Gene group for immunohistochemical typing of BRAF V600E mutant colorectal cancer and application thereof
By using CD8 and TEM4 genes for immunohistochemical detection and machine learning algorithm screening, an immunohistochemical classification method for BRAF V600E mutant colorectal cancer was established, solving the high cost and complexity of the existing classification methods and achieving efficient and reliable classification results.
Patent Information
- Application Number
- CN202510003433.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-02
- Publication Date
- 2025-04-11
AI Technical Summary
The existing molecular typing methods for BRAF V600E colorectal cancer are expensive, complex processes and poor clinical accessibility.
A set of genes for immunohistochemistry typing of BRAF V600E mutant colorectal cancer, including CD8 and TEM4 genes, is provided, and an immunohistochemistry-based classification method is established through immunohistochemistry detection and machine learning algorithm screening.
Effective immunohistochemical typing of BRAF V600E mutant colorectal cancer was achieved, overcome the high cost and complexity of existing methods, improve clinical accessibility, and is highly consistent with the typing results based on mRNA expression.
Smart Images
Figure CN119799894A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the fields of cancer typing and genomics, and in particular to a group of genes for immunohistochemical typing of BRAF V600E mutant colorectal cancer and applications thereof. Background Art
[0002] BRAF V600E mutant colorectal cancer (BRAF V600E CRC) refers to a type of colorectal cancer in which mutations in serine / threonine kinase genes downstream of EGFR and RAS in the MAPK signaling pathway lead to conformational changes in RAF kinase, activation of BRAF kinase and downstream MAPK pathways, and increased tumor cell proliferation and anti-apoptotic behavior. BRAF V600E mutant colorectal cancer occurs in about 10% of colorectal cancers. This mutation is more likely to occur in patients with advanced age, female, right-sided colon cancer, mucinous and poorly differentiated, MSI-h type colorectal cancer, and is prone to peritoneal and distant lymph node metastasis. This group has a relatively poor prognosis, and the responses achieved by the current targeted therapy strategies used for this group are still limited.
[0003] High heterogeneity is the main characteristic of malignant tumors and the main reason for the low efficacy of anti-tumor treatment. BRAF mutant colorectal cancer is no exception. Researchers used transcriptome sequencing combined with cluster analysis of gene expression profiles to prove that there are two molecular subtypes of BRAFV600E mutant colorectal cancer, which can be divided into BM1 type with high inflammatory response and BM2 type with cell cycle checkpoint disorder. Existing studies have shown that BM1 and BM2 type BRAF V600E mutant colorectal cancer patients have significant differences in responsiveness to treatment and prognosis. BM1 type colorectal cancer patients are more sensitive to dual-target or triple-target treatment regimens, and the survival time of BM1 type patients is almost three times that of BM2 type patients. At the same time, because BM1 type colorectal cancer is accompanied by a high inflammatory response, it suggests that patients with this type of colorectal cancer are expected to benefit from treatment with immune checkpoint inhibitors.
[0004] Although molecular classification of BRAF V600E colorectal cancer based on mRNA expression profile sequencing cluster analysis can simultaneously reveal the biological characteristics of BRAF V600E CRC and suggest subtype-specific treatment strategies, this method has the disadvantages of high specimen quality requirements for tumor tissue, expensive expression profile sequencing technology, complex technical processes, large batch effects, and poor clinical accessibility. Immunohistochemistry is currently the most commonly used pathological detection technology in clinical practice, and the use of immunohistochemistry can accurately and robustly identify tumor subtypes. However, there is currently no immunohistochemical typing method for BRAF V600E mutant colorectal cancer. Summary of the invention
[0005] The present invention aims to solve the problems of high price, complicated process and poor clinical accessibility of existing molecular typing methods for BRAF V600E colorectal cancer, and provide a group of genes and applications thereof for immunohistochemical typing of BRAF V600E mutant colorectal cancer.
[0006] The invention provides a group of genes for immunohistochemical typing of BRAF V600E mutant colorectal cancer, comprising the following two genes: CD8 gene and TEM4 gene.
[0007] The above-mentioned method for screening genes for immunohistochemical typing of BRAF V600E mutant colorectal cancer comprises the following steps:
[0008] Step 1: Screen out genes significantly associated with the phenotype from the differentially expressed genes of BRAF V600E mutant colorectal cancer subtypes, and build classification models based on the phenotype information. Use three machine learning algorithms, random forest algorithm, LASSO regression, and SVM-RFE, to select features, and finally take the intersection of the screening results of the three algorithms to obtain key genes with significant distinguishing ability in phenotypes;
[0009] Step 2: Based on the mRNA expression profile sequencing data of BRAF V600E mutant colorectal cancer, 15 candidate genes including the CD8 gene were identified through ROC analysis;
[0010] Step 3: Obtain paraffin-embedded tissues from patients with BRAF V600E mutant colorectal cancer for immunohistochemical testing and immunohistochemical scoring of 15 candidate genes;
[0011] Step 4: Determine the cut-off value of immunohistochemistry score;
[0012] Step 5: Based on the immunohistochemistry scoring results of step 3 and the cut-off value determined in step 4, the candidate genes were divided into two groups according to their expression levels: high expression and low expression. Subsequently, the decision tree algorithm of R language was used to screen out the gene combination with the best distinguishing ability for BM typing, the CD8 gene and TEM4 gene combination.
[0013] Furthermore, the specific method of using three machine learning algorithms for screening in step one is: independently calculate the importance values of genes through three machine learning algorithms, rank the distinguishing abilities of genes in typing according to the importance values, and screen out characteristic genes with higher importance.
[0014] Furthermore, the ROC analysis described in step 2 uses the mRNA expression values of key genes to distinguish the accuracy of each subtype, and uses the area under the curve AUC as a quantitative standard for accuracy. The candidate genes are ranked according to the AUC values, and genes that cannot provide immunohistochemical antibodies are excluded, and 15 candidate genes including the CD8 gene are screened out.
[0015] Furthermore, the immunohistochemistry scoring criteria in step 3 are:
[0016] ① Scoring criteria for tumor cell nuclear staining markers:
[0017] The percentage of positive tumor cells to all tumor cells;
[0018] ② Scoring criteria for tumor cell cytoplasmic staining markers:
[0019] 3+: >80% of the infiltrating cancer cells show strong, complete, and uniform cytoplasmic staining.
[0020] 2+: >80% of infiltrating cancer cells show moderate intensity, complete cytoplasmic staining,
[0021] 1+: >80% of infiltrating cancer cells show patchy, weak cytoplasmic staining.
[0022] -: no staining or ≤80% of infiltrating cancer cells showing incomplete, weak cytoplasmic staining;
[0023] ③CD8 immune cell marker scoring criteria:
[0024] The number of positive cells / tumor core (CT) and tumor-infiltrative junction (IM) area.
[0025] Furthermore, the specific method for determining the cut-off value of the immunohistochemistry score in step 4 is:
[0026] The immunohistochemical score was used as the independent variable and the BM typing of the patient was used as the dependent variable. The logistic regression model was used to establish a predictive relationship. The genetic algorithm was used to search for the cutoff point within the value range of the independent variable. The sensitivity and specificity of each candidate point were calculated, and the maximum value of sensitivity + specificity was used as the criterion to determine the immunohistochemical score cut-off value of each antibody.
[0027] Furthermore, in step 5, the decision tree algorithm of R language was used for modeling analysis, and the classification decision tree model was constructed with BM typing as the dependent variable and the candidate gene as the independent variable. By calculating the importance score of the candidate gene, the gene combination with the best distinguishing ability for BM typing was screened out.
[0028] The method for immunohistochemical typing of BRAF V600E mutant colorectal cancer of the present invention is:
[0029] Immunohistochemistry BM1 type: CD8 positive, TEM4 negative;
[0030] Immunohistochemistry BM2 type: ①CD8 negative, any TEM4 expression; or ②CD8 positive, TEM4 positive.
[0031] CD8 positivity was defined as the number of CD8 positive cells / tumor core (CT) and tumor infiltration junction (IM) area ≥ 350 / mm 2 TEM4 positivity was defined as strong, complete, and uniform cytoplasmic staining in >80% of the infiltrating cancer cells, or moderate, complete cytoplasmic staining in >80% of the infiltrating cancer cells. CD8 negativity was defined as CD8-positive cells / tumor core (CT) and tumor-infiltrating junction (IM) area <350 / mm 2 TEM4 negativity was defined as >80% of infiltrating cancer cells showing incomplete, weak cytoplasmic staining, or no staining or ≤80% of infiltrating cancer cells showing incomplete, weak cytoplasmic staining.
[0032] The present invention provides the use of the above-mentioned group of genes for immunohistochemical typing of BRAF V600E mutant colorectal cancer in the preparation of a BRAF V600E mutant colorectal cancer typing detection kit.
[0033] The present invention provides a BRAF V600E mutant colorectal cancer typing detection kit, which comprises the following biomarkers, wherein the biomarkers are a group of genes for immunohistochemical typing of BRAF V600E mutant colorectal cancer.
[0034] Beneficial effects of the present invention:
[0035] Currently, there is no recognized immunohistochemical typing method for BRAF V600E mutant colorectal cancer. Based on the CRC BRAFV600E multi-omics data set, the present invention identified CD8 and TEM4 as the key genes for immunohistochemical typing of this cancer, and established an immunohistochemical typing method. The immunohistochemical typing results were compared with the typing results based on mRNA expression, and the results showed that the two were highly consistent.
[0036] In addition, immunohistochemical typing not only provides important information for patient prognosis, but also reveals the unique pathological and molecular characteristics of each immunohistochemical subtype, providing a theoretical basis for the development of specific treatment for CRC BRAF V600E subtype. In addition, the present invention also overcomes the problems of complex operation, high time cost and significant batch effect of expression spectrum sequencing technology in BRAF expression spectrum cluster analysis typing. The proposed immunohistochemical typing method is not only low in cost (only 800 yuan per patient), but also has a high degree of consistency and good repeatability, and high clinical accessibility. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 It is a flow chart of the gene screening of immunohistochemical typing of BRAF V600E mutant colorectal cancer of the present invention;
[0038] Figure 2 This is the error rate curve of random forest algorithm screening typing gene;
[0039] Figure 3 is the trend of the importance of typing genes in the LASSO regression model as the degree of regularization changes;
[0040] Figure 4 The LASSO regression model determines the optimal regularization parameter λ;
[0041] Figure 5 This is a graph showing the 5-fold cross validation error changing with the number of features;
[0042] Figure 6 This is a graph showing the 5-fold cross validation accuracy changing with the number of features;
[0043] Figure 7 Intersection Venn diagram of three machine learning algorithms for screening differentially typed genes;
[0044] Figure 8 This is a schematic diagram of the consistency test results of immunohistochemical typing and expression profile cluster analysis typing of 52 training sets;
[0045] Fig. 9 This is a flow chart of the immunohistochemical typing method for BRAF V600E mutant colorectal cancer;
[0046] Fig.10 This is a schematic diagram of the results of consistency testing between immunohistochemical typing and expression profile cluster analysis typing in the 17 validation sets. DETAILED DESCRIPTION
[0047] The embodiments of the present invention are described in detail below. The following embodiments are implemented on the premise of the technical solution of the present invention, and detailed implementation schemes and specific operation processes are given, but the protection scope of the present invention is not limited to the following embodiments.
[0048] Embodiment 1:
[0049] The genes used for immunohistochemical typing of BRAF V600E mutant colorectal cancer are CD8 and TEM4. The specific gene screening process includes the following steps (see the detailed process for details). Figure 1 ):
[0050] 1. Screening of differentially expressed genes: According to the 476 BRAF V600E mutant colorectal cancer subtypes screened by David Barras and other researchers, differentially expressed genes were identified (Barras D, Missiaglia E, Wirapati P, Sieber OM, Jorissen RN, Love C, Molloy PL, Jones IT, McLaughlin S, Gibbs P, Guinney J, Simon IM, et al. BRAF V600E Mutant Colorectal Cancer Subtypes Based on Gene Expression. Clinical cancer research: an official journal of the American Association for Cancer Research 2017; 23: 104-15.), the specific list of differentially expressed genes is shown in Table 1. Using the BM typing results based on NMF clustering analysis, three machine learning algorithms, random forest, LASSO regression and SVM-RFE, were used to screen the differentially expressed genes related to typing.
[0051] First, genes significantly associated with the phenotypic phenotype were screened from the 476 pre-screened differential gene sets, and classification models were constructed based on the phenotyping information (BM1 and BM2). Feature selection was performed using three algorithms: random forest, LASSO regression, and SVM-RFE. The importance values of genes were calculated independently, and the genes were ranked according to their importance values in phenotyping to screen out characteristic genes with higher importance. Finally, the intersection of the screening results of the three algorithms was taken, and a total of 27 key genes with significant distinguishing ability in phenotyping were obtained.
[0052] The error rate curve of random forest algorithm screening typing gene is as follows Figure 2 As shown, the results showed that the random forest algorithm had a good classification effect on the research data, and the error rate of the best model was low and stable, indicating that the selected characteristic genes had a good ability to distinguish BM1 and BM2 typing.
[0053] The results of the LASSO regression model are shown in the figure Figure 3 and Figure 4As shown, Figure 3 It reflects the trend of the importance of typing genes changing with the degree of regularization. Figure 4 The optimal regularization parameter λ was determined, providing a basis for subsequent feature gene selection.
[0054] Feature Selection Cross Validation Evaluation Figure 5 and Figure 6 shown. Figure 5 This is a graph of the 5-fold cross-validation error versus the number of features. The curve shows that as the number of features increases, the cross-validation error gradually decreases, indicating that adding features helps reduce prediction errors. The figure marks the lowest error (0.0318) corresponding to the optimal number of features (98), which is the feature combination with the best model performance. Figure 6 This is a graph of the 5-fold cross-validation accuracy versus the number of features. The curve shows that as the number of features increases, the accuracy gradually increases and tends to stabilize; the figure marks the highest accuracy (96.8%) corresponding to the optimal number of features (98). Conclusion: Through cross-validation analysis, it can be found that the model achieves the best performance when using 98 features, that is, the lowest error and the highest accuracy.
[0055] The intersection of three machine learning algorithms for screening differentially expressed genes is shown in the Venn diagram. Figure 7 shown.
[0056] Table 1 Gene sets screened by three machine learning algorithms
[0057]
[0058]
[0059]
[0060]
[0061]
[0062]
[0063]
[0064]
[0065]
[0066]
[0067]
[0068]
[0069] 2. Determination of candidate genes: Based on the mRNA expression profile sequencing data of 52 cases of BRAF V600E mutant colorectal cancer from the Affiliated Tumor Hospital of Harbin Medical University, the accuracy of the mRNA expression values of 27 key genes in distinguishing each subtype was evaluated by receiver operating characteristic (ROC) analysis, and the area under the curve (AUC) was used as the quantitative standard of accuracy. The candidate genes were ranked according to the AUC value, and combined with the antibody supplier database, genes that could not provide immunohistochemical antibodies were excluded, and 14 candidate genes were screened out. Since there are significant differences in the immune microenvironment between BM1 and BM2 types, and BM1 type is a highly inflammatory response type, the CD8 gene was added to the candidate gene list. The CD8 gene was selected as a candidate gene because the CD8 molecule it encodes plays a key role in tumor immune response and its expression can be easily detected by RNA sequencing and immunohistochemistry. Studies have shown that CD8+T cells are the main effector cells of anti-tumor immunity, and their density has a significant impact on the prognosis of patients with various tumor types and is an important predictor of the efficacy of immunotherapy (such as PD-1 / PD-L1 inhibitors). Therefore, 15 candidate genes were finally identified.
[0070] 3. Immunohistochemical scoring: Paraffin-embedded tissues were obtained from the above 52 patients with BRAF V600E mutant colorectal cancer for immunohistochemical testing and the protein expression of 15 candidate genes was evaluated. The following different scoring criteria were selected according to the different expression sites of the 15 candidate genes.
[0071] ① Scoring criteria for tumor cell nuclear staining markers:
[0072] The percentage of positive tumor cells to all tumor cells;
[0073] ② Scoring criteria for tumor cell cytoplasmic staining markers:
[0074] 3+: >80% of the infiltrating cancer cells show strong, complete, and uniform cytoplasmic staining.
[0075] 2+: >80% of infiltrating cancer cells show moderate intensity, complete cytoplasmic staining,
[0076] 1+: >80% of infiltrating cancer cells show patchy, weak cytoplasmic staining.
[0077] -: no staining or ≤80% of infiltrating cancer cells showing incomplete, weak cytoplasmic staining;
[0078] ③CD8 immune cell marker scoring criteria:
[0079] The number of positive cells / tumor core (CT) and tumor-infiltrative junction (IM) area.
[0080] These scoring criteria can be used to evaluate the protein expression pattern of each candidate gene in tumor cells and immune cells, as well as the differences in expression patterns of the same candidate gene in different patients.
[0081] 4. Define the cut-off value of immunohistochemistry score: Based on the immunohistochemistry score results of 52 patients with BRAF V600E mutant colorectal cancer, the optimization criteria are to maximize the sensitivity and specificity of identifying the corresponding subtypes. A genetic algorithm based on a logistic regression model was used to screen the optimal cut-off point (cut-off) of the immunohistochemistry score. The specific method is: using the immunohistochemistry score as the independent variable and the patient's BM typing as the dependent variable, a predictive relationship was established using a logistic regression model. The genetic algorithm was used to search for the cut-off point within the range of the independent variable, calculate the sensitivity and specificity of each candidate point, and use the maximum value of sensitivity + specificity as the criterion to determine the immunohistochemistry score cut-off value of each antibody.
[0082] Specifically, the goal of this step is to find a cut-off point that maximizes the combined effect of sensitivity and specificity, thereby achieving the strongest ability to distinguish between different subtypes.
[0083] 5. Confirmation of typing gene combination: Based on the immunohistochemical scoring results of 52 patients with BRAF V600E mutant colorectal cancer in step 3, the candidate genes were divided into high expression and low expression groups according to the expression level according to the cut-off value determined in step 4, and the patients were assigned to the corresponding BM subtypes. Subsequently, the decision tree algorithm (rpart package) of R language was used to model and analyze the above data, and a classification decision tree model was constructed with BM typing as the dependent variable and candidate genes as the independent variable. By calculating the importance scores of candidate genes, the gene combination with the best distinguishing ability for BM typing was screened out. At the same time, the model complexity parameter (cp) was evaluated using cross-validation error, and the optimal cp value was determined to optimize the decision tree model and balance the model accuracy and complexity. Finally, the best immunohistochemical typing gene combination TEM4 and CD8 was screened out, and a BM subtype typing method based on immunohistochemical scoring was constructed. The typing results based on immunohistochemistry were tested for consistency with those based on mRNA expression (i.e., BM typing results based on NMF cluster analysis), and the typing consistency was evaluated by calculating the Cohen's kappa index. Figure 8 As shown, Cohen's kappa index was 0.55 and typing consistency reached 82.6%, verifying the stability and reliability of the screened gene combination.
[0084] 6. The final classification method is as follows Fig. 9 As shown, that is:
[0085] Immunohistochemistry BM1 type: CD8 positive, TEM4 negative;
[0086] Immunohistochemistry BM2 type: CD8 negative, any TEM4 expression; or CD8 positive, TEM4 positive;
[0087] CD8 positivity was defined as the number of CD8 positive cells / tumor core (CT) and tumor infiltration junction (IM) area ≥ 350 / mm 2 TEM4 positivity was defined as strong, complete, and uniform cytoplasmic staining in >80% of the infiltrating cancer cells, or moderate, complete cytoplasmic staining in >80% of the infiltrating cancer cells. CD8 negativity was defined as CD8-positive cells / tumor core (CT) and tumor-infiltrating junction (IM) area <350 / mm 2 TEM4 negativity was defined as >80% of infiltrating cancer cells showing incomplete, weak cytoplasmic staining, or no staining or ≤80% of infiltrating cancer cells showing incomplete, weak cytoplasmic staining.
[0088] Example 2: Immunohistochemical typing verification
[0089] This immunohistochemical typing method has been implemented and validated in a cohort of 17 patients with BRAF V600E mutant colorectal cancer from the Affiliated Cancer Hospital of Harbin Medical University, verifying the consistency between the immunohistochemical typing results and the gold standard - expression profile clustering analysis typing results - as well as the clinical significance of immunohistochemical typing.
[0090] 1. We obtained 17 BRAF V600E mutant colorectal cancer postoperative tumor tissue samples and prepared tissue sections. These 17 BRAF V600E mutant colorectal cancer patients all had mRNA expression profile data obtained by second-generation sequencing, as well as expression profile cluster analysis typing results obtained by gene pair analysis based on mRNA data.
[0091] 2. Immunohistochemical staining of CD8 and TEM4 was performed on the sections of the 17 patients with BRAF V600E mutant colorectal cancer. The antibodies used were: CD8 antibody (Abcam, 1:400 dilution) and TEM4 antibody (Abcam, ab67278, 1:500 dilution).
[0092] 3. Review the immunohistochemical staining results, evaluate the immunohistochemical staining scores of each indicator of each sample and take the average value.
[0093] 4. According to the typing method, immunohistochemical typing was performed on 17 cases of BRAF V600E mutant colorectal cancer in this example. The specific typing method and results are as follows:
[0094] Immunohistochemistry BM1 type (8 cases in total): CD8 positive, TEM4 negative.
[0095] Immunohistochemistry BM2 type (9 cases in total): CD8 negative, any expression of TEM4; or CD8 positive, TEM4 positive.
[0096] 5. The obtained immunohistochemical typing was checked for consistency with the mRNA-based typing. Fig.10 As shown in the figure, the consistency test Cohen's kappa index is 0.414, and the consistency of typing reaches 70.5%, which indicates that the immunohistochemical typing method of the present invention is relatively reliable.
Claims
1. A set of genes for immunohistochemical typing of BRAF V600E mutant colorectal cancer, characterized in that: The genes for immunohistochemical typing include the following two genes: CD8 gene and TEM4 gene.
2. The gene for immunohistochemical typing of BRAF V600E mutant colorectal cancer according to claim 1, characterized in that: The method for screening genes for immunohistochemical typing comprises the following steps: Step 1: Screen out genes significantly associated with the phenotype from the differentially expressed genes of BRAF V600E mutant colorectal cancer subtypes, and build classification models based on the phenotype information. Use three machine learning algorithms, random forest algorithm, LASSO regression and SVM-RFE, to select features. Finally, take the intersection of the screening results of the three algorithms to obtain key genes with significant distinguishing ability in the phenotype. Step 2: Based on the mRNA expression profile sequencing data of BRAF V600E mutant colorectal cancer, 15 candidate genes including the CD8 gene were identified through ROC analysis; Step 3: Obtain paraffin-embedded tissues from patients with BRAF V600E mutant colorectal cancer for immunohistochemical testing and immunohistochemical scoring of 15 candidate genes; Step 4: Determine the cut-off value of immunohistochemistry score; Step 5: Based on the immunohistochemistry scoring results of step 3 and the cut-off value determined in step 4, the candidate genes were divided into two groups according to their expression levels: high expression and low expression. Subsequently, the decision tree algorithm of R language was used to screen out the gene combination with the best distinguishing ability for BM typing, the CD8 gene and TEM4 gene combination.
3. The method for screening genes for immunohistochemical typing according to claim 2, characterized in that: The specific method of using three machine learning algorithms for screening in step 1 is: independently calculate the importance value of the gene through three machine learning algorithms, sort the gene's distinguishing ability in typing according to the importance value, and screen out characteristic genes with higher importance.
4. The method for screening genes for immunohistochemical typing according to claim 2, characterized in that: The ROC analysis described in step 2 uses the mRNA expression values of key genes to distinguish the accuracy of each subtype, and uses the area under the curve AUC as a quantitative standard for accuracy. The candidate genes are ranked according to the AUC values, and genes that cannot provide immunohistochemical antibodies are excluded, and 15 candidate genes including the CD8 gene are screened out.
5. The method for screening genes for immunohistochemical typing according to claim 2, characterized in that: The immunohistochemical scoring criteria in step 3 are: ① Scoring criteria for tumor cell nuclear staining markers: The percentage of positive tumor cells to all tumor cells; ② Scoring criteria for tumor cell cytoplasmic staining markers: 3+: >80% of the infiltrating cancer cells show strong, complete, and uniform cytoplasmic staining. 2+: >80% of infiltrating cancer cells show moderate intensity, complete cytoplasmic staining, 1+: >80% of infiltrating cancer cells show patchy, weak cytoplasmic staining. -: no staining or ≤80% of infiltrating cancer cells showing incomplete, weak cytoplasmic staining; ③CD8 immune cell marker scoring criteria: The number of positive cells / tumor core and tumor invasion boundary area.
6. The method for screening genes for immunohistochemical typing according to claim 2, characterized in that: The specific method for determining the cut-off value of immunohistochemistry score in step 4 is: The immunohistochemical score was used as the independent variable and the BM typing of the patient was used as the dependent variable. The logistic regression model was used to establish a predictive relationship. The genetic algorithm was used to search for the cutoff point within the value range of the independent variable. The sensitivity and specificity of each candidate point were calculated, and the maximum value of sensitivity + specificity was used as the criterion to determine the immunohistochemical score cut-off value of each antibody.
7. The method for screening genes for immunohistochemical typing according to claim 2, characterized in that: In step 5, the decision tree algorithm of R language was used for modeling analysis, with BM typing as the dependent variable and candidate genes as the independent variable to construct a classification decision tree model. By calculating the importance scores of candidate genes, the gene combination with the best ability to distinguish BM typing was screened out.
8. A method for immunohistochemical typing of BRAF V600E mutant colorectal cancer, characterized in that: The method is: Immunohistochemistry BM1 type: CD8 positive, TEM4 negative; Immunohistochemistry BM2 type: ①CD8 negative, any TEM4 expression; or ②CD8 positive, TEM4 positive; CD8 positivity was defined as the number of CD8 positive cells / tumor core and tumor infiltration junction area ≥ 350 / mm 2 TEM4 positivity was defined as strong, complete, and uniform cytoplasmic staining in >80% of the infiltrating cancer cells, or moderate, complete cytoplasmic staining in >80% of the infiltrating cancer cells; CD8 negativity was defined as the number of CD8-positive cells / tumor core and tumor invasion interface area <350 / mm 2 TEM4 negativity was defined as >80% of infiltrating cancer cells showing incomplete, weak cytoplasmic staining, or no staining or ≤80% of infiltrating cancer cells showing incomplete, weak cytoplasmic staining.
9. Use of a set of genes for immunohistochemical typing of BRAF V600E mutant colorectal cancer as claimed in claim 1 in preparing a BRAF V600E mutant colorectal cancer typing detection kit.
10. A BRAF V600E mutant colorectal cancer typing detection kit, characterized in that: The kit comprises the following biomarkers, which are a group of genes for immunohistochemical typing of BRAF V600E mutant colorectal cancer as described in claim 1.