Immune microenvironment difference detection and typing method and system for myeloblastoma

Through the differential detection and classification methods of medulloblastoma immune microenvironment, the immune microenvironment characteristics and potential therapeutic targets of different subtypes were identified, and the problem of insufficient understanding of the immune microenvironment of medulloblastoma in the prior art was solved. , new treatment and prognosis evaluation methods are provided.

CN119069004AActive Publication Date: 2024-12-03CHILDRENS HOSPITAL OF CHONGQING MEDICAL UNIV

Patent Information

Application Number
CN202411019680.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-12-03
Estimated Expiration
2044-07-29

AI Technical Summary

Technical Problem

The prior art has not yet fully understood the immune microenvironment of medulloblastoma, especially the immune characteristics and clinical significance of different immune subtypes, which affect the treatment and prognosis evaluation of the disease.

Method used

A differential detection and typing method of medulloblastoma in immune microenvironment, including pseudo-single-cell calculation of gene expression data, protein interaction network analysis, machine learning and single-cell omics analysis, through Netclass, ssGSEA, lasso regression and other technologies to identify different immune microenvironment subtypes and screen out genes with different prognosis.

Benefits of technology

The successful division of medulloblastoma into subtypes with different biological and clinical characteristics reveals its immune characteristics and potential therapeutic targets, providing new ideas and methods for the treatment and prognosis evaluation of the disease.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119069004A_ABST
    Figure CN119069004A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of biomedicine, and particularly discloses an immune microenvironment difference detection and typing method and system.The method comprises the steps that gene expression data of a sample suffering from medulloblastoma are collected and introduced into a protein interaction network, and then expression data of pseudo-single cells are combined, so that the immune microenvironment difference detection and typing of the medulloblastoma is achieved; unsupervised clustering is carried out through similarity network fusion analysis, different immune microenvironment subtypes in the medullary blastoma are determined, and then machine learning is carried out through l asso to carry out biological characteristic screening on the different immune microenvironment subtypes; and finally carrying out single-cell omics analysis. By adopting the technical scheme, through verification of immune microenvironment typing and marker screening, the immune characteristics and potential treatment targets of the medulloblastoma are obtained, and a new thought and method are provided for treatment and prognosis evaluation of diseases.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of biomedical technology and relates to a method and system for detecting and typing immune microenvironment differences of medulloblastoma. Background Art

[0002] Medulloblastoma (MB) is the most common childhood brain tumor in the posterior cranial fossa, accounting for about 25% of all childhood brain tumors. It is aggressive. Transcriptomic, genomic, epigenomic and proteomic analyses have shown that human MB is highly heterogeneous.

[0003] These tumors are divided into four more refined major subgroups with different outcomes and molecular characteristics: WNT (Wingless-related integration site, the WNT protein family is named after the "Wingless" gene of fruit flies and the "Int-1" gene of mice, which play important roles in morphogenesis, cell proliferation and differentiation, etc.), Sonic Hedgehog (SHH), Group 3 and Group 4. WNT and SHH usually have mutations in WNT and SHH pathway genes, respectively. GP3 subgroup tumors usually show MYC (Myelocytomatosis viral oncogene homolog, MYC is an important proto-oncogene that plays a key role in cell proliferation, differentiation and apoptosis) amplification and have the worst clinical outcomes. The genetic drivers of GP4 are still unclear.

[0004] Although traditional treatment options have made some progress, for some high-risk patients, the treatment effect is still unsatisfactory, and their survival rate and prognosis are still not optimistic. Therefore, it is particularly important to explore new treatment strategies and predictive markers.

[0005] Tumor heterogeneity refers to the fact that tumor tissue is composed of cell subpopulations with different expression profiles or biological functions, which leads to differences in tumor growth rate, invasion and metastasis ability, and sensitivity to drugs. This is a significant feature of malignant tumors. This heterogeneity is the main cause of tumor recurrence, metastasis, and drug resistance, and directly affects the clinical treatment effect. Therefore, analyzing tumor heterogeneity has important clinical diagnostic value. In-depth research on the formation and regulation mechanism of tumor heterogeneity will provide theoretical support for precise targeted treatment of tumors.

[0006] At present, there are two main theories about the causes of tumor heterogeneity: the clonal selection theory and the tumor stem cell theory. The clonal selection theory believes that tumor heterogeneity originates from the continued mutation of a single tumor cell population during development. In the process of adapting to the environment, only those cell populations that can survive and reproduce are retained, which eventually leads to the formation of tumor heterogeneity. The tumor stem cell theory believes that tumors are composed of a small group of tumor stem cells with self-renewal capabilities and cells with different degrees of differentiation, so tumors show heterogeneity of different cell types. Although the two theories have different explanatory angles, they both emphasize the important role of the tumor microenvironment (TME) in the formation of tumor heterogeneity.

[0007] The tumor microenvironment refers to the internal environment in which tumor cells occur and develop, including tumor cells themselves, immune cells, stromal cells, and the intercellular matrix, microvessels, and biological molecules infiltrating in the nearby area. In recent years, more and more studies have shown that the tumor immune microenvironment plays a key role in the occurrence, development, and treatment response of tumors. The tumor microenvironment not only affects the growth and metastasis of tumor cells, but is also closely related to the treatment effect and patient prognosis. Individualized treatment based on the tumor microenvironment has received more and more attention, especially immunotherapy has been widely used in the clinical treatment of various cancers and has achieved remarkable results. However, the understanding of the tumor microenvironment of medulloblastoma is still insufficient.

[0008] Therefore, a deep understanding of the cellular composition of the medulloblastoma microenvironment, the interaction between medulloblastoma cancer cells and other cells in the microenvironment, and the impact of medulloblastoma cancer cells on the microenvironment will help to discover biomarkers and potential therapeutic targets for the development of medulloblastoma, which is of great significance for understanding the pathogenesis of the tumor, predicting the disease progression of patients, and formulating individualized treatment plans.

[0009] However, the research on the immune microenvironment of medulloblastoma is still in its infancy, especially the understanding of the immune characteristics and clinical significance of medulloblastoma of different immune subtypes is not in-depth enough. Summary of the invention

[0010] The purpose of the present invention is to provide a method and system for detecting and typing the differences in the immune microenvironment of medulloblastoma, to obtain the immune characteristics and potential therapeutic targets of medulloblastoma, and to provide new ideas and methods for the treatment and prognosis evaluation of the disease.

[0011] In order to achieve the above object, the basic scheme of the present invention is: a method for detecting and typing the immune microenvironment difference of medulloblastoma, comprising the following steps:

[0012] S1, gene expression data of samples with medulloblastoma were collected and pseudo single-cell calculations were performed;

[0013] S2, using Netclass to generate a protein interaction network as an adjacency matrix, introducing gene expression data into the protein interaction network, and then combining pseudo-single-cellized gene expression data to perform unsupervised clustering through similarity network fusion analysis to determine different immune microenvironment subtypes in medulloblastoma;

[0014] S3, using lasso to perform machine learning to screen biological features of different immune microenvironment subtypes;

[0015] S4, Validation of immune microenvironment typing and biological signature screening by single-cell omics analysis.

[0016] The working principle and beneficial effects of this basic scheme are as follows: This technical scheme obtains the heterogeneity of the tumor immune microenvironment of MB patients of different subtypes by classifying the microenvironments of different subtypes of medulloblastoma, and combines single-cell sequencing data visualization to conduct in-depth research on the differences in their microenvironment landscapes.

[0017] MB is divided into different subtypes with different biological and clinical characteristics, and genes with different prognostic differences in different subtypes are screened. Different subtypes of medulloblastoma have different immune microenvironment subtypes. These subtypes have unique markers and functions, play their own unique roles in the development of tumors, and bring completely different outcomes to patients. Through immune microenvironment typing and marker screening, the immune characteristics and potential therapeutic targets of medulloblastoma are revealed, providing new ideas and methods for the treatment and prognosis evaluation of the disease.

[0018] Further, step S1 is specifically as follows:

[0019] Based on the ssGSEA pseudo single-cell analysis method, the gene expression data are compared with the given gene set, and the enrichment score of the gene set in each sample is calculated to simulate the composition of immune cells in the immune microenvironment of the tumor;

[0020] Based on the netClass analysis method, the information of protein interaction PPI network is used to integrate gene expression data and protein network topology.

[0021] The pseudo single-cell results obtained by the ssGSEA method were combined with gene expression data to perform similarity network fusion SNF analysis on the samples;

[0022] The survival package and ggplot2 package, ESTIMATE package, MCPcounter package, and msigdbr package in R language were used to visualize the survival data of different immune subtypes of MB patients, evaluate the abundance of tumor microenvironment and different cell types in the tumor microenvironment, and evaluate the gene set data related to biological processes;

[0023] Use the glmnet function in the glmnet package to perform multi-classification lasso regression analysis on gene expression data;

[0024] Based on the gene expression data, the R software package limma (3.56.2) was used to perform DEG analysis of differentially expressed genes between groups to obtain the differences in data expression between sample groups; the R software package GOplot (R 4.0.2) was used to draw bubble charts to observe pathway enrichment.

[0025] Identify different immune microenvironment subtypes and corresponding biological characteristics in medulloblastoma for subsequent analysis and processing.

[0026] Furthermore, based on the netClass analysis method, the information of the protein interaction network is used to integrate the gene expression data and the protein network topology. The specific steps are as follows:

[0027] Obtain protein interaction network data from public databases or literature, and construct a network containing the interaction relationships between proteins as an adjacency matrix;

[0028] Calculate the diffusion kernel: accept the adjacency matrix as input and calculate the diffusion kernel matrix according to the specified parameters. The specified parameters (p and a) are used to generate the diffusion kernel matrix through iterative calculation. When calculating the diffusion kernel matrix, the parameter is.adjacency is used to determine whether the input is an adjacency matrix or a Laplace matrix, and different calculation paths are selected.

[0029] Use the igraph library to handle graph operations, including creating undirected graphs and calculating Laplacian matrices;

[0030] When calculating the diffusion kernel matrix and the subsequent eigenvector multiplication, matrix operations are used to screen out the intersection with another data set, and then the diffusion kernel is calculated using part of the data in the intersection.

[0031] netClass is used to analyze gene expression data and identify biologically meaningful gene patterns in combination with the topological structure of protein interaction networks.

[0032] Furthermore, the pseudo single-cell results obtained by the ssGSEA method were combined with gene expression data to perform similar network fusion SNF analysis on the samples:

[0033] The complete matrix of expression after pseudogene removal, the ssGSEA matrix, was used as input;

[0034] The SNFtool R package (v2.2.0) was used with the number of neighbors K = 30, Gaussian kernel parameter alpha = 0.5, and number of iterations T = 10. Spectral clustering implemented by the SNF toolkit was run on the SNF fusion similarity matrix to obtain the most appropriate grouping;

[0035] SNF analysis was performed in different subtypes: in the SHH (sample number n = 233) subtype, three immune microenvironment subsets were obtained when k = 3;

[0036] In the GP3 (n=144) subtype, 2 subsets of immune microenvironments were obtained at k=2;

[0037] In the GP4 (n=326) subtype, three subsets of immune microenvironments were obtained when k=3.

[0038] Netclass analysis combined with SNF analysis can classify the tumor microenvironment of different subtypes of MB, and the operation is simple.

[0039] Furthermore, the survival package and ggplot2 package, ESTIMATE package, msigdbr package, and MCPcounter package in R language were used to visualize the survival data of MB patients, evaluate the tumor microenvironment based on gene expression data, obtain gene set data related to biological processes in the genomics database of species, and evaluate the abundance of different cell types in the tumor microenvironment as follows:

[0040] S11, prepare the data required for survival analysis, including the observation time futime and event status fustat of each sample, and the variable group used for grouping;

[0041] S12, use the survfit() function to fit the survival curve, and calculate the survival curve according to the specified observation time and event status and grouping variable group;

[0042] S13, use ggsurvplot() function to create a survival curve chart, set the parameters in the chart not to display the confidence interval (conf.int=F), display the risk table (risk.table=T), not add the total patient survival curve (add.all=F), customize the color palette (palette="Dark2"), adjust the title, axis title, font size and style, and the range of the coordinate axis;

[0043] S21, prepare the gene expression data file, use the outputGCT() function to convert the data into the input format required by the ESTIMATE package, and save it as a new file (in.gct.file);

[0044] S22, use the filterCommonGenes() function to filter out common genes from the raw data and save them as a new input file, use the estimateScore() function to calculate the ESTIMATE score of the tumor sample, and save the score result as a GCT format file (out.score.file);

[0045] S23, then use the plotPurity() function to visualize the purity score of the sample, use the read.table() function to read the score result file and convert it into a data frame format, and finally save the result as a text file (ESTIMATE_score.txt);

[0046] S31, use the msigdbr() function to obtain the gene set data of the C5 category from the species' genomics database and save it in the GO_df_all data frame;

[0047] The required columns (gs_name, gene_symbol, gs_exact_source, and gs_subcat) were selected using the dplyr::select() function, and the unnecessary data were filtered out based on the subcategory (gs_subcat) to obtain the final GO gene set data (GO_df);

[0048] S32, using the split() function to group the gene set data according to the GO entry name gs_name to obtain a gene list (go_list);

[0049] The gene expression data to be analyzed (rt1) is subjected to gene set variability analysis (GSVA) ​​using the gsva() function, the gene set variability score of each GO entry is calculated (gsva_mat), and the results are saved in a text file (gsva_hallmark.txt);

[0050] S33, set parameters: use Gaussian kernel density estimation function (kcdf = "Gaussian"), and call all available cores (parallel.sz = parallel::detectCores());

[0051] S41, read the data required for MCPcounter evaluation, including gene expression data of cell types (MCP_counter_tz.txt) and gene information data (MCP_counter_gene.txt);

[0052] S42, use the MCPcounter.estimate() function to estimate the abundance of cell types in the tumor sample, specifying the gene expression data of the cell type (probesets parameter), the gene information of the cell type (genes parameter), and the feature type (featuresType parameter);

[0053] S43, save the evaluation results in the results object and visualize them to understand the abundance of different cells in the immune microenvironment.

[0054] Multiple packages in R language are used for data analysis, which is easy to operate.

[0055] Furthermore, in step S3, the method of performing multi-classification lasso regression analysis on gene expression data using the glmnet function in the glmnet package is:

[0056] Use the glmnet function to fit a multinomial logistic regression model, where the independent variable x is the gene expression of the sample, the dependent variable y is the classification label of the sample, the family parameter is used to specify the distribution type of the multinomial as multinomial, and the type.multinomial parameter is used to set the combination mode;

[0057] Use the plot function to visualize the fitted model, set the x-axis to the coefficient lambda of the regularization penalty term and mark the value of each coefficient;

[0058] Use the cv.glmnet function to cross-validate the lasso regression model, select the optimal lambda value, and refit the lasso regression model using the glmnet function based on the optimal lambda value. The alpha parameter is set to 1 to indicate the use of lasso regression, the lambda parameter is set to the optimal value obtained by cross-validation, and the family parameter and type.multinomial parameter are the same as before.

[0059] Extract the genes whose coefficients in the lasso regression model are not 0 and save them as gene_min.

[0060] The glmnet function in the glmnet package was used to perform multi-classification lasso regression analysis on gene expression data.

[0061] Furthermore, based on the gene expression data, the R software package limma (3.56.2) was used to perform differentially expressed gene DEG analysis between groups to obtain the difference in data expression between sample groups; the R software package GOplot (R 4.0.2) was used to draw a bubble chart to observe the pathway enrichment. The specific steps are as follows:

[0062] Differentially expressed genes were analyzed using the limma package:

[0063] Design matrix creation: Use the model.matrix function to create a design matrix (design), which is used to describe the treatment groups (group) in the experimental design;

[0064] Contrast matrix creation: Use the makeContrasts function to create a control matrix (contr.matrix) and define the contrast between the two control groups (C1VSC2), where C1 and C2 are the two control groups in the design matrix;

[0065] Linear model fitting: Use the lmFit function to fit a linear model (vfit) to the expression data (y) and the design matrix (design);

[0066] Contrast fitting: Use the contrasts.fit function to perform contrast fitting on the linear model and pass in the contrast matrix (contr.matrix);

[0067] Bayesian estimation: Use the eBayes function to perform Bayesian estimation on the fitted model to obtain the significance of the gene;

[0068] Diagnostic plot drawing: The plotSA function is used to draw the diagnostic plot of the model to evaluate the suitability and stability of the model;

[0069] Significance test: Use the decideTests function to perform a significance test on the model and output summary information of the significance test results of the gene;

[0070] Differentially expressed gene analysis: Use the topTable function to obtain the differential expression of all genes based on the results of Bayesian estimation, and then filter out genes with significant differential expression based on given thresholds (padj and foldChange);

[0071] Result output: Write the significantly differentially expressed genes after screening into a CSV file;

[0072] DEGs functional annotation and enrichment analysis, R software package clusterProfiler was used to perform GO functional annotation and KEGG enrichment analysis on the screened differentially expressed genes (DEGs). The enrichment results were considered significant only when the P value was less than 0.05, specifically:

[0073] Gene ID conversion: Use the bitr function to convert gene symbols from SYMBOL to ENTREZID;

[0074] GO analysis: Use the enrichGO function to perform GO enrichment analysis, specify the type of gene ID, p-value and q-value threshold parameters, and save the results as CCGO.rdata file;

[0075] KEGG analysis: Use the enrichKEGG function to perform KEGG enrichment analysis, specify parameters such as species, p-value and q-value thresholds, and save the results as CCKEGG.rdata file;

[0076] GO / KEGG enrichment visualization: Use barplot and dotplot functions to visualize GO and KEGG enrichment results and save the images as pdf files;

[0077] Gene set comparison analysis: Use the enrichKEGG function to perform KEGG enrichment analysis on genes between different groups, compare the KEGG enrichment results of different groups, and visualize the results.

[0078] Based on the gene expression data, the R software package limma (3.56.2) was used to perform differentially expressed gene DEG analysis between groups to obtain the difference in data expression between sample groups. The R software package GOplot (R 4.0.2) was used to draw a bubble chart to observe the enrichment of pathways.

[0079] Further, the method for verifying the immune microenvironment typing and biological feature screening by single cell omics analysis in step S4 is:

[0080] Collect single-cell data of samples with medulloblastoma and use Seurat to perform quality control and dimensionality reduction clustering of single-cell data;

[0081] Combine the collected annotation information to annotate the cell subpopulations of dimensionality reduction clustering;

[0082] Use the R package scRNAtoolVis to obtain the distribution status of specific marker maker in single-cell subpopulations and the differences in single-cell levels of medulloblastoma patients with different immune microenvironment subtypes;

[0083] The method for collecting single-cell data of samples with medulloblastoma and using Seurat to perform quality control and dimensionality reduction clustering of single-cell data is as follows:

[0084] For each sample, the number of UMIs, total number of genes, number of mitochondrial genes, and number of ribosomal genes were counted, and cells with a total number of genes greater than 6000 or a number of genes less than 200, as well as cells with a mitochondrial gene ratio greater than 30%, were filtered out;

[0085] For each sample, the top 2000 genes with the largest variation were determined based on the mean and dispersion of all genes for integrated analysis to eliminate batch effects between samples. The integrated data were subjected to principal component analysis (PCA) dimensionality reduction, and the top 20 principal components were retained to capture the main changes in the data.

[0086] Seurat pipeline was used to detect hgv, and the average expression and dispersion of each gene were calculated. Genes were placed in bins and the z-score of the dispersion within each bin was calculated. A z-score of 0.5 was used as the cutoff value for dispersion, a lower limit of 0.0125 and a high cutoff of 3.0 were used as the average expression. Principal component analysis (PCA) was used for linear dimensionality reduction, and the ElbowPlot method and Jackstraw method were used to select statistically significant principal components. Seurat was used to visualize the clustering based on UMAP. The specific steps are as follows:

[0087] JackStraw: Use the JackStraw method to evaluate the PCA results and identify significant principal components;

[0088] ScoreJackStraw: Score the results of the JackStraw method and determine the principal components with statistical significance;

[0089] JackStrawPlot and ElbowPlot: visualize the results of the JackStraw method and select the number of principal components to be retained. JackStrawPlot is used to observe the p-value of each principal component, while ElbowPlot is used to observe the changes in the degree of variance explanation, usually selecting the number of principal components at the "elbow";

[0090] Cell clustering:

[0091] FindNeighbors: Calculate the neighbor relationship between cells based on the results of PCA;

[0092] FindClusters: Based on the K-nearest neighbor graph, a clustering algorithm is used to divide cells into different cell clusters;

[0093] Idents and table: View the clustering results. Use Idents to view the cluster to which each cell belongs. Table is used to count the number of cells in each cluster.

[0094] UMAP Dimensionality Reduction and Visualization:

[0095] RunUMAP: Use the UMAP algorithm to reduce the dimensionality of the data. UMAP uses a stochastic gradient descent optimization algorithm, an approximation method based on the minimum spanning tree and Gaussian mixture model, and a distance-based weighting function to adjust the similarity weights between different data points to map high-dimensional data into two-dimensional or three-dimensional space;

[0096] Minimize the loss function between the distance between data points in high-dimensional space and the distance between corresponding points in low-dimensional space, and optimize the data representation in low-dimensional space;

[0097] DimPlot: Based on the dimension reduction of UMAP, it draws the distribution map of cells and intuitively shows the relationship and distribution of different cell clusters;

[0098] The UMAP clustering visualization results showed that the cells from 26 samples of SHH, GP3, and GP4 were clustered into 21 (SHH), 18 (GP3), and 15 (GP4) subpopulations, respectively.

[0099] Single-cell data of samples with medulloblastoma were collected, and Seurat was used to perform quality control and dimensionality reduction clustering of the single-cell data for ease of use.

[0100] Furthermore, combined with the collected annotation information, the method for annotating the cell subpopulations of dimensionality reduction clustering is as follows:

[0101] For MB patient samples and a single matched sample from relapse, cells that passed quality control for scRNA sequencing were projected into 2D UMAP plots, and cell subpopulations in this experiment were manually annotated based on marker genes from the Cell Marker database and other brain tumor-related literature.

[0102] Combined with the collected annotation information, the cell subpopulations of dimensionality reduction clustering are annotated.

[0103] The present invention also provides a system for detecting and typing the immune microenvironment differences of medulloblastoma, comprising a processing unit, wherein the processing unit executes the method of the present invention to detect and type the immune microenvironment differences of medulloblastoma.

[0104] Through the processing unit, this system obtains the distribution status of the specific marker maker in the single-cell subpopulation, as well as the differences in the single-cell level of medulloblastoma patients with different immune microenvironment subtypes, providing new ideas and methods for the treatment and prognosis evaluation of the disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] Figure 1 It is a schematic diagram of the process of the method for detecting and typing the immune microenvironment differences of medulloblastoma of the present invention. DETAILED DESCRIPTION

[0106] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0107] In the description of the present invention, it is necessary to understand that the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0108] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0109] The present invention discloses a method for detecting and typing the difference in the immune microenvironment of medulloblastoma, exploring the immune microenvironment of medulloblastoma and typing it, in order to provide strong support for the formulation of individualized treatment strategies, which can be used for early diagnosis, prognosis prediction, development of specific treatment, monitoring of treatment effects, etc. Figure 1 As shown, the method for detecting and classifying differences in the immune microenvironment of medulloblastoma includes the following steps:

[0110] S1, gene expression data of samples with medulloblastoma were collected and pseudo single-cell calculations were performed;

[0111] S2, using Netclass (NetClass is a network-based classification method that converts network structure information into feature vectors and then uses traditional machine learning classifiers for prediction. Its main goal is to use the topological characteristics of the network to improve classification performance. The core idea of ​​the NetClass algorithm is to extract structural information from network data into feature vectors and then use these feature vectors for classification) to generate a protein–protein interaction (PPI) network as an adjacency matrix, introduce gene expression data into the protein interaction network, and then combine pseudo-single-cell gene expression data to perform unsupervised clustering through similarity network fusion analysis to determine different immune microenvironment subtypes in medulloblastoma;

[0112] S3, using lasso to perform machine learning to screen biological features of different immune microenvironment subtypes;

[0113] S4, Validation of immune microenvironment typing and biological signature screening by single-cell omics analysis.

[0114] In a preferred embodiment of the present invention, step S1 specifically comprises:

[0115] Based on the ssGSEA (Single-sample Gene Set Enrichment Analysis) pseudo single-cell analysis method, the gene expression data is compared with the given gene set, and the enrichment score of the gene set in each sample is calculated to simulate the composition of immune cells in the immune microenvironment of the tumor, specifically:

[0116] Sort the gene expression data of each sample from high to low (the gene expression data of each sample is usually sorted from high to low according to the expression level of the gene in the sample. This sorting reflects the expression intensity of the gene in the sample), and calculate the cumulative distribution function CDF (Cumulative Distribution Function, referred to as CDF, is an important concept in probability theory and statistics. For a random variable X, its cumulative distribution function F(x) is defined as the probability that the random variable X is less than or equal to x, that is, F(x) = P(X≤x)) of each gene set, including the gene set CDF (calculated based on the position of the genes in the gene set in the sorted list) and the background gene CDF (based on the cumulative distribution calculated based on the expression level of all genes);

[0117] The enrichment score is calculated by comparing the gene set CDF with the background gene CDF, where NNN is the total number of genes and CDF represents the cumulative distribution function:

[0118] Obtain expression data for gene sets and background genes: This can be microarray data, RNA-seq data, or the results of other gene expression measurements. Make sure that both the gene set and background gene expression values ​​are prepared.

[0119] Calculate the CDF of the gene set and background genes: For the gene set and background genes, sort them from small to large according to their expression values. Then, calculate the cumulative distribution value of each gene, that is, divide the number of genes less than or equal to the expression value of the gene by the total number of genes, so that the CDF curves of the gene set and background genes can be obtained.

[0120] Compare gene set CDF and background gene CDF: You can use visualization methods, such as drawing CDF curves, to visually compare the distribution of gene sets and background genes. Observe whether the gene set CDF is above or below the background gene CDF, and the degree of difference between them.

[0121] Calculate the enrichment score: Calculate the area difference between the gene set CDF and the background gene CDF. The specific calculation method can be selected according to the specific situation. A simple method is to calculate the integrated area between the two CDF curves as the enrichment score. The area can be calculated using numerical integration or other appropriate methods.

[0122] The enrichment scores were normalized to make them comparable across different gene sets, and they were assigned to the most likely cell, in order to identify the cell type that each cell most likely represents at the single-cell level; “most likely” here means that the cell has the highest enrichment score on a certain gene set and is therefore most likely to represent the cell type associated with that gene set.

[0123] Based on the netClass analysis method, the information of protein-protein interaction (PPI) network is used to integrate gene expression data and protein network topology.

[0124] The pseudo single-cell results obtained by the ssGSEA method were combined with the gene expression data to perform a similarity network fusion (SNF) analysis on the samples.

[0125] The survival package in R language was used (in R language: the "survival" package is mainly used for survival analysis. Survival analysis is a statistical method for processing event time data, which is often used in medical research (such as patient survival time), engineering reliability research (such as equipment failure time) and other fields) and ggplot2 package (the "ggplot2" package is a powerful tool for creating high-quality data visualization. It is based on the concept of "graphic grammar" and allows users to build complex graphics in a hierarchical manner), ESTIMATE package (ESTIMATE package is mainly used to evaluate the level of matrix and immune cell infiltration in tumor tissue. It provides researchers with tumor purity, the level of matrix cells present and the level of immune cells in tumor tissue based on gene expression data. The MCPcounter package (the MCPcounter package can use transcriptome data to quantify the absolute abundance of 8 immune cells and 2 stromal cells in heterogeneous tissues) and the msigdbr package (the msigdbr package is mainly used to evaluate the stromal and immune cell infiltration levels in tumor tissues. It provides researchers with tumor purity, the level of stromal cells present, and the score of immune cell infiltration levels in tumor tissues based on gene expression data). The survival data of different immune subtypes of MB (Medulloblastoma) patients were visualized and analyzed, the abundance of tumor microenvironment and different cell types in the tumor microenvironment were evaluated, and the gene set data related to biological processes were evaluated.

[0126] Use the glmnet function in the glmnet package (the "glmnet" package is used in the R language to fit the generalized linear model) to perform multi-classification lasso regression analysis on the gene expression data;

[0127] Based on gene expression data, the R software package limma ("limma" is a powerful software package in R language that is widely used for gene expression data analysis) (3.56.2) was used to perform differentially expressed gene DEG analysis between groups to obtain the difference in data expression between sample groups; the R software package GOplot (R4.0.2) was used to draw bubble charts to observe pathway enrichment.

[0128] In a preferred embodiment of the present invention, based on the netClass analysis method, the information of the protein interaction network is used to integrate the gene expression data and the protein network topology structure, and the specific steps are as follows:

[0129] Construction of protein interaction network: Obtain data of protein interaction network from public databases or literature, and construct a network containing the interaction relationships between proteins as an adjacency matrix;

[0130] Data processing, including data screening, dimension adjustment and other operations, to ensure the accuracy and effectiveness of subsequent calculations.

[0131] Calculate the diffusion kernel: accept the adjacency matrix as input, and calculate the diffusion kernel matrix according to the specified parameters. The specified parameters (p and a) are used to generate the diffusion kernel matrix through iterative calculation. When calculating the diffusion kernel matrix, the parameter is.adjacency (is.adjacency is a parameter set or provided by the user according to the specific situation, used to indicate whether a matrix is ​​an adjacency matrix) is used to determine whether the input is an adjacency matrix or a Laplace matrix, and different calculation paths are selected (if it is an adjacency matrix, the corresponding calculation is performed according to the node connection relationship represented by the adjacency matrix; if it is a Laplace matrix, the Laplace matrix is ​​calculated using its unique properties (such as the relationship between the degree matrix and the adjacency matrix, etc.));

[0132] Use the igraph library to handle graph operations, including creating undirected graphs and calculating Laplacian matrices;

[0133] When calculating the diffusion kernel matrix and the subsequent eigenvector multiplication, matrix operations are used to screen out the intersection with another data set, and then the diffusion kernel is calculated using part of the data in the intersection.

[0134] In a preferred embodiment of the present invention, the pseudo single cell results obtained by the ssGSEA method are combined with the gene expression data to perform similarity network fusion (SNF) analysis on the samples as follows:

[0135] The SNF method does not require any prior feature selection, so the complete matrix of expression after pseudogene removal, the ssGSEA matrix, is used as input;

[0136] The SNFtool R package (v2.2.0) was used with the number of neighbors K = 30, Gaussian kernel parameter alpha = 0.5, and number of iterations T = 10. Spectral clustering implemented by the SNF toolkit was run on the SNF fusion similarity matrix to obtain the most appropriate grouping;

[0137] SNF analysis was performed in different subtypes: in the SHH (Sonic Hedgehog) subtype (sample size n = 233, the best grouping recommended by SNF, this parameter is the cluster number parameter in the spectral clustering algorithm. During the spectral clustering process, the algorithm will divide the data into K clusters), 3 immune microenvironment subgroups were obtained when k = 3;

[0138] In the GP3 (n=144) subtype, 2 subsets of immune microenvironments were obtained at k=2;

[0139] In the GP4 (n=326) subtype, three subsets of immune microenvironments were obtained when k=3.

[0140] K is the number of neighbors of each node when building a similarity network. In SNF (Similarity Network Fusion), each sample node is connected to its most similar K sample nodes to build a K-nearest neighbors graph. Setting K = 30 means that each sample is connected to its 30 most similar samples. Choosing a suitable K value is important because it affects the sparsity of the network and the final clustering results.

[0141] Alpha is a parameter that controls the width of the Gaussian kernel. When calculating the similarity matrix, the Gaussian kernel function is used to measure the similarity between samples. Specifically, the elements of the similarity matrix are calculated using the Gaussian kernel function. Setting alpha = 0.5 means that 0.5 is used as the parameter of the Gaussian kernel function when calculating the similarity. A larger alpha value will make the similarity values ​​more concentrated, while a smaller alpha value will make the similarity values ​​more dispersed.

[0142] T is the number of iterations in the SNF algorithm. In the SNF process, the information from different data sources is fused by updating the similarity matrix multiple times. In each iteration, the current similarity matrix is ​​fused and updated with the similarity matrices of other data sources. Setting T = 10 means that 10 iterations are performed to update and fuse the similarity matrix. Generally speaking, more iterations can make the fusion result more stable and accurate, but it will also increase the calculation time.

[0143] In a preferred embodiment of the present invention, the survival package and ggplot2 package, ESTIMATE package, msigdbr package, and MCPcounter package in R language are used to respectively visualize the survival data of MB patients, evaluate the tumor microenvironment based on the gene expression data, obtain the gene set data related to the biological process in the genomic database of the species, and evaluate the abundance of different cell types in the tumor microenvironment in the following steps:

[0144] S11, prepare the data required for survival analysis, including the observation time futime and event status fustat of each sample, and the variable group used for grouping;

[0145] S12, use the survfit() function to fit the survival curve, and calculate the survival curve according to the specified observation time and event status and grouping variable group;

[0146] S13, use ggsurvplot() function to create a survival curve chart, set the parameters in the chart not to display the confidence interval (conf.int=F), display the risk table (risk.table=T), not add the total patient survival curve (add.all=F), and customize the color palette (palette="Dark2"), adjust the title, axis title, font size and style, and the range of the coordinate axis, to help understand the survival status of different groups and provide a basis for subsequent data interpretation and clinical research;

[0147] S21, prepare the gene expression data file, use the outputGCT() function to convert the data into the input format required by the ESTIMATE package, and save it as a new file (in.gct.file);

[0148] S22, use the filterCommonGenes() function to filter out common genes from the raw data and save them as a new input file, use the estimateScore() function to calculate the ESTIMATE score of the tumor sample, and save the score result as a GCT format file (out.score.file);

[0149] S23, then use the plotPurity() function to visualize the purity score of the sample, use the read.table() function to read the score result file and convert it into a data frame format, and finally save the result as a text file (ESTIMATE_score.txt) for subsequent analysis and visualization;

[0150] S31, use the msigdbr() function to obtain the gene set data of the C5 category from the species' genomics database and save it in the GO_df_all data frame;

[0151] The required columns (gs_name, gene_symbol, gs_exact_source, and gs_subcat) were selected using the dplyr::select() function, and the unnecessary data were filtered out based on the subcategory (gs_subcat) to obtain the final GO gene set data (GO_df);

[0152] S32, use the split() function to group the gene set data according to gs_name (GO entry name) to obtain a gene list (go_list);

[0153] The gene expression data to be analyzed (rt1) was subjected to gene set variability analysis (GSVA) ​​using the gsva() function, the gene set variability score for each GO entry was calculated (gsva_mat), and the results were saved in a text file (gsva_hallmark.txt);

[0154] S33, set parameters: use the Gaussian kernel density estimation function (kcdf = "Gaussian") and call all available cores (parallel.sz = parallel::detectCores()); these analysis steps can help understand the relative expression levels of gene sets in samples and reveal the activity levels of different biological processes in samples.

[0155] S41, read the data required for MCPcounter (Microenvironment Cell Populations-counter) evaluation, including gene expression data of cell types (MCP_counter_tz.txt) and gene information data (MCP_counter_gene.txt);

[0156] S42, use the MCPcounter.estimate() function to estimate the abundance of cell types in the tumor sample, specifying the gene expression data of the cell type (probesets parameter), the gene information of the cell type (genes parameter), and the feature type (featuresType parameter);

[0157] S43, save the evaluation results in the results object and visualize them to understand the abundance of different cells in the immune microenvironment.

[0158] In a preferred embodiment of the present invention, in step S3, the method of performing multi-classification lasso regression analysis (Least Absolute Shrinkage and Selection Operator Regression, a regularization method for linear regression) on the gene expression data using the glmnet function in the glmnet package is:

[0159] Use the glmnet function to fit a multinomial logistic regression model, where the independent variable x is the gene expression of the sample, the dependent variable y is the classification label of the sample, and the family parameter (in the glmnet package of R language, the family parameter is used to specify the type of regression model. For example, for Lasso regression, by setting the family parameter, you can choose different distribution types to fit the corresponding model) specifies the distribution type of the multinomial as multinomial (when using the glmnet package for regression analysis, "multinomial" is usually used to specify the type of regression model as multi-classification logistic regression (MultinomialLogistic Regression)), and the type.multinomial parameter (the type.multinomial parameter is used to specify the type of processing method in multi-classification problems. In the glmnet package, it has two optional values: "ungrouped" (default) and "grouped") sets the combination method;

[0160] Use the plot function to visualize the fitted model, set the x-axis to the coefficient lambda of the regularization penalty term and mark the value of each coefficient;

[0161] Use the cv.glmnet function to cross-validate the lasso regression model, select the optimal lambda value, and refit the lasso regression model using the glmnet function based on the optimal lambda value. The alpha parameter is set to 1 to indicate the use of lasso regression, the lambda parameter is set to the optimal value obtained by cross-validation, and the family parameter and type.multinomial parameter are the same as before.

[0162] Extract the genes whose coefficients in the lasso regression model are not 0 and save them as gene_min, identifying the characteristic genes that have a significant impact on the classification results in the multi-classification problem.

[0163] In a preferred embodiment of the present invention, based on gene expression data, R software package limma (3.56.2) was used to perform differentially expressed gene DEG analysis between groups to obtain the difference in data expression between sample groups; R software package GOplot (R4.0.2) was used to draw a bubble chart to observe the enrichment of pathways. The specific steps are as follows:

[0164] Differentially expressed genes were analyzed using the limma package:

[0165] Design matrix creation: Use the model.matrix function to create a design matrix (design), which is used to describe the treatment groups (group) in the experimental design;

[0166] Contrast matrix creation: Use the makeContrasts function to create a control matrix (contr.matrix) and define the contrast between the two control groups (C1VSC2), where C1 and C2 are the two control groups in the design matrix;

[0167] Linear model fitting: Use the lmFit function to fit a linear model (vfit) to the expression data (y) and the design matrix (design);

[0168] Contrast fitting: Use the contrasts.fit function to perform contrast fitting on the linear model and pass in the contrast matrix (contr.matrix);

[0169] Bayesian estimation: Use the eBayes function to perform Bayesian estimation on the fitted model to obtain the significance of the gene;

[0170] Diagnostic plot drawing: The plotSA function is used to draw the diagnostic plot of the model to evaluate the suitability and stability of the model;

[0171] Significance test: Use the decideTests function to perform a significance test on the model and output summary information of the significance test results of the gene;

[0172] Differentially expressed gene analysis: Use the topTable function to obtain the differential expression of all genes based on the results of Bayesian estimation, and then filter out genes with significant differential expression based on given thresholds (padj and foldChange);

[0173] Result output: Write the significantly differentially expressed genes after screening into a CSV file;

[0174] Functional annotation and enrichment analysis of DEGs (Different express genes) were performed using the R software package clusterProfiler to perform GO functional annotation and KEGG enrichment analysis on the screened differentially expressed genes (DEGs). The enrichment results were considered significant only when the P value was less than 0.05. Specifically:

[0175] Gene ID conversion: Use the bitr function to convert gene symbols from SYMBOL to ENTREZID;

[0176] GO (Gene Ontology) analysis: Use the enrichGO function to perform GO enrichment analysis, specify the type of gene ID, p-value and q-value threshold parameters, and save the results as CCGO.rdata file;

[0177] KEGG (Kyoto Encyclopedia of Genes and Genomes) analysis: Use the enrichKEGG function to perform KEGG enrichment analysis, specify parameters such as species, p-value and q-value thresholds, and save the results as CCKEGG.rdata file;

[0178] GO / KEGG enrichment visualization: Use barplot and dotplot functions to visualize GO and KEGG enrichment results and save the images as pdf files;

[0179] Gene set comparison analysis: Use the enrichKEGG function to perform KEGG enrichment analysis on genes between different groups, compare the KEGG enrichment results of different groups, and visualize the results.

[0180] For example, using SNF, the tumor microenvironment of MB samples of the SHH subtype was further divided into three subtypes: C1, C2, and C3. Among them, the C2 subtype was significantly enriched in activated CD8 T cells, γδT cells, activated CD4 T cells, and type II T helper cells, and its expression was stronger than the other two subtypes; most immune cells such as macrophages, natural killer cells, central memory CD4 T cells, type I T helper cells, activated dendritic cells, immunosuppressive myeloid cells, central memory CD8 T cells, and regulatory T cells were enriched in C2 and C3 subtypes, but the expression of C3 subtype was stronger than that of C2 subtype; C1 subtype had almost no immune cells, and was only slightly expressed in a small number of immune cells such as immature dendritic cells, CD56 bright natural killer cells, CD56 dark natural killer cells, and plasmacytoid dendritic cells, showing an overall immune cold state.

[0181] Survival analysis showed that the survival rates of different subtypes were significantly different, among which the survival of C2 subtype was the worst, and the overall survival rate of C1 subtype was the highest. The patients of C1 subtype were obviously older, mainly adults, the patients of C3 subtype were mainly infants, and the patients of C2 subtype were mainly children, but the gender distribution difference was not obvious.

[0182] The immune microenvironment of GP3 MB was divided into two subtypes, C1 and C2, by SNF. For γδT cells, type II T helper cells, activated CD8 T cells, activated CD4 T cells and memory B cells, the C2 subtype was significantly enriched; while for CD56 bright natural killer cells, effector memory CD4 T cells, plasmacytoid dendritic cells, immunosuppressive myeloid cells (immunosuppressive myeloid-derived cells), neutrophils, central memory CD4 T cells, natural killer cells, eosinophils, macrophages, etc., both subtypes were expressed, but the expression of C2 subtype was slightly stronger than that of C1 subtype; both subtypes were also expressed for activated B cells, monocytes, type 17 T helper cells, immature B cells, T follicular helper cells, mast cells, and type I T helper cells, but the expression of C1 subtype was stronger than that of C2 subtype.

[0183] In the survival analysis, the survival of C2 subtype within 100 months was significantly worse than that of C1.

[0184] The immune microenvironment of GP4 MB was divided into three subtypes: C1, C2, and C3 by SNF. Mast cells, type 17 T helper cells, activated B cells, monocytes, CD56 bright natural killer cells, CD56 dark natural killer cells, plasmacytoid dendritic cells and effector memory CD4 T cells are enriched in both C1 and C3 subtypes, with similar enrichment levels, but are almost not expressed in C2 subtype; memory B cells, activated CD8 T cells, activated CD4 T cells, type II T helper cells, macrophages, T follicular helper cells, eosinophils and immature dendritic cells are expressed in all three subtypes, but the C3 subtype is the weakest, followed by the C2 subtype, and the C1 subtype is the strongest; the remaining immune cell types including natural killer T cells, type I T helper cells, immature B cells, neutrophil populations, activated dendritic cells, effector memory CD8 T cells, etc. are most strongly expressed in the C1 subtype and are almost not expressed in the other two subtypes. In the survival analysis, the C1 subtype had the worst survival within 100 months.

[0185] In a preferred embodiment of the present invention, the method for verifying the immune microenvironment typing and biological feature screening by single cell omics analysis in step S4 is:

[0186] Collect single-cell data of samples with medulloblastoma, and use Seurat (a single-cell pipeline analysis method) to perform quality control and dimensionality reduction clustering of single-cell data;

[0187] Combine the collected annotation information to annotate the cell subpopulations of dimensionality reduction clustering;

[0188] Using the R package scRNAtoolVis ("scRNAtoolVis" is an R package for single-cell RNA sequencing (scRNA-seq) data analysis and visualization. An R package is a collection of functions, datasets, and documents in the R language environment to extend its functionality), we obtained the distribution status of specific marker makers in single-cell subpopulations and the differences in single-cell levels among medulloblastoma patients with different immune microenvironment subtypes.

[0189] The method for collecting single-cell data of samples with medulloblastoma and using Seurat to perform quality control and dimensionality reduction clustering of single-cell data is as follows:

[0190] For each sample, the number of UMIs, total number of genes, number of mitochondrial genes, and number of ribosomal genes were counted, and cells with a total number of genes greater than 6000 or a number of genes less than 200, as well as cells with a mitochondrial gene ratio greater than 30%, were filtered out;

[0191] For each sample, the top 2000 genes with the largest variation were determined based on the mean and dispersion of all genes for integrated analysis to eliminate batch effects between samples. The integrated data were subjected to principal component analysis (PCA) dimensionality reduction, and the top 20 principal components were retained to capture the main changes in the data.

[0192] Seurat's pipeline was used to detect hgv, calculate the average expression and dispersion of each gene, put the genes into boxes and calculate the z-score of the dispersion within each box; a z-score of 0.5 was used as the cutoff value of dispersion, a lower limit of 0.0125 and a high cutoff of 3.0 were used as the average expression, principal component analysis PCA was used for linear dimensionality reduction, and the ElbowPlot method and Jackstraw method were used to select statistically significant principal components; Seurat was used to visualize the clusters based on UMAP (Uniform Manifold Approximation and Projection), and the specific steps were as follows:

[0193] JackStraw: Use the JackStraw method to evaluate the PCA results and identify significant principal components;

[0194] ScoreJackStraw: Score the results of the JackStraw method and determine the principal components with statistical significance;

[0195] JackStrawPlot and ElbowPlot: visualize the results of the JackStraw method and select the number of principal components to be retained. JackStrawPlot is used to observe the p-value of each principal component, while ElbowPlot is used to observe the changes in the degree of variance explanation, usually selecting the number of principal components at the "elbow";

[0196] Cell clustering:

[0197] FindNeighbors: Calculate the neighbor relationship between cells based on the results of PCA;

[0198] FindClusters: Based on the K-nearest neighbor graph, a clustering algorithm (such as Louvain algorithm) is used to divide cells into different cell clusters (or clusters);

[0199] Idents and table: View the clustering results. Use Idents to view the cluster to which each cell belongs. Table is used to count the number of cells in each cluster.

[0200] UMAP Dimensionality Reduction and Visualization:

[0201] RunUMAP: Use the UMAP algorithm to reduce the dimensionality of the data. UMAP uses a stochastic gradient descent optimization algorithm, an approximation method based on the minimum spanning tree and Gaussian mixture model, and a distance-based weighting function to adjust the similarity weights between different data points to map high-dimensional data into two-dimensional or three-dimensional space;

[0202] Minimize the loss function between the distance between data points in high-dimensional space and the distance between corresponding points in low-dimensional space, and optimize the data representation in low-dimensional space;

[0203] DimPlot: Based on the dimension reduction of UMAP, it draws the distribution map of cells and intuitively shows the relationship and distribution of different cell clusters;

[0204] The UMAP clustering visualization results showed that the cells from 26 samples of SHH, GP3, and GP4 were clustered into 21 (SHH), 18 (GP3), and 15 (GP4) subpopulations, respectively.

[0205] In a preferred embodiment of the present invention, the method for annotating the cell subpopulations of dimensionality reduction clustering in combination with the collected annotation information is as follows:

[0206] For MB patient samples and a single matched sample from relapse, cells that passed quality control for scRNA sequencing were projected into 2D UMAP plots, and cell subpopulations in this experiment were manually annotated based on marker genes from the Cell Marker database and other brain tumor-related literature.

[0207] In the first part of the present invention, the gene expression data of MB patients (SHH=223, GP3=144, GP4=326) were pseudo-single-celled, and a PPI network was generated using Netclass as an adjacency matrix. The gene expression matrix (16116 genes) after pseudogene removal was introduced to perform a similarity network fusion analysis (SNF, Similarity Network Fusion) analysis:

[0208] (1) ssGSEA analysis was performed on the MB data of different subtypes. Based on the pseudo-single cell results obtained, SNF analysis was performed in combination with gene expression data. Based on the differences in the tumor immune microenvironment, different immune microenvironment subtypes in the three subtypes of medulloblastoma, SHH, GP3, and GP4, were identified.

[0209] (2) Combined with other information, such as histopathological information and clinical information, bioinformatics analysis was performed to study and discuss the biological characteristics of each subtype;

[0210] (3) Perform multiple immune infiltration analyses to verify the stability of SNF typing;

[0211] (4) Screen out marker genes with significant survival differences in each subtype, and observe the functional differences of different subtypes through gene enrichment analysis. MB is clearly divided into different subtypes with different biological and clinical characteristics, and genes with prognostic differences in different subtypes are screened. Each subtype has its own unique biological markers. The C2 subtype with the worst prognosis of SHH subtype MB specifically expresses PAWR; among GP3 MB, the C1 subtype with the best prognosis specifically expresses ZMYND8; the C1 subtype GP4 MB has the worst prognosis and specifically expresses PLAT.

[0212] The second part of the present invention uses single cell data of medulloblastoma patients (SHH subtype = 9, GP3 subtype = 6, GP4 subtype = 12) for analysis:

[0213] (1) Use Seurat to perform quality control and dimensionality reduction clustering of single-cell data;

[0214] (2) Annotate the cell subpopulations obtained by dimensionality reduction clustering based on the collected annotation information;

[0215] (3) Observe the distribution status of specific makers in single cell subsets and the differences in single cell levels among MB patients with different immune microenvironment subtypes.

[0216] At the single-cell level, in SHH MB, C2 has the least myeloid cells compared to C1 and C3, and has many unipolar brush cells (UBC); C1 has a large number of granule cell precursors (GCPs), which are almost non-existent in C2 and C3. In GP3 MB, the only difference between C1 and C2 is still the UBC cells. C2 contains a large number of UBC cells, while C1 has almost none. The situation in GP4 MB is completely opposite. The C3 subtype with the best prognosis expresses a large number of UBC cells and has the least myeloid cells, while C1 with the worst prognosis is the opposite, with high expression of transitional cerebellar progenitors (TCP). Low expression of myeloid cells is always accompanied by high expression of UBC, which indicates that UBC cells and myeloid cells may play a key role in promoting the occurrence and development of tumors, and there may be some relationship between them.

[0217] The three subtypes of medulloblastoma have different immune microenvironment subtypes. These subtypes have unique markers and functions, play their own unique roles in the development of tumors, and bring completely different outcomes to patients. Through immune microenvironment typing and marker screening, the immune characteristics and potential therapeutic targets of medulloblastoma are revealed, providing new ideas and methods for the treatment and prognosis evaluation of the disease.

[0218] For example, by drawing violin plots of differentially expressed genes in different samples and cells, their distribution at the single-cell level can be observed. FBXO16 and HCAR1, which are highly expressed in the C1 subtype, are more distributed in oligodendrocytes and GCP cells, and less distributed in myeloid and endothelial cells; ZNF467 and MTHFD1L, which are highly expressed in the C3 subtype, are more distributed in the myeloid system; PAWR, which is specifically expressed in the C2 subtype, is highly enriched in UBC cells.

[0219] The present invention also provides a system for detecting and typing the immune microenvironment differences of medulloblastoma, comprising a processing unit, which executes the method of the present invention to detect and type the immune microenvironment differences of medulloblastoma.

[0220] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0221] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A method for detecting and typing differences in the immune microenvironment of medulloblastoma, characterized in that: The steps include: S1, gene expression data of samples with medulloblastoma were collected and pseudo single-cell calculations were performed; S2, using Netclass to generate a protein interaction network as an adjacency matrix, introducing gene expression data into the protein interaction network, and then combining pseudo-single-cellized gene expression data to perform unsupervised clustering through similarity network fusion analysis to determine different immune microenvironment subtypes in medulloblastoma; S3, using lasso to perform machine learning to screen biological features of different immune microenvironment subtypes; S4, Validation of immune microenvironment typing and biological signature screening by single-cell omics analysis.

2. The method for detecting and typing immune microenvironment differences of medulloblastoma according to claim 1, characterized in that: Step S1 is specifically as follows: Based on the ssGSEA pseudo single-cell analysis method, the gene expression data are compared with the given gene set, and the enrichment score of the gene set in each sample is calculated to simulate the composition of immune cells in the immune microenvironment of the tumor; Based on the netClass analysis method, the information of protein interaction PPI network is used to integrate gene expression data and protein network topology. The pseudo single-cell results obtained by the ssGSEA method were combined with gene expression data to perform similarity network fusion SNF analysis on the samples; The survival package and ggplot2 package, ESTIMATE package, MCPcounter package, and msigdbr package in R language were used to visualize the survival data of different immune subtypes of MB patients, evaluate the abundance of tumor microenvironment and different cell types in the tumor microenvironment, and evaluate the gene set data related to biological processes; Use the glmnet function in the glmnet package to perform multi-classification lasso regression analysis on gene expression data; Based on the gene expression data, the R software package l imma was used to perform DEG analysis of differentially expressed genes between groups to obtain the differences in data expression between sample groups; the R software package GOplot was used to draw bubble charts to observe the enrichment of pathways.

3. The method for detecting and typing the immune microenvironment differences of medulloblastoma according to claim 2, characterized in that: Based on the netClass analysis method, the information of protein interaction network is used to integrate gene expression data and protein network topology. The specific steps are as follows: Obtain protein interaction network data from public databases or literature, and construct a network containing the interaction relationships between proteins as an adjacency matrix; Calculate the diffusion kernel: accept the adjacency matrix as input and calculate the diffusion kernel matrix according to the specified parameters. The specified parameters p and a are used to generate the diffusion kernel matrix through iterative calculation. When calculating the diffusion kernel matrix, the parameter is.adjacency is used to determine whether the input is an adjacency matrix or a Laplace matrix, and different calculation paths are selected. Use the igraph library to handle graph operations, including creating undirected graphs and calculating Laplacian matrices; When calculating the diffusion kernel matrix and the subsequent eigenvector multiplication, matrix operations are used to screen out the intersection with another data set, and then the diffusion kernel is calculated using part of the data in the intersection.

4. The method for detecting and typing the immune microenvironment differences of medulloblastoma according to claim 2, characterized in that: The method of combining the pseudo single-cell results obtained by the ssGSEA method with the gene expression data to perform similar network fusion SNF analysis on the samples is as follows: The complete matrix of expression after pseudogene removal, the ssGSEA matrix, was used as input; Using the SNFtool R package, the number of neighbors K = 30, the Gaussian kernel parameter alpha = 0.5, and the number of iterations T = 10, spectral clustering implemented by the SNF toolkit was run on the SNF fusion similarity matrix to obtain the most appropriate grouping; SNF analysis was performed in different subtypes: in the SHH subtype, the number of samples was n = 233, and three immune microenvironment subgroups were obtained when k = 3; In the GP3 subtype, n = 144, 2 subsets of immune microenvironments were obtained at k = 2; In the GP4 subtype, n=326, 3 subsets of immune microenvironments were obtained at k=3.

5. The method for detecting and typing the immune microenvironment differences of medulloblastoma according to claim 2, characterized in that: The survival package, ggplot2 package, ESTIMATE package, msigdbr package, and MCPcounter package in R language were used to visualize the survival data of MB patients, evaluate the tumor microenvironment based on gene expression data, obtain gene set data related to biological processes in the genomics database of species, and evaluate the abundance of different cell types in the tumor microenvironment as follows: S11, prepare the data required for survival analysis, including the observation time futime and event status fustat of each sample, and the variable group used for grouping; S12, use the survfit() function to fit the survival curve, and calculate the survival curve according to the specified observation time and event status and grouping variable group; S13, use ggsurvplot() function to create a survival curve chart, set the chart not to display the confidence interval conf.int = F, display the risk table risk.table = T, do not add the total patient survival curve add.all = F, customize the palette palette = "Dark2", adjust the title, axis title, font size and style, and the range of the coordinate axis; S21, prepare the gene expression data file, use the outputGCT() function to convert the data into the input format required by the ESTIMATE package, and save it as a new file in.gct.file; S22, filterCommonGenes() function is used to filter out common genes from the raw data and save them as a new input file, estimateScore() function is used to calculate the ESTIMATE score of the tumor sample, and the score result is saved as a GCT format file out.score.file; S23, then use the plotPurity() function to visualize the purity score of the sample, use the read.table() function to read the score result file and convert it into a data frame format, and finally save the result as a text file ESTIMATE_score.txt; S31, use the msigdbr() function to obtain the gene set data of the C5 category from the species' genomics database and save it in the GO_df_al l data frame; The required columns gs_name, gene_symbol, gs_exact_source, and gs_subcat were selected using the dplyr::select() function, and unnecessary data were filtered out based on the subcategory gs_subcat to obtain the final GO gene set data GO_df; S32, using the split() function to group the gene set data according to the GO item name gs_name to obtain a gene list go_list; Use the gsva() function to perform gene set variability analysis GSVA on the gene expression data rt1 to be analyzed, calculate the gene set variability score gsva_mat for each GO entry, and save the result in the text file gsva_hallmark.txt; S33, set parameters: use Gaussian kernel density estimation function kcdf = "Gaussian", and call all available kernels parallel.sz = parallel::detectCores(); S41, read the data required for MCPcounter evaluation, including gene expression data of cell types MCP_counter_tz.txt and gene information data MCP_counter_gene.txt; S42, use the MCPcounter.estimate() function to estimate the abundance of cell types in the tumor sample, specify the probesets parameter for gene expression data of the cell type, the genes parameter for gene information of the cell type, and the featuresType parameter for feature type; S43, save the evaluation results in the results object and visualize them to understand the abundance of different cells in the immune microenvironment.

6. The method for detecting and typing immune microenvironment differences of medulloblastoma according to claim 2, characterized in that: In step S3, the method of performing multi-classification lasso regression analysis on gene expression data using the glmnet function in the glmnet package is: Use the glmnet function to fit a multinomial logistic regression model, where the independent variable x is the gene expression of the sample, the dependent variable y is the classification label of the sample, the family parameter is used to specify the distribution type of the multinomial as multinomial, and the type.multinomial parameter is used to set the combination mode; Use the plot function to visualize the fitted model, set the x-axis to the coefficient lambda of the regularization penalty term and mark the value of each coefficient; Use the cv.glmnet function to cross-validate the lasso regression model, select the optimal lambda value, and refit the lasso regression model using the glmnet function based on the optimal lambda value. The alpha parameter is set to 1 to indicate the use of lasso regression, the lambda parameter is set to the optimal value obtained by cross-validation, and the family parameter and type.multinomial parameter are the same as before. Extract the genes whose coefficients in the lasso regression model are not 0 and save them as gene_min.

7. The method for detecting and typing immune microenvironment differences of medulloblastoma according to claim 2, characterized in that: Based on the gene expression data, the R software package l imma was used to perform differentially expressed gene DEG analysis between groups to obtain the difference in data expression between sample groups; the R software package GOplot was used to draw a bubble chart to observe the pathway enrichment. The specific steps are as follows: Differentially expressed genes were analyzed using the limma package: Design matrix creation: Use the model.matrix function to create a design matrix design, which is used to describe the treatment group group in the experimental design; Contrast matrix creation: Use the makeContrasts function to create the control matrix contr.matrix and define the contrast between the two control groups C1VSC2, where C1 and C2 are the two control groups in the design matrix; Linear model fitting: Use the lmFit function to fit a linear model vfit, which is used to fit the expression data y and the design matrix design; Contrast fitting: Use the contrasts.fit function to perform contrast fitting on the linear model and pass in the contrast matrix contr.matrix; Bayesian estimation: Use the eBayes function to perform Bayesian estimation on the fitted model to obtain the significance of the gene; Diagnostic plot drawing: The plotSA function is used to draw the diagnostic plot of the model to evaluate the suitability and stability of the model; Significance test: Use the decideTests function to perform a significance test on the model and output summary information of the significance test results of the gene; Differentially expressed gene analysis: Use the topTable function to obtain the differential expression of all genes based on the results of Bayesian estimation, and then filter out genes with significant differential expression based on the given thresholds padj and foldChange; Result output: Write the significantly differentially expressed genes after screening into a CSV file; Functional annotation and enrichment analysis of DEGs. The R software package clusterProfiler was used to perform GO functional annotation and KEGG enrichment analysis on the screened differentially expressed gene DEGs. The enrichment results were considered significant only when the P value was less than 0.

05. Specifically: Gene ID conversion: Use the bitr function to convert gene symbols from SYMBOL to ENTREZID; GO analysis: Use the enrichGO function to perform GO enrichment analysis, specify the type of gene ID, p-value and q-value threshold parameters, and save the results as CCGO.rdata file; KEGG analysis: Use the enrichKEGG function to perform KEGG enrichment analysis, specify parameters such as species, p-value and q-value thresholds, and save the results as CCKEGG.rdata file; GO / KEGG enrichment visualization: Use barplot and dotplot functions to visualize GO and KEGG enrichment results and save the images as pdf files; Gene set comparison analysis: Use the enrichKEGG function to perform KEGG enrichment analysis on genes between different groups, compare the KEGG enrichment results of different groups, and visualize the results.

8. The method for detecting and typing immune microenvironment differences of medulloblastoma according to claim 2, characterized in that: Step S4: The method for verifying the immune microenvironment typing and biological feature screening by single cell omics analysis is: Collect single-cell data of samples with medulloblastoma and use Seurat to perform quality control and dimensionality reduction clustering of single-cell data; Combine the collected annotation information to annotate the cell subpopulations of dimensionality reduction clustering; Use the R package scRNAtoolVis to obtain the distribution status of specific marker maker in single-cell subpopulations and the differences in single-cell levels of medulloblastoma patients with different immune microenvironment subtypes; The method for collecting single-cell data of samples with medulloblastoma and using Seurat to perform quality control and dimensionality reduction clustering of single-cell data is as follows: For each sample, the number of UMIs, total number of genes, number of mitochondrial genes, and number of ribosomal genes were counted, and cells with a total number of genes greater than 6000 or a number of genes less than 200, as well as cells with a mitochondrial gene ratio greater than 30%, were filtered out; For each sample, the top 2000 genes with the largest variation were determined based on the mean and dispersion of all genes for integrated analysis to eliminate batch effects between samples. The integrated data were subjected to principal component analysis (PCA) dimensionality reduction, and the top 20 principal components were retained to capture the main changes in the data. Seurat pipeline was used to detect hgv, and the average expression and dispersion of each gene were calculated. Genes were placed in bins and the z-score of the dispersion within each bin was calculated. A z-score of 0.5 was used as the cutoff value for dispersion, a lower limit of 0.0125 and a high cutoff of 3.0 were used as the average expression. Principal component analysis (PCA) was used for linear dimensionality reduction, and the ElbowPlot method and Jackstraw method were used to select statistically significant principal components. Seurat was used to visualize the clustering based on UMAP. The specific steps are as follows: JackStraw: Use the JackStraw method to evaluate the PCA results and identify significant principal components; ScoreJackStraw: Score the results of the JackStraw method and determine the principal components with statistical significance; JackStrawPlot and ElbowPlot: visualize the results of the JackStraw method and select the number of principal components to be retained. JackStrawPlot is used to observe the p-value of each principal component, while ElbowPlot is used to observe the changes in the degree of variance explanation, usually selecting the number of principal components at the "elbow"; Cell clustering: FindNeighbors: Calculate the neighbor relationship between cells based on the results of PCA; FindClusters: Based on the K-nearest neighbor graph, a clustering algorithm is used to divide cells into different cell clusters; Idents and table: View the clustering results. Use Idents to view the cluster to which each cell belongs. Table is used to count the number of cells in each cluster. UMAP Dimensionality Reduction and Visualization: RunUMAP: Use the UMAP algorithm to reduce the dimensionality of the data. UMAP uses a stochastic gradient descent optimization algorithm, an approximation method based on the minimum spanning tree and Gaussian mixture model, and a distance-based weighting function to adjust the similarity weights between different data points to map high-dimensional data into two-dimensional or three-dimensional space; Minimize the loss function between the distance between data points in high-dimensional space and the distance between corresponding points in low-dimensional space, and optimize the data representation in low-dimensional space; DimPlot: Based on the dimension reduction of UMAP, it draws the distribution map of cells and intuitively shows the relationship and distribution of different cell clusters; The UMAP clustering visualization results showed that the cells from 26 samples of SHH, GP3, and GP4 were clustered into 21SHH, 18GP3, and 15GP4 subpopulations, respectively.

9. The method for detecting and typing immune microenvironment differences of medulloblastoma according to claim 1, characterized in that: Combining the collected annotation information, the method for annotating the cell subpopulations of dimensionality reduction clustering is as follows: For MB patient samples and a single matched sample from relapse, cells that passed quality control for scRNA sequencing were projected into 2D UMAP plots, and cell subpopulations in this experiment were manually annotated based on marker genes from the Cell Marker database and other brain tumor-related literature.

10. A system for detecting and typing differences in the immune microenvironment of medulloblastoma, characterized in that: The method comprises a processing unit, wherein the processing unit executes the method according to any one of claims 1 to 9 to detect and classify the immune microenvironment differences of medulloblastoma.

Citation Information

Patent Citations

  • Tumor immune typing method based on lncRNA markers

    CN113355411A

  • Application of cancer-related fibroblast-related typing system to prediction of glioblastoma prognosis and treatment responsiveness

    CN117805379A

  • Esophageal cancer treatment target screening method based on single cell space transcriptome

    CN118116453A

  • Scoring model for evaluating tumor immune microenvironment and construction method therefor

    WO2023123005A1

Cited By

  • Application of E33 region methylation level in evaluation of myeloblastoma prognosis products

    CN120138153A

  • Use of e33 region methylation levels in assessing medulloblastoma prognosis

    CN120138153B

  • Cell infiltration inference method and system fusing GO function annotation and PPI network information

    CN121075448A

  • Cell infiltration inference method and system fusing go function annotation and ppi network information

    CN121075448B

  • Brain glioma immune microenvironment complexity assessment method and application

    CN121096452A