Method and system for processing cytokine detection data and storage medium

Abstract

The invention relates to the technical field of data processing, in particular to a cell factor detection data processing method and system and a storage medium. The method comprises the following steps: acquiring original single cell cytokine detection data; performing cytokine secretion amount analysis on the original single cell cytokine detection data, and performing multi-cell-cytokine matrix construction to obtain a cytokine secretion characteristic matrix; carrying out secretion mode typing degree difference calculation on the cell factor secretion characteristic matrix to generate a cell dynamic heterogeneous distribution diagram; performing cell subset center tracking according to the cell dynamic heterogeneous distribution diagram to obtain cell detection subset data; and carrying out subpopulation cell clustering processing on the cytokine secretion characteristic matrix through cell detection subpopulation data to obtain cytokine subpopulation characteristic data. According to the method disclosed by the invention, the cell subpopulation is identified by finely analyzing the single cell data, so that the characteristic quantification of the cytokine secretion heterogeneity is realized.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, system and storage medium for processing cytokine detection data. Background Art

[0002] Cytokines are key molecules for intercellular communication and play a vital role in a variety of physiological and pathological processes such as immune response, inflammatory response, and tissue repair. Cell heterogeneity refers to the differences in gene expression, protein expression, metabolic state, and function between different cells in the same cell population. This heterogeneity is particularly prominent in the tumor microenvironment, immune response, stem cell differentiation, and other processes. In terms of cytokine secretion, different cells secrete different types and amounts of cytokines, and this difference reflects the functional state of the cells and the influence of the microenvironment. Therefore, studying the heterogeneity of cytokine secretion is crucial to understanding cell behavior, disease development, and drug response. However, traditional methods for processing cytokine detection data usually use statistical models or simple algorithms to process data, focusing on analyzing the overall behavioral characteristics of cell populations. Although this method can provide statistical results at the population level, it is easy to ignore the microscopic differences between cells, especially when dealing with heterogeneity between cells, there is a risk of averaging individual differences. This averaging effect not only masks the characteristic differences between cell subpopulations, but also leads to the loss of key biological information. Summary of the invention

[0003] Based on this, the present invention provides a method, system and storage medium for processing cytokine detection data to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for processing cytokine detection data comprises the following steps:

[0005] Step S1: obtaining original single-cell cytokine detection data; performing cytokine secretion analysis on the original single-cell cytokine detection data to obtain cytokine secretion vector data; constructing a multi-cell-cytokine matrix based on the cytokine secretion vector data to obtain a cytokine secretion feature matrix;

[0006] Step S2: performing cytokine secretion pattern analysis on the cytokine secretion feature matrix to generate cytokine secretion pattern data; using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and performing fractal degree color mapping to generate a cell dynamic heterogeneous distribution map;

[0007] Step S3: tracking the center of a cell subpopulation according to the cell dynamic heterogeneous distribution map to generate cell subpopulation center point data; performing cell neighborhood growth processing according to the cell subpopulation center point data to obtain cell detection subpopulation data;

[0008] Step S4: using the cell detection subgroup data to determine the cell attribution of the cytokine secretion feature matrix, and performing subgroup cell clustering processing to generate cell detection factor subgroup clustering data; and performing quantitative processing based on the heterogeneity of the cell detection factor subgroup clustering data to obtain cytokine subgroup feature data.

[0009] The present invention converts the cytokine secretion information of a single cell into a quantifiable data form by constructing a cytokine secretion feature matrix, which can retain the difference information between individual cells and avoid the averaging effect from covering up key biological information, such as the specific secretion pattern of rare cell subpopulations. Cytokine secretion pattern analysis and typing degree difference calculation can identify different cytokine secretion patterns in cell populations and quantify the typing degree of each cell in different patterns. This pattern-based analysis method can not only identify cell subpopulations with similar secretion patterns, but also reveal functional differences between different subpopulations, such as pro-inflammatory and anti-inflammatory cell subpopulations. The cell dynamic heterogeneous distribution map generated by fractal degree color mapping shows the heterogeneous distribution of cytokine secretion patterns in cell populations in a visual way, which can more clearly show the distribution and mutual relationship of cell subpopulations, and help to discover new cell subpopulations and potential biomarkers. Cell subpopulation center tracking and neighborhood growth processing realize automatic identification and boundary definition of cell subpopulations, and can more accurately identify cell subpopulations with biological significance. Automatic identification of cell subpopulations according to the intrinsic structure of data is closer to the real situation of biological systems. Through neighborhood growth, cells with similar secretion patterns can be gathered together to form clearer subgroup boundaries, so as to more accurately quantify the characteristics of each subgroup. By carrying out attribution judgment and subgroup clustering processing on cells, and associating cytokine secretion characteristics with cell subgroups, by the refined analysis of single cell data, the characteristic quantification of cytokine secretion heterogeneity is achieved. Therefore, a processing method for cytokine detection data of the present invention converts the cytokine secretion information of single cells into quantifiable data form by constructing a cytokine secretion feature matrix, retains the difference information between individual cells, and avoids the averaging effect from covering up key biological information. Through cytokine secretion pattern analysis and typing degree difference calculation, the different cytokine secretion patterns present in the cell population are identified, and the typing degree of each cell in different patterns is quantified, and the distribution and mutual relationship of cell subgroups are clearly shown.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: obtaining original single cell cytokine detection data;

[0012] Step S12: preprocessing the original single cell cytokine detection data, and performing quality screening on the detection data to generate quality-controlled cytokine detection data;

[0013] Step S13: calculating the total amount of cytokine expression for the cytokine detection data after quality control to generate total amount of cytokine expression data;

[0014] Step S14: performing logarithmic transformation and normalization processing on the total cytokine expression data to generate normalized total cytokine expression data;

[0015] Step S15: rank the cytokines according to the normalized total cytokine expression data to generate cell detection factor ranking data;

[0016] Step S16: analyzing the secretion amount of different cytokines according to the cell detection factor ranking data to obtain cytokine secretion vector data;

[0017] Step S17: construct a multi-cell-cytokine matrix based on the cytokine secretion vector data to obtain a cytokine secretion feature matrix.

[0018] The present invention pre-processes and screens the raw data for quality, removes low-quality data and noise interference, and ensures the reliability of subsequent analysis results. This avoids errors and deviations caused by data quality problems and improves the accuracy of analysis. Logarithmic transformation and normalization further improve the comparability and stability of data. Logarithmic transformation can compress the data range, reduce the influence of extreme values, make data distribution more normal distribution, and be more suitable for subsequent statistical analysis. Normalization eliminates the influence of expression differences between different cytokines, makes the comparison between different cytokines more fair, avoids the phenomenon that high expression cytokines dominate the analysis results, and thus more accurately reflects the differences in cytokine secretion patterns. The secretion analysis on different cytokines converts cytokine expression information into multidimensional vector data, and more comprehensively depicts the cytokine secretion characteristics of cells. The multi-cell-cytokine matrix finally constructed integrates the cytokine secretion information of all cells, and provides an ideal data structure for subsequent pattern analysis and subgroup identification.

[0019] Preferably, step S2 comprises the following steps:

[0020] Step S21: performing matrix transposition processing on the cytokine secretion characteristic matrix to obtain a transposed cytokine secretion characteristic matrix;

[0021] Step S22: performing local similarity analysis between cells according to the transposed cytokine secretion feature matrix to generate local similarity fractal values ​​of cells; performing cytokine secretion pattern analysis according to the local similarity fractal values ​​of cells to generate cytokine secretion pattern data;

[0022] Step S23: using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and performing nonlinear dimensionality reduction processing to generate optimized cell two-dimensional coordinate data;

[0023] Step S24: performing kernel density distribution estimation based on the optimized cell two-dimensional coordinate data to generate cell density distribution estimation data;

[0024] Step S25: Perform fractal color mapping based on the local similarity fractal value of the cells and the estimated data of cell density distribution to generate a dynamic heterogeneous distribution map of the cells.

[0025] The present invention effectively identifies the cytokine secretion pattern by conducting an in-depth analysis of the cytokine secretion feature matrix, and the cytokine secretion pattern analysis converts these similarities into specific pattern data, revealing the different secretion patterns present in the cell population. Using these pattern data to calculate the typing degree difference of the cytokine secretion vector, the typing degree of each cell in different patterns can be quantified, thereby more finely depicting the heterogeneity of the cell. By estimating the density distribution of cells in two-dimensional space, the distribution and mutual relationship of cell subpopulations can be more clearly shown. Finally, the fractal degree color mapping combines the local similarity fractal degree value of the cell with the cell density distribution information to generate a dynamic heterogeneous distribution map of the cell. This picture intuitively shows the heterogeneity of the cell population in the form of color and position. For example, color can represent the typing degree of the cell, and position can represent the distribution of the cell in two-dimensional space.

[0026] Preferably, step S22 comprises the following steps:

[0027] Step S221: calculating the Euclidean distance between cells based on the transposed cytokine secretion feature matrix to generate an initial cell distance matrix;

[0028] Step S222: performing cytokine distribution feature statistics according to the transposed cytokine secretion feature matrix to generate cytokine distribution feature data;

[0029] Step S223: setting a dynamic neighborhood K value according to the cytokine distribution characteristic data to obtain a dynamic neighborhood K value;

[0030] Step S224: selecting candidate cell neighborhoods for the initial cell distance matrix using the dynamic neighborhood K value to obtain cytokine candidate neighborhood list data;

[0031] Step S225: performing local cosine similarity calculation on the transposed cytokine secretion feature matrix through the cytokine candidate neighborhood list data, and performing neighborhood similarity iterative convergence to obtain local similarity data between cells;

[0032] Step S226: Calculating the cell weighted fractal value according to the local similarity data between cells to generate the local similarity fractal value of the cell;

[0033] Step S227: Based on the cytokine distribution feature data, the cytokine secretion pattern analysis is performed on the cytokine candidate neighborhood list data using the cell local similarity fractal value to generate cytokine secretion pattern data.

[0034] The present invention avoids the limitations of fixed K values ​​in different cell density areas by setting dynamic neighborhood K values. In densely populated areas, using smaller K values ​​can more finely characterize local similarities; while in sparsely populated areas, using larger K values ​​can avoid errors caused by insufficient information. The candidate neighborhood selected by the dynamic K value, combined with the local cosine similarity calculation and iterative convergence process, can more accurately quantify the local similarity between cells. Compared with the global similarity calculation, the local similarity pays more attention to the similarity of cells in a specific neighborhood, so as to better reflect the microenvironment characteristics and functional differences of the cells. The local similarity fractal value of the cell calculated using the cell weighted fractal value is combined with the cytokine distribution characteristics to perform cytokine secretion pattern analysis, which can more comprehensively characterize the characteristics of the cell and reveal the functional differences between different cell subpopulations.

[0035] Preferably, step S23 includes the following steps:

[0036] Step S231: using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and generating a similar fractal degree difference matrix between cells;

[0037] Step S232: performing Gaussian kernel probability value conversion according to the inter-cell similarity fractal degree difference matrix, and performing secretion mode transition probability processing to generate a cell secretion mode transition probability matrix;

[0038] Step S233: generating random two-dimensional spatial coordinates of cells according to the cell secretion mode transition probability matrix to obtain initial two-dimensional coordinate data of cells;

[0039] Step S234: using the cell secretion mode transition probability matrix to perform gradient descent optimization on the initial two-dimensional coordinate data of the cell to generate optimized two-dimensional coordinate data of the cell.

[0040] The present invention quantifies the degree of difference between cells in different secretion patterns by calculating the difference matrix of similar fractal degree between cells. The difference information between cells is converted into probability form, and a cell secretion pattern transfer probability matrix is ​​constructed, which reflects the potential association between cells based on the similarity of secretion patterns. The secretion pattern transfer probability matrix is ​​used for gradient descent optimization, so that the distribution of cells in two-dimensional space can better reflect the similarity of their secretion patterns. In other words, cells with similar secretion patterns will be closer in two-dimensional space, while cells with different secretion patterns will be further away.

[0041] Preferably, step S3 comprises the following steps:

[0042] Step S31: tracking the center of a cell subpopulation according to the cell dynamic heterogeneous distribution map to generate cell subpopulation center point data;

[0043] Step S32: Initializing the seed points of the cell subpopulation center point data, and constructing a queue of seed points to be grown, thereby generating the queue data of seed points to be grown;

[0044] Step S33: looping through the queue to be grown of each subgroup according to the queue data of the seed point to be grown, and when the queue is not empty, taking the seed point at the head of the queue as the current growing point to obtain the coordinate data of the current growing point;

[0045] Step S34: performing neighborhood cell evaluation on the cell dynamic heterogeneous distribution map through the current growth point coordinate data to generate neighborhood cell evaluation data;

[0046] Step S35: Based on the neighborhood cell evaluation data, neighborhood cell growth is judged using the preset seed point growth conditions, and the coordinates of the neighborhood cells that meet the seed point growth conditions are extended to the tail of the corresponding growth queue to generate the extended queue data to be grown;

[0047] Step S36: marking the coordinates of neighboring cells that do not meet the seed point growth conditions as cell accesses, and performing subpopulation profile update processing on the current growth point coordinate data to obtain initial subpopulation profile data and visited cell data;

[0048] Step S37: Perform growth iteration judgment on the expanded queue data to be grown through the accessed cell data. When the expanded queue data to be grown is not empty, return to step S33. When the expanded queue data to be grown is empty, perform growth region integration on the initial subpopulation profile data to obtain cell subpopulation profile data.

[0049] Step S38: Optimize the subpopulation boundaries of the cell subpopulation contour data and assign subpopulation labels to obtain cell detection subpopulation data.

[0050] The present invention accurately locates the core area of ​​potential subpopulations through cell subpopulation center tracking, providing reliable seed points for subsequent neighborhood growth. Through seed point initialization and queue construction, the queue to be grown of each subpopulation is traversed in a loop, and the neighborhood cells are evaluated and growth judged, thereby realizing the gradual expansion of the subpopulation. The preset seed point growth conditions, such as the similarity threshold between cells, ensure that only cells that are sufficiently similar to the seed point will be included in the subpopulation, thereby ensuring the internal consistency of the subpopulation. Cells that do not meet the growth conditions are marked for access, avoiding repeated calculations and improving the efficiency of the algorithm. The iterative growth process ensures that the subpopulation can expand to its natural boundaries without being restricted by preset parameters. Compared with traditional distance-based clustering methods, this is more adaptable to the complexity and unevenness of cell distribution, and avoids the erroneous classification of cells that originally belonged to different subpopulations.

[0051] Preferably, step S31 includes the following steps:

[0052] Step S311: performing density map smoothing processing according to the cell dynamic heterogeneous distribution map, and performing cytokine gradient vector calculation to generate cytokine gradient vector data;

[0053] Step S312: setting a gradient modulus threshold according to the cytokine gradient vector data to obtain a density gradient modulus threshold;

[0054] Step S313: performing gradient region judgment on the cytokine gradient vector data based on the density gradient modulus threshold, and marking the cell dynamic heterogeneous distribution map as low gradient region marking data when the cytokine gradient vector data is lower than the density gradient modulus threshold;

[0055] Step S314: when the cytokine gradient vector data is higher than or equal to the density gradient modulus threshold, the cell dynamic heterogeneous distribution map is marked as high gradient area marking data;

[0056] Step S315: performing gradient diffusion of density values ​​of neighboring points on the cell dynamic heterogeneous distribution map based on the high gradient region marking data through the cytokine gradient vector data to obtain an adjusted dynamic heterogeneous distribution map;

[0057] Step S316: Tracking the local density maximum point according to the adjusted dynamic heterogeneous distribution map and the low gradient area marking data, thereby obtaining the cell subpopulation center point data.

[0058] The present invention effectively distinguishes high gradient areas and low gradient areas by setting density gradient modulus thresholds for dynamic heterogeneous distribution maps of cells. Low gradient areas usually correspond to the interior of subpopulations, while high gradient areas correspond to the boundaries between subpopulations. This distinction is crucial for accurately identifying subpopulation centers. Gradient diffusion based on high gradient area marker data effectively enhances the density value of subpopulation centers, making them more prominent in the local range. This makes it easier to accurately identify subpopulation centers in the subsequent local density maximum point tracking process, avoiding the mistaken identification of cells in the boundary area as subpopulation centers. By combining the adjusted dynamic heterogeneous distribution map and low gradient area marker data to track local density maximum points, the subpopulation center points can be more accurately located. The cell subpopulation center point data finally obtained accurately reflects the central position of the cell subpopulation, thereby improving the accuracy and reliability of subpopulation identification.

[0059] Preferably, step S4 comprises the following steps:

[0060] Step S41: using the cell detection subgroup data to determine the cell attribution of the cytokine secretion feature matrix, and performing subgroup cell clustering processing to generate cell detection factor subgroup clustering data;

[0061] Step S42: Calculating the subgroup secretion mean of the cell detection factor subgroup clustering data to generate cytokine subgroup secretion data;

[0062] Step S43: calculating the degree of subgroup variation of the cell detection factor subgroup clustering data using the cytokine subgroup secretion data to generate a subgroup cytokine variation coefficient;

[0063] Step S44: performing heterogeneity quantification processing according to the coefficient of variation of subgroup cytokines and the cytokine subgroup secretion data, thereby obtaining cytokine subgroup characteristic data.

[0064] The present invention clearly assigns cells to different subgroups through cell attribution judgment and subgroup cell clustering processing, and generates clustering data containing cytokine secretion information of each subgroup. The average secretion level of each subgroup on different cytokines is quantified, reflecting the overall secretion characteristics of the subgroup. The calculation of subgroup variation captures the discrete degree of each subgroup on the secretion of different cytokines through the coefficient of variation. A larger coefficient of variation indicates that there are large differences in the secretion of the cytokine within the subgroup, that is, the heterogeneity is high. Combining the coefficient of variation of subgroup cytokines and the mean value of subgroup secretion for heterogeneity quantification processing, the cytokine subgroup characteristic data is obtained. These data not only include the average secretion level of each subgroup, but also reflect the degree of heterogeneity within the subgroup, thereby more comprehensively characterizing the characteristics of the subgroup. For example, a subgroup has a higher average secretion level of a certain cytokine, but its internal coefficient of variation is also larger, which indicates that there are different subgroups secreting the cytokine within the subgroup, suggesting a more refined functional differentiation.

[0065] Preferably, the present invention further provides a cytokine detection data processing system, which executes the cytokine detection data processing method as described above, and the cytokine detection data processing system comprises:

[0066] The detection data preprocessing module is used to obtain the original single-cell cytokine detection data; analyze the cytokine secretion amount of the original single-cell cytokine detection data to obtain the cytokine secretion vector data; construct a multi-cell-cytokine matrix based on the cytokine secretion vector data to obtain the cytokine secretion feature matrix;

[0067] The cytokine secretion analysis module is used to analyze the cytokine secretion pattern of the cytokine secretion feature matrix and generate cytokine secretion pattern data; the cytokine secretion pattern data is used to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and perform fractal degree color mapping to generate a cell dynamic heterogeneous distribution map;

[0068] The cell subpopulation center tracking module is used to track the cell subpopulation center according to the cell dynamic heterogeneous distribution map and generate the cell subpopulation center point data; perform cell neighborhood growth processing according to the cell subpopulation center point data to obtain the cell detection subpopulation data;

[0069] The cytokine heterogeneity clustering module is used to judge the cell attribution of the cytokine secretion feature matrix through the cell detection subgroup data, and perform subgroup cell clustering processing to generate cell detection factor subgroup clustering data; according to the heterogeneity quantitative processing of the cell detection factor subgroup clustering data, the cytokine subgroup characteristic data is obtained.

[0070] Preferably, the present invention further provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed, implements the method for processing cytokine detection data as described in any one of the above. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A schematic diagram of the steps of the method for processing cytokine detection data of the present invention;

[0072] Figure 2 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0073] Figure 3 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0074] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0075] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0076] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0077] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0078] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for processing cytokine detection data, comprising the following steps:

[0079] Step S1: obtaining original single-cell cytokine detection data; performing cytokine secretion analysis on the original single-cell cytokine detection data to obtain cytokine secretion vector data; constructing a multi-cell-cytokine matrix based on the cytokine secretion vector data to obtain a cytokine secretion feature matrix;

[0080] Step S2: performing cytokine secretion pattern analysis on the cytokine secretion feature matrix to generate cytokine secretion pattern data; using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and performing fractal degree color mapping to generate a cell dynamic heterogeneous distribution map;

[0081] Step S3: tracking the center of a cell subpopulation according to the cell dynamic heterogeneous distribution map to generate cell subpopulation center point data; performing cell neighborhood growth processing according to the cell subpopulation center point data to obtain cell detection subpopulation data;

[0082] Step S4: using the cell detection subgroup data to determine the cell attribution of the cytokine secretion feature matrix, and performing subgroup cell clustering processing to generate cell detection factor subgroup clustering data; and performing quantitative processing based on the heterogeneity of the cell detection factor subgroup clustering data to obtain cytokine subgroup feature data.

[0083] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of the steps of the method for processing cytokine detection data of the present invention. In this embodiment, the method for processing cytokine detection data includes the following steps:

[0084] Step S1: obtaining original single-cell cytokine detection data; performing cytokine secretion analysis on the original single-cell cytokine detection data to obtain cytokine secretion vector data; constructing a multi-cell-cytokine matrix based on the cytokine secretion vector data to obtain a cytokine secretion feature matrix;

[0085] In an embodiment of the present invention, single-cell sequencing technology (e.g., CITE-seq) is used to obtain raw single-cell cytokine detection data. The data contains the expression or secretion of various cytokines detected by each single cell. For example, cytokines such as TNF-α, IFN-γ, IL-2, IL-4, IL-6, and IL-10 can be detected. Assuming that the experiment detected 1,000 cells and 10 cytokines, the raw data will be a matrix of 1,000 rows (cells) × 10 columns (cytokines), and each element represents the expression level of a specific cytokine in a specific cell. The raw data is quality controlled, such as removing low-quality cells and low-expression cytokines; the data is standardized, such as using z-score standardization to make the data between different cytokines comparable; if necessary, the data can be reduced in dimension, such as using principal component analysis (PCA) to reduce the data dimension. After processing, each cell can be represented by a cytokine secretion vector. For example, the secretion vector of a cell can be [0.5, 1.2, -0.8, 0.2, 1.5, 0.1, -0.5, 0.9, 1.1, -0.2], which represents the secretion of 10 cytokines of the cell. A multi-cell-cytokine matrix, i.e., a cytokine secretion feature matrix, is constructed based on the cytokine secretion vector data. The cytokine secretion vectors of all cells are arranged in rows to obtain a cytokine secretion feature matrix. For example, the cytokine secretion vector data of 1,000 cells can form a matrix of 1,000 rows × 10 columns, which is the cytokine secretion feature matrix.

[0086] Step S2: performing cytokine secretion pattern analysis on the cytokine secretion feature matrix to generate cytokine secretion pattern data; using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and performing fractal degree color mapping to generate a cell dynamic heterogeneous distribution map;

[0087] In an embodiment of the present invention, a clustering algorithm (e.g., k-means clustering, hierarchical clustering) is used to divide cells into different groups according to their cytokine secretion patterns, and each group represents a unique cytokine secretion pattern. For example, cells can be divided into 5 groups, representing different secretion patterns such as pro-inflammatory, anti-inflammatory, regulatory, mixed and low secretion types. The clustering result is the cytokine secretion pattern data, and the cytokine secretion vector of each cell is used to perform secretion pattern typing difference calculation. The distance between each cell and the center of each cytokine secretion pattern is calculated, and the closer the distance is, the closer the secretion pattern representing the cell is to the center, and the higher the typing degree is. For example, the distance between a cell and the center of 5 secretion patterns can be calculated to obtain 5 distance values. These distance values ​​are normalized so that their sum is 1. These normalized distance values ​​can be regarded as the probability that the cell belongs to different secretion patterns. Fractal degree color mapping is performed according to the difference in the secretion pattern typing degree of each cell to generate a cell dynamic heterogeneous distribution map. Different secretion patterns are represented by different colors, for example, pro-inflammatory type is red, anti-inflammatory type is blue, regulatory type is green, mixed type is yellow, and low secretion type is gray. According to the probability of each cell belonging to a different secretion pattern, different colors are superimposed to obtain the final color of each cell. By plotting the colors of all cells in two-dimensional or three-dimensional space, a dynamic heterogeneous distribution map of cells can be generated, which intuitively shows the heterogeneity of cytokine secretion patterns.

[0088] Step S3: tracking the center of a cell subpopulation according to the cell dynamic heterogeneous distribution map to generate cell subpopulation center point data; performing cell neighborhood growth processing according to the cell subpopulation center point data to obtain cell detection subpopulation data;

[0089] In an embodiment of the present invention, in the distribution map, cells with similar colors are clustered together to form different cell subpopulations. By identifying the center points of these clustered areas, the center position of each cell subpopulation can be determined. The density peak clustering algorithm (DBSCAN) or other clustering algorithms can be used to identify these center points, and their coordinates are recorded to generate cell subpopulation center point data. For example, in a two-dimensional distribution map, five cell subpopulations can be identified, and their center point coordinates are (1,2), (3,4), (5,6), (7,8) and (9,10), respectively. Cell neighborhood growth processing is performed according to the cell subpopulation center point data. Taking each center point as a seed, it is gradually expanded to the surrounding area, and cells whose distance is less than a set threshold are included in the subpopulation. This threshold can be adjusted according to the cell density and distribution. Cell neighborhood growth processing can effectively classify discrete cell points in the distribution map into different subpopulations and generate cell detection subpopulation data. For example, taking the center point (1,2) as a seed and setting the threshold to 1, all cells whose distance (1,2) is less than 1 are included in the subpopulation. Ultimately, each cell is assigned to a specific subset.

[0090] Step S4: using the cell detection subgroup data to determine the cell attribution of the cytokine secretion feature matrix, and performing subgroup cell clustering processing to generate cell detection factor subgroup clustering data; and performing quantitative processing based on the heterogeneity of the cell detection factor subgroup clustering data to obtain cytokine subgroup feature data.

[0091] In an embodiment of the present invention, the cell attribution judgment is performed on the cytokine secretion feature matrix obtained in step S1 according to the cell detection subgroup data obtained in step S3. Each cell is assigned to the subgroup to which it belongs, and the cytokine secretion feature of each subgroup is summarized. For example, statistical indicators such as the average secretion amount, median secretion amount, and secretion amount variance of various cytokines in each subgroup can be calculated. Subgroup cell clustering is performed on the cells in each subgroup to generate cell detection factor subgroup clustering data. The purpose of this step is to further refine the cell heterogeneity within the subgroup. For example, in a subgroup, there are cell subtypes that secrete different cytokine combinations. Clustering algorithms can be used to further divide the cells in the subgroup into smaller subtypes, and the cytokine secretion features of each subtype are recorded. Heterogeneity quantification is performed on the cell detection factor subgroup clustering data to obtain cytokine subgroup feature data. A variety of indicators can be used to quantify the heterogeneity within the subgroup, for example, the proportion of different subtype cells in each subgroup can be calculated, or the coefficient of variation of the cytokine secretion amount in each subgroup can be calculated. These heterogeneity indicators can help us better understand the functions and characteristics of different subpopulations. For example, if the heterogeneity within a subpopulation is high, it means that the subpopulation contains multiple cell subtypes with different functions; while if the heterogeneity within a subpopulation is low, it means that the cell functions within the subpopulation are relatively consistent. These quantitative heterogeneity indicators constitute the characteristic data of cytokine subpopulations.

[0092] Preferably, step S1 comprises the following steps:

[0093] Step S11: obtaining original single cell cytokine detection data;

[0094] Step S12: preprocessing the original single cell cytokine detection data, and performing quality screening on the detection data to generate quality-controlled cytokine detection data;

[0095] Step S13: calculating the total amount of cytokine expression for the cytokine detection data after quality control to generate total amount of cytokine expression data;

[0096] Step S14: performing logarithmic transformation and normalization processing on the total cytokine expression data to generate normalized total cytokine expression data;

[0097] Step S15: rank the cytokines according to the normalized total cytokine expression data to generate cell detection factor ranking data;

[0098] Step S16: analyzing the secretion amount of different cytokines according to the cell detection factor ranking data to obtain cytokine secretion vector data;

[0099] Step S17: construct a multi-cell-cytokine matrix based on the cytokine secretion vector data to obtain a cytokine secretion feature matrix.

[0100] In an embodiment of the present invention, single-cell sequencing technology (e.g., CITE-seq) or single-cell proteomics technology (e.g., CyTOF) is used to obtain raw single-cell cytokine detection data. These technologies can detect the expression levels of multiple cytokines in a single cell. Assuming that the experiment detected 500 cells and 15 cytokines, the raw data will be a matrix of 500 rows (cells) × 15 columns (cytokines), each element representing the expression or secretion of a specific cytokine in a specific cell, for example, it can be expressed by UMI counts or protein mean fluorescence intensity (MFI). The purpose of quality screening is to remove low-quality cell and cytokine data. For example, low-quality cells can be filtered out based on indicators such as total cell cytokine expression, mitochondrial gene expression, and the number of detected cytokines, while removing cytokines with extremely low expression in most cells. For each cell, the sum of the expression levels of all detected cytokines is calculated. This can reflect the overall cytokine secretion activity of the cell. For example, if the expression levels of 12 cytokines detected in a cell are [10, 5, 8, 12, 3, 7, 9, 11, 4, 6, 13, 2], then the total cytokine expression of the cell is 80. The total cytokine expression of all cells is summarized to generate the total cytokine expression data, which is a vector containing 450 values. The logarithmic transformed data is normalized, such as using min-max normalization or z-score standardization, to make the data between different cells comparable. For example, the total cytokine expression of all cells can be scaled to between 0 and 1, and the processed data is called normalized total cytokine expression data. Sort the cells according to the normalized total cytokine expression data obtained in step S14, from high to low. This can help identify cells with high cytokine secretion activity. The sorting result is represented by a vector containing 450 cell indexes, which is called the cell detection factor sorting data. For example, [cell 321, cell 123, cell 45, ...] means that the total cytokine expression of cell 321 is the highest, followed by cell 123, and so on. For each cell, the corresponding cytokine expression in the quality control cytokine detection data obtained in step S12 is extracted to form a vector. For example, the cytokine secretion vector of a cell can be [0.8, 0.2, 0.5, 0.9, 0.1, 0.3, 0.6, 1.0, 0.15, 0.25, 1.1, 0.05], which represents the normalized expression of 12 cytokines in the cell. The cytokine secretion vectors of all cells in step S16 are arranged in rows to construct a multi-cell-cytokine matrix, that is, the cytokine secretion feature matrix. This is a matrix of 450 rows × 12 columns, in which each row represents a cell, each column represents a cytokine, and each element in the matrix corresponds to the amount of a cytokine secreted by a cell.

[0101] Preferably, step S2 comprises the following steps:

[0102] Step S21: performing matrix transposition processing on the cytokine secretion characteristic matrix to obtain a transposed cytokine secretion characteristic matrix;

[0103] Step S22: performing local similarity analysis between cells according to the transposed cytokine secretion feature matrix to generate local similarity fractal values ​​of cells; performing cytokine secretion pattern analysis according to the local similarity fractal values ​​of cells to generate cytokine secretion pattern data;

[0104] Step S23: using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and performing nonlinear dimensionality reduction processing to generate optimized cell two-dimensional coordinate data;

[0105] Step S24: performing kernel density distribution estimation based on the optimized cell two-dimensional coordinate data to generate cell density distribution estimation data;

[0106] Step S25: Perform fractal color mapping based on the local similarity fractal value of the cells and the estimated data of cell density distribution to generate a dynamic heterogeneous distribution map of the cells.

[0107] In an embodiment of the present invention, the cytokine secretion feature matrix obtained in step S17 is transposed to obtain a transposed cytokine secretion feature matrix. The dimension of the original matrix is ​​the number of cells × the number of cytokines (e.g., 450 × 12), and the dimension of the transposed matrix becomes the number of cytokines × the number of cells (e.g., 12 × 450). The local similarity between cells is calculated based on the transposed cytokine secretion feature matrix. The similarity of the cytokine secretion pattern between each pair of cells can be calculated using methods such as the Pearson correlation coefficient or the Spearman correlation coefficient. The average similarity of each cell to its k nearest neighbor cells is taken as the local similarity fractal value of the cell. For example, if k=10 is set, the average similarity of each cell to its 10 nearest neighbor cells is calculated. Cytokine secretion pattern analysis is performed according to the local similarity fractal value of the cell. The density peak clustering algorithm (DBSCAN) or other clustering algorithms can be used to divide cells into different clusters according to the local similarity fractal value and density of the cells, and each cluster represents a unique cytokine secretion pattern. The clustering result is the cytokine secretion pattern data. For example, cells can be divided into 4 clusters, representing pro-inflammatory, anti-inflammatory, regulatory and low secretion types. The distance between each cell and the center of each cytokine secretion pattern is calculated. The smaller the distance, the more similar the cell is to the secretion pattern. These distance values ​​can be normalized to obtain the probability that each cell belongs to a different secretion pattern, that is, the difference in secretion pattern typing. Using nonlinear dimensionality reduction methods such as t-SNE or UMAP, the high-dimensional cytokine secretion vector data is reduced to a two-dimensional space, and the coordinates after dimensionality reduction are saved as optimized cell two-dimensional coordinate data. The dimensionality reduction process retains the local similarity between cells, so that in two-dimensional space, similar cells are still clustered together, and different types of cells are separated from each other. By placing a kernel function (such as a Gaussian kernel function) at the two-dimensional coordinate position of each cell, the density of each cell location is calculated to generate cell density distribution estimation data. High-density areas indicate high cell aggregation, and low-density areas indicate sparse cell distribution. Usually, a heat map can be used for visualization. Cells with high fractal degree are represented by warm colors, such as red and orange, and cells with low fractal degree are represented by cold colors, such as blue and green. In order to make the cell dynamic heterogeneous distribution map more readable, the cell density distribution estimation data can be combined. For example, the color of the area with high cell density can be made darker, and the color of the area with low cell density can be made lighter. The two-dimensional coordinates of the cell can be used as the position of the cell on the map, and combined with the color of the fractal degree mapping, the cell dynamic heterogeneous distribution map can be drawn. This map can intuitively show the heterogeneous distribution of cells, that is, different cell secretion patterns and their distribution.For example, on the heat map, redder areas represent cell clusters with high fractal degrees, and the cytokine secretion patterns of these cells are quite different from those of surrounding cells; bluer areas represent cell clusters with low fractal degrees, and the cytokine secretion patterns of these cells are similar to those of surrounding cells.

[0108] Preferably, step S22 comprises the following steps:

[0109] Step S221: calculating the Euclidean distance between cells based on the transposed cytokine secretion feature matrix to generate an initial cell distance matrix;

[0110] Step S222: performing cytokine distribution feature statistics according to the transposed cytokine secretion feature matrix to generate cytokine distribution feature data;

[0111] Step S223: setting a dynamic neighborhood K value according to the cytokine distribution characteristic data to obtain a dynamic neighborhood K value;

[0112] Step S224: selecting candidate cell neighborhoods for the initial cell distance matrix using the dynamic neighborhood K value to obtain cytokine candidate neighborhood list data;

[0113] Step S225: performing local cosine similarity calculation on the transposed cytokine secretion feature matrix through the cytokine candidate neighborhood list data, and performing neighborhood similarity iterative convergence to obtain local similarity data between cells;

[0114] Step S226: Calculating the cell weighted fractal value according to the local similarity data between cells to generate the local similarity fractal value of the cell;

[0115] Step S227: Based on the cytokine distribution feature data, the cytokine secretion pattern analysis is performed on the cytokine candidate neighborhood list data using the cell local similarity fractal value to generate cytokine secretion pattern data.

[0116] In an embodiment of the present invention, for the transposed cytokine secretion feature matrix, each column represents the cytokine secretion vector of a cell. The Euclidean distance between two cells is calculated by first calculating the square of the difference between the corresponding elements in the cytokine secretion vectors of the two cells, then adding the square sum of the differences of all elements, and finally taking the square root. For example, if the cytokine secretion vector of cell A is [2,3,1,4,5] and the cytokine secretion vector of cell B is [1,2,3,2,4], then the Euclidean distance between cell A and cell B is sqrt((2-1)^2+(3-2)^2+(1-3)^2+(4-2)^2+(5-4)^2)=sqrt(1+1+4+4+1)=sqrt(11)≈3.32. For all cells in the transposed cytokine secretion feature matrix, the Euclidean distance is calculated pairwise to obtain an initial cell distance matrix. For example, statistics such as the average expression level, variance, and skewness of each cytokine can be calculated. The correlation coefficient between cytokines can also be calculated to describe the co-expression relationship between cytokines. These statistics constitute the cytokine distribution feature data. The dynamic neighborhood K value setting can determine the appropriate K value according to the cell density and cytokine distribution characteristics around each cell. For example, for areas with high cell density, a smaller K value can be set to select only nearby cells as neighborhoods; for areas with low cell density, a larger K value can be set to expand the neighborhood range. For example, based on local density estimation, a smaller K value can be selected for areas with high density and a larger K value can be selected for areas with low density. Methods based on cytokine distribution characteristics can also be used, such as based on the variance of the cytokine expression of each cell, cells with large variance use larger K values, and cells with small variance use smaller K values. For each cell, according to its corresponding dynamic neighborhood K value, the K cells closest to the cell are selected from the initial cell distance matrix as candidate neighborhoods of the cell. For example, if the dynamic neighborhood K value of cell A is 8, then in the initial cell distance matrix, starting from the cell closest to cell A, the 8 cells closest to cell A are selected as candidate neighborhoods of cell A. The cytokine secretion vectors of the two cells are regarded as two vectors, and then the cosine values ​​of the two vectors are calculated. The higher the cosine value, the smaller the angle between the two vectors and the higher the similarity. For each cell, calculate its cosine similarity with all candidate neighboring cells. After calculating the initial local cosine similarity, perform iterative convergence of neighborhood similarity. The purpose of iterative convergence is to make the local similarity better reflect the local similarity between cells rather than the global similarity. The concept of fractal dimension is used to calculate the fractal value. For example, based on the distribution of local similarity, the box counting method or other fractal dimension calculation methods can be used to calculate the fractal dimension of each cell as the weighted fractal value of the cell.For example, the higher the similarity of cells around a cell, the lower the weighted fractal value of the cell, indicating that the local homogeneity of the cell is higher; the lower the similarity of cells around a cell, the higher the weighted fractal value, indicating that the local heterogeneity of the cell is higher. Combined with the cytokine distribution feature data obtained in step S222 and the local similarity fractal value of the cell obtained in step S226, the cytokine candidate neighborhood list data obtained in step S224 is analyzed for cytokine secretion pattern. Density peak clustering algorithm (DBSCAN) or other clustering algorithms can be used to divide cells into different clusters according to the local similarity fractal value, density and cytokine distribution characteristics of cells, and each cluster represents a unique cytokine secretion pattern. The clustering result is the cytokine secretion pattern data.

[0117] Preferably, step S23 includes the following steps:

[0118] Step S231: using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and generating a similar fractal degree difference matrix between cells;

[0119] Step S232: performing Gaussian kernel probability value conversion according to the inter-cell similarity fractal degree difference matrix, and performing secretion mode transition probability processing to generate a cell secretion mode transition probability matrix;

[0120] Step S233: generating random two-dimensional spatial coordinates of cells according to the cell secretion mode transition probability matrix to obtain initial two-dimensional coordinate data of cells;

[0121] Step S234: using the cell secretion mode transition probability matrix to perform gradient descent optimization on the initial two-dimensional coordinate data of the cell to generate optimized two-dimensional coordinate data of the cell.

[0122] In an embodiment of the present invention, for example, cells are divided into a high secretion group, a low secretion group or a mixed secretion group. The secretion pattern typing difference calculation is intended to quantify the difference between the cytokine secretion pattern of each cell and the typical pattern of the secretion pattern group to which it belongs. A cell-to-cell similarity fractal degree difference matrix can be constructed, and each element of the matrix represents the absolute value difference between the typing degree difference values ​​of two cells. For example, assuming that cell A belongs to the high secretion group, the calculated distance from the average value of the high secretion group is 0.8; cell B belongs to the low secretion group, and the distance from the average value of the low secretion group is 0.2, then the similarity fractal degree difference value between cell A and cell B is |0.8-0.2|=0.6. After the construction is completed, the matrix can be used to measure the heterogeneity between cells. The diagonal elements of the cell-to-cell similarity fractal degree difference matrix are all 0, and the matrix is ​​symmetrical, that is, the similarity fractal degree difference value between cell A and cell B is equal to the similarity fractal degree difference value between cell B and cell A. The cell-to-cell similarity fractal degree difference matrix obtained in step S231 is converted into a probability value. The Gaussian kernel function can be used to convert the distance into probability. The smaller the distance, the greater the probability. Then, for each cell, the probability of belonging to different secretion modes is normalized so that the sum of the probabilities of each cell is 1. For example, for each cell, the sum of the conversion probability values ​​of the cell and all other cells is first calculated, and then the conversion probability values ​​of the cell and all other cells are divided by this sum to obtain the normalized transition probability. For example, if the conversion probability value between cell A and cell B is 0.7 after Gaussian kernel conversion, the conversion probability value between cell A and cell C is 0.5, and the sum of the transition probabilities between cell A and all cells is 2, then the transition probability value between cell A and cell B is 0.7 / 2=0.35, and the transition probability value between cell A and cell C is 0.5 / 2=0.25. After the secretion mode transition probability processing, the cell secretion mode transition probability matrix is ​​obtained. Each row of the matrix represents a cell, each column represents a cell, and the elements in the matrix represent the secretion mode transition probability between the corresponding cells. Each secretion mode is regarded as an attraction point in two-dimensional space, and cells are distributed around these attraction points according to their probability of belonging to different secretion modes. For example, if the probability of a cell belonging to mode 1 is 0.8 and the probability of belonging to mode 2 is 0.2, the initial two-dimensional coordinates of the cell are closer to the attraction point of mode 1. The generated initial two-dimensional coordinate data of the cell is an N×2 matrix, where N is the number of cells, such as 450. Gradient descent is an optimization algorithm that iteratively adjusts the two-dimensional coordinates of the cell so that the distribution of the cell in two-dimensional space can better reflect the probability of secretion mode transition between cells. The goal of gradient descent optimization is to minimize the objective function, which defines the difference between the actual position of the cell and the transition probability.For example, the objective function can be defined as when the transition probability of two cells is high, their distance in two-dimensional space should be closer, and when the transition probability of two cells is low, their distance in two-dimensional space should be farther. By iteratively adjusting the coordinates of each cell, the distribution of cells can be made more consistent with the expected secretion mode transition probability. In each iteration, the gradient of each cell is calculated, and the gradient points to the direction where the objective function decreases fastest, and then the position of the cell is updated according to a certain step size. In order to avoid falling into a local minimum, a momentum term can be added or an adaptive learning rate can be used. After multiple iterations, when the change of the objective function tends to stabilize, the optimization process can be stopped. This step outputs the optimized cell two-dimensional coordinate data, which contains the optimized two-dimensional coordinates of all cells.

[0123] Preferably, step S3 comprises the following steps:

[0124] Step S31: tracking the center of a cell subpopulation according to the cell dynamic heterogeneous distribution map to generate cell subpopulation center point data;

[0125] Step S32: Initializing the seed points of the cell subpopulation center point data, and constructing a queue of seed points to be grown, thereby generating the queue data of seed points to be grown;

[0126] Step S33: looping through the queue to be grown of each subgroup according to the queue data of the seed point to be grown, and when the queue is not empty, taking the seed point at the head of the queue as the current growing point to obtain the coordinate data of the current growing point;

[0127] Step S34: performing neighborhood cell evaluation on the cell dynamic heterogeneous distribution map through the current growth point coordinate data to generate neighborhood cell evaluation data;

[0128] Step S35: Based on the neighborhood cell evaluation data, neighborhood cell growth is judged using the preset seed point growth conditions, and the coordinates of the neighborhood cells that meet the seed point growth conditions are extended to the tail of the corresponding growth queue to generate the extended queue data to be grown;

[0129] Step S36: marking the coordinates of neighboring cells that do not meet the seed point growth conditions as cell accesses, and performing subpopulation profile update processing on the current growth point coordinate data to obtain initial subpopulation profile data and visited cell data;

[0130] Step S37: Perform growth iteration judgment on the expanded queue data to be grown through the accessed cell data. When the expanded queue data to be grown is not empty, return to step S33. When the expanded queue data to be grown is empty, perform growth region integration on the initial subpopulation profile data to obtain cell subpopulation profile data.

[0131] Step S38: Optimize the subpopulation boundaries of the cell subpopulation contour data and assign subpopulation labels to obtain cell detection subpopulation data.

[0132] As an example of the present invention, refer to Figure 2 As shown, Figure 1 Detailed implementation steps of step S3 in the flowchart, in this example, step S3 includes:

[0133] Step S31: tracking the center of a cell subpopulation according to the cell dynamic heterogeneous distribution map to generate cell subpopulation center point data;

[0134] In an embodiment of the present invention, after obtaining the cell dynamic heterogeneous distribution map, it is necessary to track the center of the cell subpopulation according to the map to generate the cell subpopulation center point data. The cell dynamic heterogeneous distribution map shows the heterogeneity of the cells in a color-coded manner, wherein cells with similar cytokine secretion patterns tend to gather together to form different regions. For example, the cell dynamic heterogeneous distribution map can be image segmented, the regions with similar colors can be segmented out, and then the center point of each segmented region can be calculated. The selection of the center point needs to consider two aspects, one is the color feature, that is, the center point is usually located in an area where the color is more concentrated and more obvious, and the other is the spatial feature, that is, the center point is usually located in the center of the region. For each identified cell subpopulation center point, it is necessary to record its coordinates in two-dimensional space and the corresponding cell index. The cell subpopulation center point data can be stored in a table, with each row representing a center point, and the column containing the x-coordinate, y-coordinate and corresponding cell index of the center point.

[0135] Step S32: Initializing the seed points of the cell subpopulation center point data, and constructing a queue of seed points to be grown, thereby generating the queue data of seed points to be grown;

[0136] In an embodiment of the present invention, seed point initialization is performed on these center points, and a queue of seed points to be grown is constructed to generate queue data of seed points to be grown. Seed point initialization refers to creating an initial state for each cell subpopulation center point, including its coordinates, the subpopulation number to which it belongs, and other necessary growth parameters. For example, each center point needs to be assigned a unique subpopulation number and a flag indicating its current growth state, which is usually set to "to be grown" at the beginning. The queue of seed points to be grown refers to a queue for storing all seed points to be grown, and the queue follows the first-in-first-out principle to ensure that each seed point can be processed in sequence. The process of constructing the queue of seed points to be grown is to add all initialized seed points (i.e., all cell subpopulation center points) to the queue. The queue data of seed points to be grown can be stored in a list or other data structure, and each element in the queue represents a seed point, including information such as the coordinates of the seed point, the subpopulation number to which it belongs, and the growth state. For example, if there are 5 subpopulation center points, after initialization, these 5 seed points are all added to the queue of seed points to be grown and arranged in sequence in the order of addition.

[0137] Step S33: looping through the queue to be grown of each subgroup according to the queue data of the seed point to be grown, and when the queue is not empty, taking the seed point at the head of the queue as the current growing point to obtain the coordinate data of the current growing point;

[0138] In the embodiment of the present invention, for each subgroup, an independent queue to be grown is maintained. During the loop traversal process, it is first determined whether the queue to be grown of the current subgroup is empty. If it is empty, the current subgroup is skipped and the next subgroup is processed. If the queue is not empty, the seed point at the head of the queue is taken out as the current growth point. The current growth point refers to the cell that is currently undergoing neighborhood cell evaluation, and its coordinate information will be used as the basis for evaluating the neighborhood cells in the next step. For example, assuming that the queue of seed points to be grown of the currently processed subgroup A is not empty and contains seed points S1, S2, and S3, then S1 at the head of the queue is first taken out as the current growth point. At the same time, S1 is deleted from the queue. Then, the coordinate data of the growth point is recorded, including the x-coordinate and y-coordinate of the point, as well as the subgroup number to which it belongs. The current growth point coordinate data will be used as the input of step S34 to evaluate the neighborhood cells and determine which cells can be included in the subgroup.

[0139] Step S34: performing neighborhood cell evaluation on the cell dynamic heterogeneous distribution map through the current growth point coordinate data to generate neighborhood cell evaluation data;

[0140] In an embodiment of the present invention, according to the coordinate data of the current growth point obtained in step S33, the neighborhood cells around the current growth point are evaluated in the dynamic heterogeneous distribution map of cells generated in step S25. The evaluation index may include the distance from the current growth point, the local similarity fractal value, the density, etc. For example, a distance threshold may be set, and the cells whose distance from the current growth point is less than the threshold are used as candidate neighborhood cells; then, further screening is performed based on the local similarity fractal value and density of the candidate neighborhood cells, and cells that meet specific conditions are selected as the final neighborhood cells. If a neighborhood cell meets the similarity condition, the cell is marked as belonging to the current subpopulation and added to the queue to be grown of the subpopulation. It should be noted that a cell belongs to the neighborhood of multiple cell subpopulations at the same time. At this time, a certain strategy can be used to determine which subpopulation the cell ultimately belongs to, such as selecting the subpopulation to which the growth point closest to it belongs, or selecting the subpopulation to which the growth point with the highest similarity belongs. For example, if the coordinates of the current growth point are (x1, y1), the coordinates of the neighboring cells are (x2, y2), the set radius is r, and the similarity threshold is t, if sqrt((x1-x2)^2+(y1-y2)^2)<=r, and the similarity between the cell (x2, y2) and the current growth point is greater than t, then the cell (x2, y2) is marked as belonging to the current subpopulation.

[0141] Step S35: Based on the neighborhood cell evaluation data, neighborhood cell growth is judged using the preset seed point growth conditions, and the coordinates of the neighborhood cells that meet the seed point growth conditions are extended to the tail of the corresponding growth queue to generate the extended queue data to be grown;

[0142] In an embodiment of the present invention, after obtaining the neighborhood cell evaluation data, it is necessary to perform neighborhood cell growth judgment based on the data and the preset seed point growth conditions. The preset seed point growth conditions refer to the rules for judging whether the neighborhood cells meet the rules for being included in the current subgroup. These rules can be based on the similarity, distance, and fractal difference characteristics of the cells. For example, a common growth condition is that if the similarity between the neighborhood cell and the current growth point is higher than a certain threshold, and the fractal difference is less than a certain threshold, the cell meets the growth conditions. For neighborhood cells that meet the growth conditions, their coordinates are added to the tail of the queue to be grown of the corresponding subgroup to achieve the expansion of the queue. For example, if the current growth point belongs to subgroup A, and the neighborhood cell B meets the growth conditions of subgroup A, then the coordinates of cell B are added to the tail of the queue to be grown of subgroup A. In this way, cell B will also become a new growth point in subsequent iterations, and the subgroup will continue to expand outward, thereby obtaining the data of the queue to be grown after the expansion, and the queue contains the newly added seed points. For example, Python's list or queue data structure can be used to store the expanded queue to be grown.

[0143] Step S36: marking the coordinates of neighboring cells that do not meet the seed point growth conditions as cell accesses, and performing subpopulation profile update processing on the current growth point coordinate data to obtain initial subpopulation profile data and visited cell data;

[0144] In an embodiment of the present invention, neighboring cells that do not meet the growth conditions are added to the visited cell data set, which records all the cells that have been visited. In subsequent evaluations, if a visited cell is encountered, the evaluation is skipped. The purpose of updating the subpopulation outline is to record the boundary information of the subpopulation. The current growth point coordinates can be used as a component of the subpopulation outline. For example, the coordinates of the current growth point can be added to the outline list of the subpopulation, or a more complex method can be used, such as calculating the convex hull of the current growth point and its neighboring cells to update the subpopulation outline. The output of this step includes initial subpopulation outline data, which contains the outline information of each subpopulation, and visited cell data, which records all visited cells to ensure that cells are not visited repeatedly.

[0145] Step S37: Perform growth iteration judgment on the expanded queue data to be grown through the accessed cell data. When the expanded queue data to be grown is not empty, return to step S33. When the expanded queue data to be grown is empty, perform growth region integration on the initial subpopulation profile data to obtain cell subpopulation profile data.

[0146] In the embodiment of the present invention, it is checked whether the data of the queue to be grown after expansion is empty. If the queue to be grown is empty, it means that the growth of the subpopulation has been completed, and the subpopulation contour data can be integrated; if the queue to be grown is not empty, it means that the subpopulation still has seed points to be processed, and the process returns to step S33, and the head seed point of the queue to be grown is continuously taken out for processing. The purpose of the growth iteration judgment is to control the growth process of the cell subpopulation. When the growth of the subpopulation stops, the contour information of the subpopulation is integrated together. When the queues to be grown of all subpopulations are empty, it means that the growth of all cell subpopulations has been completed, and the subpopulation contours can be integrated. The purpose of the subpopulation contour data integration is to combine the initial subpopulation contour data into a complete subpopulation contour. Since each cell can only update the local subpopulation contour in the process of cell neighborhood growth, it is necessary to combine these local contour information together to form a complete contour of each subpopulation. The integration method can combine the coordinates of all contour points into a list, or use a more complex contour fitting algorithm, for example, the contour points can be fitted into a polygon or a spline curve.

[0147] Step S38: Optimize the subpopulation boundaries of the cell subpopulation contour data and assign subpopulation labels to obtain cell detection subpopulation data.

[0148] In an embodiment of the present invention, the subgroup boundary is adjusted according to the average expression level of the cells in the subgroup to ensure that the cells on the boundary have a high similarity with the cells in the subgroup. After completing the optimization of the subgroup boundary, it is necessary to assign a subgroup label to each cell subgroup. The subgroup label refers to the number or name of the subgroup to which each cell belongs. Different subgroups can be marked with numbers or characters, for example, subgroup 1, subgroup 2, and so on. The purpose of assigning labels is to facilitate subsequent subgroup analysis. For each cell, it is assigned according to the subgroup outline divided in step S37, and the subgroup label of the cell is recorded to obtain cell detection subgroup data.

[0149] Preferably, step S31 includes the following steps:

[0150] Step S311: performing density map smoothing processing according to the cell dynamic heterogeneous distribution map, and performing cytokine gradient vector calculation to generate cytokine gradient vector data;

[0151] Step S312: setting a gradient modulus threshold according to the cytokine gradient vector data to obtain a density gradient modulus threshold;

[0152] Step S313: performing gradient region judgment on the cytokine gradient vector data based on the density gradient modulus threshold, and marking the cell dynamic heterogeneous distribution map as low gradient region marking data when the cytokine gradient vector data is lower than the density gradient modulus threshold;

[0153] Step S314: when the cytokine gradient vector data is higher than or equal to the density gradient modulus threshold, the cell dynamic heterogeneous distribution map is marked as high gradient area marking data;

[0154] Step S315: performing gradient diffusion of density values ​​of neighboring points on the cell dynamic heterogeneous distribution map based on the high gradient region marking data through the cytokine gradient vector data to obtain an adjusted dynamic heterogeneous distribution map;

[0155] Step S316: Tracking the local density maximum point according to the adjusted dynamic heterogeneous distribution map and the low gradient area marking data, thereby obtaining the cell subpopulation center point data.

[0156] In an embodiment of the present invention, a method such as Gaussian filtering and median filtering is used to smooth the dynamic heterogeneous distribution map of cells. For example, Gaussian filtering convolves the image through a Gaussian kernel function to achieve a smoothing effect, and the width of the Gaussian kernel function determines the degree of smoothing. The smoothed image will be used as the input of the gradient vector calculation. The purpose of the cytokine gradient vector calculation is to calculate the direction and intensity of the change of the cytokine intensity around each pixel. The gradient vector represents the direction and magnitude of the fastest change in the cytokine intensity. The gradient vector can be obtained by calculating the partial derivatives of the image in the x direction and the y direction. For example, the Sobel operator, Prewitt operator or other gradient operators can be used for calculation. The calculated gradient vector includes the gradient component in the x direction and the gradient component in the y direction, and these two components can form a vector, indicating the gradient direction and magnitude of the change in the cytokine intensity. The threshold is set by a statistical method, for example, the mean and standard deviation of all gradient modulus values ​​can be calculated, and then the mean value plus several times the standard deviation is used as the threshold. The gradient modulus value of the point is obtained, and then the gradient modulus value is compared with the density gradient modulus value threshold. If the gradient modulus is less than the threshold, the pixel is marked as a low gradient area, which can be represented by a specific color or marker value. For example, the pixel value of the low gradient area can be set to 0, or other marker values ​​different from those of the high gradient area can be used. High gradient areas usually represent areas where the cytokine intensity changes dramatically, which usually correspond to the boundaries of cell subpopulations or the separation areas between cell subpopulations. The gradient modulus of the point is obtained, and then the gradient modulus is compared with the density gradient modulus threshold. If the gradient modulus is greater than or equal to the threshold, the pixel is marked as a high gradient area, which can be represented by a specific color or marker value. For each pixel in the high gradient area, the density value of the point is diffused to the neighboring pixels around it along the reverse direction of its gradient vector. The intensity of diffusion can be adjusted according to the modulus of the gradient vector. The larger the gradient modulus, the greater the intensity of diffusion. The range of diffusion can also be controlled within a certain neighborhood range. For example, a Gaussian kernel function can be used to simulate the range and intensity of diffusion. The purpose of gradient diffusion is to make the cell density in the high gradient area gather to the low gradient area, so that the low gradient area forms a local maximum point of the cell density value. Gradient diffusion can be performed multiple times in an iterative manner, and each iteration updates the density value of each pixel until the change in density value tends to stabilize. Select unvisited low-gradient area points in the low-gradient area marking data. Then, with this point as the center, search for cell density values ​​in its neighborhood. If the cell density value of this point is greater than the cell density values ​​of all points in its neighborhood, the point is marked as the local density maximum point and recorded as the center point of the cell subpopulation. A minimum local density value threshold can be set to ensure that only points with cell density values ​​greater than the threshold are considered to be the center points of cell subpopulations.For example, a gradient ascent method can be used, starting from an initial point and iteratively searching along the gradient direction until a local maximum is found. Alternatively, a neighborhood search method can be used to traverse each pixel point and then check the cell density value in its neighborhood to determine whether the point is a local maximum. The output of this step is the cell subpopulation center point data, which contains the coordinates of the cell subpopulation center point and the corresponding cell index information.

[0157] Preferably, step S4 comprises the following steps:

[0158] Step S41: using the cell detection subgroup data to determine the cell attribution of the cytokine secretion feature matrix, and performing subgroup cell clustering processing to generate cell detection factor subgroup clustering data;

[0159] Step S42: Calculating the subgroup secretion mean of the cell detection factor subgroup clustering data to generate cytokine subgroup secretion data;

[0160] Step S43: calculating the degree of subgroup variation of the cell detection factor subgroup clustering data using the cytokine subgroup secretion data to generate a subgroup cytokine variation coefficient;

[0161] Step S44: performing heterogeneity quantification processing according to the coefficient of variation of subgroup cytokines and the cytokine subgroup secretion data, thereby obtaining cytokine subgroup characteristic data.

[0162] As an example of the present invention, refer to Figure 3 As shown, Figure 1 Detailed implementation steps of step S4 in the embodiment are shown in the flowchart. In this embodiment, step S4 includes:

[0163] Step S41: using the cell detection subgroup data to determine the cell attribution of the cytokine secretion feature matrix, and performing subgroup cell clustering processing to generate cell detection factor subgroup clustering data;

[0164] In an embodiment of the present invention, after obtaining the cell detection subgroup data, it is necessary to use the data to perform cell attribution judgment on the cytokine secretion feature matrix. The purpose of cell attribution judgment is to correspond the subgroup label of each cell in the cell detection subgroup data to the cell in the cytokine secretion feature matrix. The cell detection subgroup data records the subgroup information to which each cell belongs. For example, cell A belongs to subgroup 1 and cell B belongs to subgroup 2. Cell attribution judgment is to add the subgroup label in the cell detection subgroup data to the cytokine secretion feature matrix according to the cell index. For example, a column is added to the cytokine secretion feature matrix to indicate the subgroup to which each cell belongs. After completing the cell attribution judgment, cell clustering processing needs to be performed for each cell subgroup. The purpose of subgroup cell clustering is to reveal the cell heterogeneity within the subgroup. For example, a subgroup contains multiple different cell subclasses. Clustering can use a variety of algorithms, for example, K-means clustering, hierarchical clustering, or other clustering algorithms can be used. When clustering, the number of cells in each subpopulation needs to be considered. If the number of cells is small, fewer clustering categories can be used, and vice versa, more clustering categories can be used. After clustering is completed, the cells in each subpopulation are divided into different subcategories and given new subcategory labels.

[0165] Step S42: Calculating the subgroup secretion mean of the cell detection factor subgroup clustering data to generate cytokine subgroup secretion data;

[0166] In an embodiment of the present invention, for each subgroup or subclass, all cells in the group are traversed, the amount of each cytokine secreted by each cell is accumulated, and then divided by the total number of cells in the group to obtain the mean cytokine secretion of the group. For example, assuming that subgroup 1 contains 3 cells, and the amounts of cytokine A secreted by these 3 cells are 2, 4, and 6, respectively, then the mean cytokine A secretion of subgroup 1 is (2+4+6) / 3=4. The subgroup secretion mean can reflect the overall characteristics of each subgroup or subclass in cytokine secretion, and can be used to compare the differences between different subgroups or subclasses. The calculated subgroup secretion mean can be stored as a matrix, each row of the matrix represents a subgroup or subclass, each column represents a cytokine, and the elements of the matrix represent the cytokine secretion mean of the corresponding subgroup or subclass.

[0167] Step S43: calculating the degree of subgroup variation of the cell detection factor subgroup clustering data using the cytokine subgroup secretion data to generate a subgroup cytokine variation coefficient;

[0168] In an embodiment of the present invention, for each subgroup or subclass, the degree of difference in cytokine secretion of the group of cells is calculated, and the coefficient of variation can be used to quantify the difference. The coefficient of variation refers to the standard deviation divided by the mean value, which is a normalized measure of the degree of variation and can be used to compare the degree of variation between different cytokines. The calculation method is that, for each subgroup or subclass, the standard deviation of each cytokine secreted by the group of cells is first calculated, and then the standard deviation is divided by the corresponding cytokine secretion mean to obtain the coefficient of variation of the group on each cytokine. For example, if the standard deviation of cytokine A secreted by subgroup 1 is 2, and the mean of cytokine A is 4, then the coefficient of variation of cytokine A of subgroup 1 is 2 / 4=0.5. The coefficient of variation of the subgroup can reflect the homogeneity of cells within the cell subgroup. The larger the coefficient of variation, the higher the heterogeneity of cells within the subgroup.

[0169] Step S44: performing heterogeneity quantification processing according to the coefficient of variation of subgroup cytokines and the cytokine subgroup secretion data, thereby obtaining cytokine subgroup characteristic data.

[0170] In an embodiment of the present invention, the cytokine subgroup secretion data obtained in step S42 and the coefficient of variation of the subgroup cytokine obtained in step S43 are combined to perform heterogeneity quantification. A variety of indices can be used to quantify heterogeneity, for example, the cell proportions of different clusters in each subgroup can be calculated, or the mean value of the coefficient of variation of the cytokine secretion amount in each subgroup can be calculated. Other information, such as the expression of cell surface markers, can also be combined to more comprehensively describe the subgroup characteristics. The cytokine subgroup characteristic data finally obtained can be used for subsequent analysis, such as comparing the differences between different subgroups, or studying the relationship between a subgroup and a disease. For example, the information such as the cytokine secretion mean, coefficient of variation, and cell proportion of different clusters of each subgroup can be integrated into a table as cytokine subgroup characteristic data.

[0171] Preferably, the present invention further provides a cytokine detection data processing system, which executes the cytokine detection data processing method as described above, and the cytokine detection data processing system comprises:

[0172] The detection data preprocessing module is used to obtain the original single-cell cytokine detection data; analyze the cytokine secretion amount of the original single-cell cytokine detection data to obtain the cytokine secretion vector data; construct a multi-cell-cytokine matrix based on the cytokine secretion vector data to obtain the cytokine secretion feature matrix;

[0173] The cytokine secretion analysis module is used to analyze the cytokine secretion pattern of the cytokine secretion feature matrix and generate cytokine secretion pattern data; the cytokine secretion pattern data is used to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and perform fractal degree color mapping to generate a cell dynamic heterogeneous distribution map;

[0174] The cell subpopulation center tracking module is used to track the cell subpopulation center according to the cell dynamic heterogeneous distribution map and generate the cell subpopulation center point data; perform cell neighborhood growth processing according to the cell subpopulation center point data to obtain the cell detection subpopulation data;

[0175] The cytokine heterogeneity clustering module is used to judge the cell attribution of the cytokine secretion feature matrix through the cell detection subgroup data, and perform subgroup cell clustering processing to generate cell detection factor subgroup clustering data; according to the heterogeneity quantitative processing of the cell detection factor subgroup clustering data, the cytokine subgroup characteristic data is obtained.

[0176] Preferably, a computer-readable storage medium stores a computer program, and when the computer program is executed, the method for processing cytokine detection data as described in any one of the above is implemented.

[0177] The present application is that, through cytokine secretion pattern analysis and typing degree difference calculation, different cytokine secretion patterns present in cell populations can be identified, and the typing degree of each cell in different patterns can be quantified. This pattern-based analysis method can not only identify cell subpopulations with similar secretion patterns, but also reveal functional differences between different subpopulations, such as pro-inflammatory and anti-inflammatory cell subpopulations. The cell dynamic heterogeneous distribution map generated by fractal degree color mapping shows the heterogeneous distribution of cytokine secretion patterns in cell populations in a visual way, and can more clearly show the distribution and mutual relationship of cell subpopulations, which is helpful to discover new cell subpopulations and potential biomarkers. Cell subpopulation center tracking and neighborhood growth processing realize automatic identification and boundary definition of cell subpopulations, and can more accurately identify cell subpopulations with biological significance. Cell subpopulations are automatically identified according to the intrinsic structure of data, which is closer to the real situation of biological systems. Through neighborhood growth, cells with similar secretion patterns can be gathered together to form clearer subpopulation boundaries, thereby more accurately quantifying the characteristics of each subpopulation. By performing cell attribution judgment and subgroup clustering processing, and associating cytokine secretion characteristics with cell subgroups, the present invention achieves the characteristic quantification of cytokine secretion heterogeneity. This method not only includes the average secretion level of each subgroup, but also reflects the degree of heterogeneity within the subgroup, thereby more comprehensively characterizing the characteristics of the subgroup. For example, a subgroup has a higher average secretion level of a certain cytokine, but its internal coefficient of variation is also larger, which indicates that there are different subgroups within the subgroup that secrete the cytokine, suggesting more refined functional differentiation.

[0178] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0179] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for processing cytokine detection data, characterized in that: The following steps are involved: Step S1: obtaining original single cell cytokine detection data; The original single-cell cytokine detection data is analyzed for cytokine secretion to obtain cytokine secretion vector data; a multi-cell-cytokine matrix is ​​constructed based on the cytokine secretion vector data to obtain a cytokine secretion feature matrix; Step S2: performing cytokine secretion pattern analysis on the cytokine secretion feature matrix to generate cytokine secretion pattern data; using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and performing fractal degree color mapping to generate a cell dynamic heterogeneous distribution map; Step S3: tracking the center of a cell subpopulation according to the cell dynamic heterogeneous distribution map to generate cell subpopulation center point data; performing cell neighborhood growth processing according to the cell subpopulation center point data to obtain cell detection subpopulation data; Step S4: using the cell detection subgroup data to determine the cell attribution of the cytokine secretion feature matrix, and performing subgroup cell clustering processing to generate cell detection factor subgroup clustering data; The heterogeneity of cell detection factor subgroup clustering data was quantitatively processed to obtain cytokine subgroup characteristic data.

2. The method for processing cytokine detection data according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: obtaining original single cell cytokine detection data; Step S12: preprocessing the original single cell cytokine detection data, and performing quality screening on the detection data to generate quality-controlled cytokine detection data; Step S13: calculating the total amount of cytokine expression for the cytokine detection data after quality control to generate total amount of cytokine expression data; Step S14: performing logarithmic transformation and normalization processing on the total cytokine expression data to generate normalized total cytokine expression data; Step S15: rank the cytokines according to the normalized total cytokine expression data to generate cell detection factor ranking data; Step S16: analyzing the secretion amount of different cytokines according to the cell detection factor ranking data to obtain cytokine secretion vector data; Step S17: construct a multi-cell-cytokine matrix based on the cytokine secretion vector data to obtain a cytokine secretion feature matrix.

3. The method for processing cytokine detection data according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing matrix transposition processing on the cytokine secretion characteristic matrix to obtain a transposed cytokine secretion characteristic matrix; Step S22: performing local similarity analysis between cells according to the transposed cytokine secretion feature matrix to generate local similarity fractal values ​​of cells; performing cytokine secretion pattern analysis according to the local similarity fractal values ​​of cells to generate cytokine secretion pattern data; Step S23: using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and performing nonlinear dimensionality reduction processing to generate optimized cell two-dimensional coordinate data; Step S24: performing kernel density distribution estimation based on the optimized cell two-dimensional coordinate data to generate cell density distribution estimation data; Step S25: Perform fractal color mapping based on the local similarity fractal value of the cells and the estimated data of cell density distribution to generate a dynamic heterogeneous distribution map of the cells.

4. The method for processing cytokine detection data according to claim 3, characterized in that: Step S22 includes the following steps: Step S221: calculating the Euclidean distance between cells based on the transposed cytokine secretion feature matrix to generate an initial cell distance matrix; Step S222: performing cytokine distribution feature statistics according to the transposed cytokine secretion feature matrix to generate cytokine distribution feature data; Step S223: setting a dynamic neighborhood K value according to the cytokine distribution characteristic data to obtain a dynamic neighborhood K value; Step S224: selecting candidate cell neighborhoods for the initial cell distance matrix using the dynamic neighborhood K value to obtain cytokine candidate neighborhood list data; Step S225: performing local cosine similarity calculation on the transposed cytokine secretion feature matrix through the cytokine candidate neighborhood list data, and performing neighborhood similarity iterative convergence to obtain local similarity data between cells; Step S226: Calculating the cell weighted fractal value according to the local similarity data between cells to generate the local similarity fractal value of the cell; Step S227: Based on the cytokine distribution feature data, the cytokine secretion pattern analysis is performed on the cytokine candidate neighborhood list data using the cell local similarity fractal value to generate cytokine secretion pattern data.

5. The method for processing cytokine detection data according to claim 3, characterized in that: Step S23 includes the following steps: Step S231: using the cytokine secretion pattern data to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and generating a similar fractal degree difference matrix between cells; Step S232: performing Gaussian kernel probability value conversion according to the inter-cell similarity fractal degree difference matrix, and performing secretion mode transition probability processing to generate a cell secretion mode transition probability matrix; Step S233: generating random two-dimensional spatial coordinates of cells according to the cell secretion mode transition probability matrix to obtain initial two-dimensional coordinate data of cells; Step S234: using the cell secretion mode transition probability matrix to perform gradient descent optimization on the initial two-dimensional coordinate data of the cell to generate optimized two-dimensional coordinate data of the cell.

6. The method for processing cytokine detection data according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: tracking the center of a cell subpopulation according to the cell dynamic heterogeneous distribution map to generate cell subpopulation center point data; Step S32: Initializing the seed points of the cell subpopulation center point data, and constructing a queue of seed points to be grown, thereby generating the queue data of seed points to be grown; Step S33: looping through the queue to be grown of each subgroup according to the queue data of the seed point to be grown, and when the queue is not empty, taking the seed point at the head of the queue as the current growing point to obtain the coordinate data of the current growing point; Step S34: performing neighborhood cell evaluation on the cell dynamic heterogeneous distribution map through the current growth point coordinate data to generate neighborhood cell evaluation data; Step S35: Based on the neighborhood cell evaluation data, neighborhood cell growth is judged using the preset seed point growth conditions, and the coordinates of the neighborhood cells that meet the seed point growth conditions are extended to the tail of the corresponding growth queue to generate the extended queue data to be grown; Step S36: marking the coordinates of neighboring cells that do not meet the seed point growth conditions as cell accesses, and performing subpopulation profile update processing on the current growth point coordinate data to obtain initial subpopulation profile data and visited cell data; Step S37: Perform growth iteration judgment on the expanded queue data to be grown through the accessed cell data. When the expanded queue data to be grown is not empty, return to step S33. When the expanded queue data to be grown is empty, perform growth region integration on the initial subpopulation profile data to obtain cell subpopulation profile data. Step S38: Optimize the subpopulation boundaries of the cell subpopulation contour data and assign subpopulation labels to obtain cell detection subpopulation data.

7. The method for processing cytokine detection data according to claim 6, characterized in that: Step S31 includes the following steps: Step S311: performing density map smoothing processing according to the cell dynamic heterogeneous distribution map, and performing cytokine gradient vector calculation to generate cytokine gradient vector data; Step S312: setting a gradient modulus threshold according to the cytokine gradient vector data to obtain a density gradient modulus threshold; Step S313: performing gradient region judgment on the cytokine gradient vector data based on the density gradient modulus threshold, and marking the cell dynamic heterogeneous distribution map as low gradient region marking data when the cytokine gradient vector data is lower than the density gradient modulus threshold; Step S314: when the cytokine gradient vector data is higher than or equal to the density gradient modulus threshold, the cell dynamic heterogeneous distribution map is marked as high gradient area marking data; Step S315: performing gradient diffusion of density values ​​of neighboring points on the cell dynamic heterogeneous distribution map based on the high gradient region marking data through the cytokine gradient vector data to obtain an adjusted dynamic heterogeneous distribution map; Step S316: Tracking the local density maximum point according to the adjusted dynamic heterogeneous distribution map and the low gradient area marking data, thereby obtaining the cell subpopulation center point data.

8. The method for processing cytokine detection data according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: using the cell detection subgroup data to determine the cell attribution of the cytokine secretion feature matrix, and performing subgroup cell clustering processing to generate cell detection factor subgroup clustering data; Step S42: Calculating the subgroup secretion mean of the cell detection factor subgroup clustering data to generate cytokine subgroup secretion data; Step S43: calculating the degree of subgroup variation of the cell detection factor subgroup clustering data using the cytokine subgroup secretion data to generate a subgroup cytokine variation coefficient; Step S44: performing heterogeneity quantification processing according to the coefficient of variation of subgroup cytokines and the cytokine subgroup secretion data, thereby obtaining cytokine subgroup characteristic data.

9. A system for processing cytokine detection data, characterized in that: For executing the method for processing cytokine detection data as claimed in claim 1, the cytokine detection data processing system comprises: The detection data preprocessing module is used to obtain the original single-cell cytokine detection data; analyze the cytokine secretion amount of the original single-cell cytokine detection data to obtain the cytokine secretion vector data; construct a multi-cell-cytokine matrix based on the cytokine secretion vector data to obtain the cytokine secretion feature matrix; The cytokine secretion analysis module is used to analyze the cytokine secretion pattern of the cytokine secretion feature matrix and generate cytokine secretion pattern data; the cytokine secretion pattern data is used to calculate the secretion pattern typing degree difference of the cytokine secretion vector data, and perform fractal degree color mapping to generate a cell dynamic heterogeneous distribution map; The cell subpopulation center tracking module is used to track the cell subpopulation center according to the cell dynamic heterogeneous distribution map and generate the cell subpopulation center point data; perform cell neighborhood growth processing according to the cell subpopulation center point data to obtain the cell detection subpopulation data; The cytokine heterogeneity clustering module is used to judge the cell attribution of the cytokine secretion feature matrix through the cell detection subgroup data, and perform subgroup cell clustering processing to generate cell detection factor subgroup clustering data; according to the heterogeneity quantitative processing of the cell detection factor subgroup clustering data, the cytokine subgroup characteristic data is obtained.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed, the method for processing cytokine detection data according to any one of claims 1 to 8 is implemented.

Images