic flux dynamic data of each cell after alignment; The time series similarity of metabolic fluxes between cells is incorporated into the neighborhood matrix to obtain a weighted domain matrix; Obtain the weighted centrality of each cell according to the weighted neighborhood matrix; Each weighted centrality was normalized separately to obtain the vitality index of each cell; The vitality index of each cell is embedded in the multi-scale metabolic space, and the multi-scale metabolic space embedding coordinates are obtained by combining local linear embedding and global manifold learning; The activity distribution map was drawn using data visualization tools based on the multi-scale metabolic space embedding coordinates.

9. The method for detecting cell viability and purity in chimeric antigen receptor T cell therapy according to claim 8, wherein: The obtaining of the purity index and the purity distribution graph according to the weighted neighborhood graph guided by the metabolic network Laplace matrix includes: Obtaining the flux of each cell in each metabolic pathway according to the weighted neighborhood graph guided by the Laplacian matrix of the metabolic network; The total metabolic flux of each cell was obtained based on the flux of each cell in each metabolic pathway; Calculate the flux proportion of each cell in each metabolic pathway based on the total metabolic flux of each cell; The initial purity index of each cell was calculated based on the flux proportion of each metabolic pathway; Normalizing the initial purity index to obtain the purity index of the cell; The purity index of each cell is embedded in the multi-scale metabolic space, and the multi-scale purity space embedding coordinates are obtained by combining local linear embedding and global manifold learning; The purity distribution map was plotted using data visualization tools based on the multi-scale purity space embedding coordinates.

10. A cell viability and purity detection system for chimeric antigen receptor T cell therapy, characterized in that: The cell viability and purity detection system in the chimeric antigen receptor T cell therapy includes: A flow cytometer, wherein the flow cytometer is used to detect the stained cell suspension using a flow cytometer to obtain a fluorescent signal; The viability and purity analysis module is used to perform cell viability detection and cell purity analysis based on the fluorescence signal, thereby obtaining cell viability information and cell purity information.

Images