Construction method and application of colorectal cancer prognosis and treatment resistance prediction model based on fibroblast-related genes

Abstract

The invention discloses a construction method and application of a colorectal cancer prognosis and treatment resistance prediction model based on fibroblast related genes. By detecting the expression level of three specific fibroblast related genes of a colon cancer patient, the kit can be used for judging the total lifetime of the colon cancer patient, evaluating the immune escape potential of tumor cells and the responsiveness of the patient to immunotherapy, and evaluating the sensitivity of tumors to targeted therapy and chemotherapy drugs, so that the kit can be used for detecting the total lifetime of the colon cancer patient. The formulation of a treatment strategy of the colon cancer patient is guided. The prediction model constructed by the fibroblast-related gene can improve the accuracy of predicting the five-year total lifetime of the colon cancer, and compared with the prediction of the prognosis of the colon cancer patient directly through the existing next-generation sequencing technology, the number of the genes needing to be detected is remarkably reduced, the efficiency is improved, the cost is reduced, and the clinical application prospect is broad. And accurate prediction of prognosis and accurate implementation of a treatment scheme are realized.

Description

Technical Field

[0001] The invention relates to a method for constructing and applying a colon cancer prognosis and treatment resistance prediction model based on fibroblast-related genes, and belongs to the technical field of biomedicine. Background Art

[0002] In recent years, the treatment of colon cancer has made some progress. In terms of prognosis prediction and treatment response prediction, there has been no significant progress in biomarkers related to colon cancer. The existing clinical prognostic classifications include the TNM staging system, the Duke staging system, the consensus molecular classification (CMS) molecular classification system for colorectal cancer, etc.; there are also a series of biomarkers, such as CEA, CA199, etc. Although more and more evidence shows that new biomarkers are closely related to the clinical prognosis of colon cancer, especially biomarkers developed by next-generation sequencing technology (NGS), these biomarkers are still some time away from clinical application. In addition, in terms of treatment response prediction, most colon cancers are insensitive to immunotherapy. Studies have found that about 5% of patients with advanced colon cancer carrying high microsatellite instability (MSI-H) can benefit from immunotherapy in the long term. Microsatellite stable (MSS) patients were previously considered to be insensitive to immunotherapy, but recent studies have found that patients with MSS colon cancer carrying POLE / POLD1 mutations are highly sensitive to immunotherapy. This suggests that existing evaluation indicators related to colon cancer immunotherapy are flawed, lack accuracy in screening of advantaged populations, and cannot evaluate the effectiveness of treatment.

[0003] The expression levels of tumor-associated fibroblast (CAF)-related genes reflect the molecular heterogeneity of colon cancer, which determines the different clinical outcomes of patients with colon cancer. More and more basic studies have shown that immune escape is inseparable from tumor progression. Considering that fibroblasts, as the main component of the matrix, have the biological characteristics of forming an immune barrier locally in the tumor microenvironment and physically blocking the infiltration of immune cells. Fibroblast-related genes are a series of genes involved in the immune escape of colon cancer cells, including a series of processes such as the synthesis, decomposition and conversion of TGFβ. However, there is still a lack of comprehensive methods to quantify the activity of fibroblasts in the tumor microenvironment. Summary of the invention

[0004] In order to solve the problem of how to determine the overall survival of colon cancer patients, evaluate the potential of tumor cells to escape immune system and the responsiveness of patients to immunotherapy, and evaluate the sensitivity of tumors to targeted therapy and chemotherapy drugs through tumor tissue detection, the first purpose of the present invention is to provide a biomarker for predicting the prognosis and treatment resistance of colon cancer.

[0005] The technical solution to achieve the above purpose is as follows:

[0006] Biomarkers for predicting prognosis and treatment resistance of colon cancer, wherein the biomarkers are a combination of differentially expressed genes, including cysteine-rich protein 2 gene (CRIP2), follistatin-like protein 3 gene (FSTL3) and glucose transporter 3 gene (SLC2A3).

[0007] The second object of the present invention is to provide the use of the above-mentioned biomarker detection product in the preparation of colon cancer prognosis and treatment resistance prediction products.

[0008] The third object of the present invention is to provide a method for constructing a colon cancer prognosis and treatment resistance prediction model based on fibroblast-related genes, comprising:

[0009] S1 obtains transcriptome sequencing data and clinical information data of tumor tissue RNA from patients with colon cancer, constructs a gene expression matrix, and randomly divides it into a training set and a validation set;

[0010] S2 performs fibroblast infiltration analysis on the gene expression matrix and calculates the fibroblast infiltration ratio of each sample;

[0011] S3 performed WGCNA analysis (Weighted gene co-expression network analysis) on the gene expression matrix to screen out gene modules closely related to fibroblast infiltration;

[0012] S4 used univariate Cox regression analysis to analyze the genes in the gene module closely related to fibroblast infiltration, screened out the fibroblast-related prognostic core genes that were statistically correlated with overall survival, and further used LASSO regression analysis to determine the genes with the largest weights and their corresponding coefficients to construct a fibroblast gene-based prediction model CAFscore. The CAFscore is the sum of the mRNA expression levels of the genes with the largest weights weighted by the corresponding coefficients, specifically CAFscore = (mRNA expression level of CRIP2) * 0.0189 + (mRNA expression level of FSTL3) * 0.1223 + (mRNA expression level of SLC2A3) * 0.1043.

[0013] Furthermore, the method for constructing the prediction model further includes the step of evaluating the prognosis and treatment resistance prediction of the prediction model in the training set, as follows:

[0014] S5 Calculate the CAFscore of each sample in the training set, take the CAFscore at the minimum P value of the training set CAFscore as the optimal threshold, divide the samples into high-risk group and low-risk group, compare the survival difference between the two groups by Kaplan-Meier method, and evaluate the effectiveness of the prediction model by drawing the receiver operating characteristic curve ROC;

[0015] S6 A nomogram combining CAFscore, gender, age, and TNM stage was constructed. The Kaplan-Meier analysis, univariate and multivariate Cox regression, and time-dependent receiver operating characteristic curve (tROC) were used to evaluate the predictive effect of the nomogram on prognosis. The Harrell consistency index (C index) and decision curve analysis (DCA) were used to evaluate the potential of CAFscore in predicting prognosis.

[0016] S7 analyzed the immune microenvironment-related scores of the training set, compared the immune microenvironment-related scores of the CAFscore high-risk group and the low-risk group, and evaluated the accuracy of CAFscore in predicting tumor immune escape potential and immunotherapy efficacy;

[0017] S8 analyzed the immunotherapy cohort, calculated the CAFscore of each tumor sample in the immunotherapy cohort, used the CAFscore at the minimum P value in the immunotherapy cohort as the optimal threshold, divided the samples into a high-risk group and a low-risk group, used the Kaplan-Meier method to compare the overall survival of the two groups of colon cancer patients, and evaluated the accuracy of CAFscore in predicting the prognosis of immunotherapy;

[0018] S9 screens effective small molecule targeted therapeutic compounds and chemotherapeutic drugs, calculates the half-maximal inhibitory concentration (IC50) by ridge regression to evaluate the sensitivity of colon cancer patients in the training set to targeted therapy and chemotherapeutic drugs, and uses 10-fold cross-validation to evaluate the accuracy of CAFscore in predicting sensitivity to targeted therapy and chemotherapeutic drugs.

[0019] Furthermore, the method for constructing the above prediction model also includes verifying the sensitivity of the prediction model to the prognosis and treatment resistance prediction of colon cancer in a validation set, which is the same as S5 to S9, except that the training set is replaced by the validation set.

[0020] Furthermore, in step S2, EPIC and MCPcounter algorithms are used to perform fibroblast infiltration analysis.

[0021] Furthermore, step S3 specifically includes:

[0022] S3-1 performs WGCNA analysis on the gene expression matrix to obtain the Pearson correlation coefficient between any two genes, selects the best soft threshold according to the distribution of gene connectivity, and divides the genes into modules;

[0023] S3-2 calculated the Pearson correlation coefficient between the module and the fibroblast infiltration ratio, and screened out the core gene module with the highest correlation with fibroblast infiltration;

[0024] S3-3 calculated the correlation between genes and module feature values ​​(MM) and the correlation between genes and phenotypic characteristics (GS) in the core gene module, and screened module genes with MM>0.6 and GS>0.5 as hub genes closely related to fibroblast infiltration.

[0025] The fourth object of the present invention is to provide a colon cancer prognosis and treatment resistance prediction model based on fibroblast-related genes constructed by the above-mentioned construction method.

[0026] The fifth purpose of the present invention is to provide the application of the above-mentioned colon cancer prognosis and treatment resistance prediction model based on fibroblast-related genes in establishing a Nomogram model for predicting the probability of overall survival of colon cancer. The specific application method is: using CAFscore, gender, age, and TNM stage as input factors of the Nomogram model, and outputting the probability value of the overall survival of colon cancer patients after surgery; the method for obtaining the probability value is to correspond the sum of the scores corresponding to the four input factors to the value on the probability axis.

[0027] The sixth object of the present invention is to provide a colon cancer prognosis and treatment resistance prediction system based on fibroblast-related genes, comprising:

[0028] A data acquisition module, used to obtain transcriptome sequencing data of RNA from tumor tissues of patients with colon cancer;

[0029] A model scoring module is used to provide the mRNA expression levels of CRIP2, FSTL3 and SLC2A3 obtained by the data acquisition module as input data to the above prediction model to obtain CAFscore;

[0030] The result output module is used to obtain the CAFscore in the model scoring module and output the prediction result, compare the CAFscore with the threshold, and if the risk value is greater than the threshold, the prediction result is output as high risk, and if the risk value is less than or equal to the threshold, the prediction result is output as low risk.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] (1) By detecting the expression levels of three fibroblast-related genes in colon cancer tumor tissues and calculating the individualized CAFscore score, the overall survival of colon cancer patients can be objectively evaluated. CAFscore is an independent risk factor for overall survival. The nomogram value obtained by combining it with current clinical indicators is superior to existing clinical indicators and other biomarkers, such as age, gender, and stage.

[0033] (2) At the same time, CAFscore is clinically helpful in evaluating the potential of tumor cells to escape immune system and the ability of colon cancer patients to respond to immunotherapy. CAFscore can be used to timely screen out patients with poor therapeutic effects, improve treatment measures as early as possible, and improve the prognosis of colon cancer patients.

[0034] (3) In addition, CAFscore can be used clinically to evaluate the sensitivity of colon cancer to targeted therapy and chemotherapy drugs, improve the accuracy of screening for treatment-sensitive populations, reduce the incidence of adverse events, and thus improve treatment efficacy, which helps promote the implementation of precision treatment for colon cancer. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is the model construction diagram, where Figure 1 A is the gene module (cyan) closely related to fibroblast infiltration screened out in the training set; Figure 1 B is the gene module (cyan) closely related to fibroblast infiltration screened out in the validation set; Figure 1 C is the core module genes closely related to fibroblast infiltration screened out according to MM>0.6 and GS>0.5 in the training set; Figure 1 D is the core module genes closely related to fibroblast infiltration screened out according to MM>0.6 and GS>0.5 in the validation set; Figure 1 E is the gene set defined by taking the intersection of the above four core module genes closely related to fibroblast infiltration (n=244); Figure 1 F is 11 graphs significantly correlated with overall survival; Figure 1 G and Figure 1 H is the Lasso model construction graph.

[0036] Figure 2 is the survival time prediction graph, where Figure 2 A-2E is the training set, Figure 2 F-2J is the validation set; Figure 2 A and 2F are survival curves; Figure 2 B and 2G are multifactorial COX analysis diagrams; Figure 2 C and 2H are the component diagrams of the nomogram; Figure 2 D and 2I are DCA curve graphs; Figure 2 E and 2J are the ROC curves for 1, 3, and 5 years.

[0037] Figure 3 is the immune escape potential prediction map, where Figure 3 A is a bar graph of the expression levels of immunosuppression-related genes; Figure 3 B is the ssGSEA score bar chart of the immune co-suppression related gene set; Figure 3C is the box plot of T cell dysfunction and exclusion scores calculated by the TIDE algorithm; Figure 3 D is a bar graph of TIDE scores; Figure 3 E is a bar graph of the infiltration fraction of regulatory T cells.

[0038] Figure 4 is an immunotherapy efficacy evaluation chart, where Figure 4 A is a bar graph of the predicted response rate to immunotherapy; Figure 4 B is the ROC curve of CAFscore predicting treatment response; Figure 4 C is the survival curve of CAFscore for the prognostic analysis of the immunotherapy cohort; Figure 4 DF is the survival curve of CAFscore combined with existing immunotherapy-related markers.

[0039] Figure 5 is the treatment sensitivity assessment graph, where Figure 5 A-5B is a bar graph of the prediction of sensitivity to targeted therapy drugs; Figure 5 C is a bar graph of the prediction of chemotherapy drug sensitivity in the CTRP database; Figure 5 D is a bar graph of the prediction of chemotherapy drug sensitivity in the Primer database; Figure 5 E is a mountain plot of the sensitivity prediction of commonly used chemotherapy drugs for colon cancer in the Primer database. DETAILED DESCRIPTION

[0040] The technical solution of the present invention is clearly and completely described below in conjunction with specific embodiments and drawings.

[0041] Example 1

[0042] 1. Obtain transcriptome sequencing data and clinical information data of tumor tissue RNA from patients with colon cancer, construct a gene expression matrix, and randomly divide it into training set and validation set.

[0043] Study subjects: 446 patients with colorectal cancer from The Cancer Genome Atlas (TCGA) were selected as the training set and 562 patients with colorectal cancer from the GSE39582 cohort in the Gene Expression Omnibus (GEO) were selected as the validation set. The following inclusion criteria were used:

[0044] (1) Have mRNA sequencing data that can be used for CAFscore calculation;

[0045] (2) A clear pathological diagnosis of colon cancer;

[0046] (3) no history of radiotherapy, chemotherapy, targeted therapy, immunotherapy, or other anticancer drugs (including neoadjuvant therapy);

[0047] (4) It has clinical information such as survival time, survival status, age, gender, and stage.

[0048] 2. The fibroblast infiltration analysis was performed on the gene expression matrices of the training set and validation set using the “EPIC” and “MCPcounter” algorithms, respectively, and the fibroblast infiltration ratio of each sample was calculated.

[0049] 3. Perform WGCNA analysis on the gene expression matrix to screen out gene modules closely related to fibroblast infiltration, including:

[0050] (1) Perform WGCNA analysis on the gene expression matrix of the training set and validation set to obtain the Pearson correlation coefficient between any two genes, select the optimal soft threshold according to the distribution of gene connectivity, and divide the genes into modules; the optimal soft threshold in the training set TCGA is 3, and the optimal soft threshold in the validation set GSE39582 is 4;

[0051] (2) The Pearson correlation coefficient between the gene module and the fibroblast infiltration ratio was calculated by the “corPvalueStudent” function to screen out the core gene module with the highest correlation with fibroblast infiltration;

[0052] (3) The correlation between genes and module feature values ​​(MM) and the correlation between genes and phenotypic characteristics (GS) in the core gene module were calculated, and module genes with MM>0.6 and GS>0.5 were selected as hub genes closely related to fibroblast infiltration.

[0053] 4. The hub genes closely related to fibroblast infiltration were analyzed by univariate Cox regression analysis using the "survival" R package to screen out the core fibroblast-related prognostic genes that were statistically related to overall survival. The specific screening criteria were: |hazard ratio (HR)|>1 and p<0.05. The "glmnet" R package was further used to analyze the genes with the largest weights and their corresponding coefficients using the LASSO regression method to construct a prediction model CAFscore based on fibroblast genes. The CAFscore is the sum of the mRNA expression levels of the genes with the largest weights weighted by the corresponding coefficients, specifically CAFscore=(mRNA expression level of CRIP2)*0.0189+(mRNA expression level of FSTL3)*0.1223+(mRNA expression level of SLC2A3)*0.1043.

[0054] 5. The prognostic and therapeutic resistance prediction models based on fibroblast-related genes for colon cancer were evaluated as follows:

[0055] (1) The CAFscore of each sample in the training set was calculated, and the CAFscore at the minimum P value of the training set CAFscore was used as the optimal threshold using the "Survminer" R package to divide the samples into a high-risk group and a low-risk group. The survival difference between the two groups was compared using the Kaplan-Meier method, and the effectiveness of the prognostic model was evaluated by drawing the receiver operating characteristic (ROC) curve.

[0056] (2) The prognostic model was combined with CAFscore, gender, age, and TNM stage by using the nomogram constructed by the "rms" R package to calculate the nomogram score. The Kaplan-Meier analysis, univariate and multivariate Cox regression, and time-dependent receiver operating characteristic curve (tROC) were used to evaluate the predictive effect of the nomogram score on prognosis. The Harrell consistency index (C index) and decision curve analysis (DCA) were calculated by the "survcomp" and "ggDCA" R packages to evaluate the potential of CAFscore for prognostic prediction;

[0057] (3) Calculate the immune scores of the training set samples, compare the immune scores of the CAFscore high-risk group and the low-risk group, and evaluate the accuracy of CAFscore in predicting the potential of tumor immune escape and the efficacy of tumors on immunotherapy;

[0058] (4) An external cohort (IMvigor210) of patients with urothelial carcinoma treated with anti-PD-L1 was used to verify the association between CAFscore and immunotherapy outcomes. The CAFscore of each tumor sample in the immunotherapy cohort was calculated, and the CAFscore at the minimum P value in the immunotherapy cohort was used as the optimal threshold. The samples were divided into high-risk and low-risk groups. The Kaplan-Meier method was used to compare the two groups in predicting the overall survival of patients with colon cancer, and the accuracy of CAFscore in predicting the prognosis of immunotherapy was evaluated.

[0059] 6. Analysis of CAFscore on immune microenvironment score is as follows:

[0060] (1) Immune, stromal, and ESTIMATE scores: The "ESTIMATE" algorithm is used to estimate the relative proportions of stromal cells and immune cells in the tumor microenvironment and is presented in the form of StromalScore, ImmuneScore, and ESTIMATEScore;

[0061] (2) Expression levels of immune checkpoint-related genes: Comparison of mRNA expression levels of immunosuppression-related genes such as CD274, CTLA4, and LAG3 between the high- and low-risk CAFscore groups;

[0062] (3) Gene set enrichment analysis (ssGSEA): Normalize the gene expression profile of a single sample using “GSVA”

[0063] The “gsva” function in the R package performs ssGSEA analysis on the gene set and expression matrix, and calculates the corresponding ssGSEA scores of the gene set composed of immune cell-related markers, immune co-suppression gene set, etc.;

[0064] (4) Tumor immune dysfunction and rejection score: TIDE is a transcriptome-based computational method used to calculate T cell dysfunction and rejection scores to predict tumor immune escape potential and tumor response to immunotherapy;

[0065] (5) Tumor mutation burden (TMB) calculation: the total number of somatic nonsynonymous mutations in coding regions, including missense, nonsense, splice site, and frameshift mutations;

[0066] (6) Calculation of microsatellite stability (MS): Microsatellite sequences are composed of short, repetitive nucleotide sequences. Standard microsatellite loci include BAT25, BAT26, D5S346, D2S123, and D17S250.

[0067] 7. CAFscore evaluates the differences in sensitivity of tumor cells to targeted therapy and chemotherapy drugs, as follows:

[0068] Through the CTRP2.0 and PRISM databases including small molecule compounds, chemotherapeutic drugs, etc., effective small molecule compounds were screened through cancer drug sensitivity genomics (https: / / www.cancerrxgene.org / ) and the "pRRopheticR" package. Effective targeted therapy small molecule compounds and chemotherapeutic drugs were screened, and the half-maximal inhibitory concentration (IC50) was calculated by ridge regression to evaluate the sensitivity to targeted therapy and chemotherapeutic drugs, and 10-fold cross-validation was used to evaluate the accuracy of CAFscore in predicting sensitivity to targeted therapy and chemotherapeutic drugs.

[0069] 8. Verify the sensitivity of the prediction model for colon cancer prognosis and treatment resistance prediction in the validation set, which is the same as steps 5 to 7 above, except that the training set is replaced by the validation set.

[0070] R software (version 4.0.3, http: / / www.r-project.org) involved in the above-mentioned embodiment is used to analyze relevant data and drawing. " EPIC " algorithm, " MCPcounter " algorithm, " survival " R package, " rms " R package, " ggDCA " R package, " glmnet " R package, " limma " R package, " Survminer " R package, " gsva " R package, Kaplan-Meier method etc. are involved in the algorithm and R language package for existing technology, by directly calling realization. Student's t test, chi-square test, Fisher's exact probability test or Mann-Whitney U test are used to compare differences between groups. Statistical significance is set to p < 0.05, and all p values ​​are two-tailed.

[0071] 9. Experimental results:

[0072] (1) Association between CAFscore and clinical characteristics

[0073] First, the fibroblast infiltration analysis was performed on the gene expression matrices of TCGA and GSE39582 using the “EPIC” and “MCPcounter” algorithms, and the green gene module with the highest correlation with fibroblast infiltration was screened out based on Pearson correlation analysis ( Figure 1 A and 1B). Secondly, by MM>0.6 and GS>0.5 ( Figure 1 C and 1D), 244 overlapping genes were screened from the green gene module and were closely related to fibroblast infiltration ( Figure 1 E). By univariate Cox regression analysis, 11 genes were significantly associated with overall survival ( Figure 1 F). Subsequently, three genes with the largest weights were finally screened out through LASSO regression analysis to construct a fibroblast gene prognostic model ( Figure 1 G and 1H). The specific genes and their corresponding coefficients are shown in Table 1. The risk score of each sample is the sum of the fibroblast-related genes weighted by the corresponding coefficients, and the scores of each sample in the training set and the validation set are calculated. CAFscore = (mRNA expression of CRIP2) * 0.0189 + (mRNA expression of FSTL3) * 0.1223 + (mRNA expression of SLC2A3) * 0.1043.

[0074] Table 1 Three fibroblast-related genes and their coefficients included in the constructed molecular signature

[0075]

[0076] (2) CAFscore is an independent risk factor for the prognosis of colon cancer

[0077] Patients were divided into CAFscore high-risk group and low-risk group, and a prognostic prediction model was established in TCGA and validated using GSE39582. Kaplan-Meier survival analysis confirmed that patients in the CAFscore high-risk group had a poor prognosis, which was significantly different from that in the CAFscore low-risk group (p<0.001, Figure 2 A and 2F). After univariate COX analysis of the prognostic significance of the variables, multivariate COX analysis of TCGA and GSE39582 showed that high CAFscore was an independent predictor of poor overall survival (TCGA: hazard ratio = 2.744, 95% confidence interval = 1.017-7.410, p < 0.001; GEO: hazard ratio = 2.33, 95% confidence interval = 1.20-4.52, p < 0.001) ( Figure 2 B and 2G).

[0078] (3) CAFscore is superior to other biomarkers in prognostic prediction

[0079] The nomogram shows that CAFscore combined with existing clinical indicators can effectively evaluate the patient's 1, 3, and 5-year prognosis ( Figure 2 C and Figure 2 H). DCA curve analysis showed that the Nomogram score generated by combining CAFscore with existing clinical indicators was significantly superior to other biomarkers in predicting overall survival ( Figure 2 D and 2I). ROC curve analysis showed that the constructed Nomogram score had a high predictive efficacy for the 1-, 3-, and 5-year prognosis of colorectal cancer patients in TCGA and GSE39582 ( Figure 2 E and 2J).

[0080] (4) CAFscore is associated with immune escape characteristics of tumor cells

[0081] Compared with patients in the CAFscore low-risk group, patients in the CAFscore high-risk group had significantly increased expression of immune checkpoint-related genes, including CD274, CTLA-4, LAG3, etc. in TCGA and GSE39582 ( Figure 3 A). In addition, we found that the scores of APC co-inhibitory signals and T cell co-inhibitory signals were significantly increased in the CAFscore high-risk group ( Figure 3 B). Based on the above results, we further calculated T cell dysfunction and exclusion scores using the TIDE algorithm. Compared with the CAFscore low-risk group, the CAFscore high-risk group had significantly increased T cell dysfunction, tumor escape-related scores, and TIDE scores ( Figure 3C and 3D). Regarding immune cell infiltration in tumors, the CAFscore high-risk group had significantly increased abundance of regulatory T cells (Tregs) ( Figure 3 E). Therefore, the above findings suggest that the CAFscore high-risk group may lead to the formation of a “cold” tumor immune microenvironment.

[0082] (5) CAFscore predicts responsiveness to immunotherapy and is validated in immunotherapy cohorts

[0083] The accuracy of CAFscore in predicting the prognosis of immunotherapy was evaluated using an immunotherapy-treated urothelial carcinoma cohort, and it was found that the objective response rate in the high-risk CAFscore group was significantly lower than that in the low-risk CAFscore group (TCGA: 16% vs. 56%, p < 0.001; GEO: 20% vs. 60%, p < 0.001; Figure 4 A). ROC curve analysis showed that the constructed CAFscore had a high predictive efficacy for immunotherapy responsiveness in patients with colorectal cancer in TCGA and GSE39582 ( Figure 4 B). More importantly, in the immunotherapy cohort, the overall survival of the CAFscore high-risk group was significantly reduced compared with the CAFscore low-risk group (p = 0.039; Figure 4 C). CAFscore further optimizes the stratification of patients by existing immunotherapy markers. Among patients whose existing indicators suggest they are sensitive to immunotherapy, survival analysis predicts that patients in the CAFscore high-risk group have a poor prognosis and a potential mechanism of resistance to immunotherapy ( Figure 4 DF).

[0084] (6) CAFscore predicts tumor resistance to targeted therapy and chemotherapy drugs

[0085] The sensitivity of patients in the CAFscore low-risk group to targeted drugs was significantly higher than that in the CAFscore high-risk group (p<0.001; Figure 5 A and 5B). In addition, among the drugs in the CTRP and PRIME datasets, patients in the CAFscore low-risk group were significantly more sensitive to small molecule compounds than those in the CAFscore high-risk group (p < 0.001; Figure 5 A and 5B). It is worth noting that patients in the low-risk CAFscore group were significantly more sensitive to chemotherapy drugs than those in the high-risk CAFscore group, especially first-line chemotherapy drugs for colon cancer, such as fluorouracil, capecitabine, and oxaliplatin. The above conclusions provide a scientific theoretical basis for precise treatment of patients, reduce the occurrence of adverse reactions, and improve the prognosis of patients with colon cancer.

[0086] The above is only a preferred embodiment of the present invention, and is not any formal or substantial limitation of the present invention. It should be pointed out that ordinary technicians in this technical field can make several improvements and supplements without departing from the present invention, and these improvements and supplements should also be regarded as the protection scope of the present invention. Any technician familiar with this profession, without departing from the spirit and scope of the present invention, can make some changes, modifications and equivalent changes made by using the technical content disclosed above, which are equivalent embodiments of the present invention; at the same time, any changes, modifications and evolutions of any equivalent changes made to the above embodiments based on the essential technology of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A biomarker for predicting prognosis and treatment resistance of colon cancer, characterized in that: The biomarkers are differentially expressed gene combinations, including cysteine-rich protein 2 gene CRIP2 , follistatin-like protein 3 gene FSTL3 and glucose transporter 3 gene SLC2A3 .

2. Use of the biomarker detection product according to claim 1 in the preparation of a colon cancer prognosis and treatment resistance prediction product.

3. A method for constructing a colon cancer prognosis and treatment resistance prediction model based on fibroblast-related genes, characterized in that: include: S1 obtains transcriptome sequencing data and clinical information data of tumor tissue RNA from patients with colon cancer, constructs a gene expression matrix, and randomly divides it into a training set and a validation set; S2 performs fibroblast infiltration analysis on the gene expression matrix and calculates the fibroblast infiltration ratio of each sample; S3 performed WGCNA analysis on the gene expression matrix to screen out gene modules closely related to fibroblast infiltration; S4 used univariate Cox regression analysis to screen out the fibroblast-related prognostic core genes that were statistically correlated with overall survival. LASSO regression analysis was further used to determine the genes with the largest weights and their corresponding coefficients to construct a fibroblast gene-based prediction model CAFscore. The CAFscore is the sum of the mRNA expression levels of the genes with the largest weights weighted by the corresponding coefficients, specifically CAFscore=( CRIP2 mRNA expression level)*0.0189+( FSTL3 mRNA expression level)*0.1223+( SLC2A3 mRNA expression level)*0.1043.

4. The construction method according to claim 3, characterized in that: The steps of evaluating the prediction model for prognosis and treatment resistance prediction in the training set are also included, as follows: S5 Calculate the CAFscore of each sample in the training set, take the CAFscore at the minimum P value of the training set CAFscore as the optimal threshold, divide the samples into high-risk group and low-risk group, compare the survival difference between the two groups by Kaplan-Meier method, and evaluate the effectiveness of the prediction model by drawing the receiver operating characteristic curve ROC; S6 A nomogram combining CAFscore, gender, age, and TNM stage was constructed. The Kaplan-Meier analysis, univariate and multivariate Cox regression, and time-dependent receiver operating characteristic curve were used to evaluate the predictive effect of the nomogram on prognosis. The Harrell consistency index and decision curve analysis were used to evaluate the potential of CAFscore in predicting prognosis. S7 analyzed the immune microenvironment-related scores of the training set, compared the immune microenvironment-related scores of the CAFscore high-risk group and the low-risk group, and evaluated the accuracy of CAFscore in predicting tumor immune escape potential and immunotherapy efficacy; S8 By analyzing the immunotherapy cohort, the CAFscore of each tumor sample in the immunotherapy cohort was calculated, and the CAFscore at the minimum P value in the immunotherapy cohort was used as the optimal threshold. The samples were divided into a high-risk group and a low-risk group. The Kaplan-Meier method was used to compare the overall survival of the two groups of colorectal cancer patients and evaluate the accuracy of CAFscore in predicting the prognosis of immunotherapy. S9 screens effective small molecule targeted therapeutic compounds and chemotherapeutic drugs, calculates IC50 by ridge regression to evaluate the sensitivity of colon cancer patients in the training set to targeted therapy and chemotherapeutic drugs, and uses 10-fold cross-validation to evaluate the accuracy of CAFscore in predicting sensitivity to targeted therapy and chemotherapeutic drugs.

5. The construction method according to claim 3, characterized in that: The sensitivity of the prediction model for colon cancer prognosis and treatment resistance prediction was also validated in the validation set.

6. The construction method according to claim 3, characterized in that: In step S2, fibroblast infiltration analysis was performed using EPIC and MCPcounter algorithms.

7. The construction method according to claim 3, characterized in that: Step S3 specifically includes: S3-1 performs WGCNA analysis on the gene expression matrix to obtain the Pearson correlation coefficient between any two genes, selects the best soft threshold according to the distribution of gene connectivity, and divides the genes into modules; S3-2 calculated the Pearson correlation coefficient between the module and the fibroblast infiltration ratio, and screened out the core gene module with the highest correlation with fibroblast infiltration; S3-3 Calculate the correlation (MM) between genes and module feature values ​​in the core gene module, as well as the correlation (GS) between genes and phenotypic characteristics, and screen module genes with MM > 0.6 and GS > 0.5 as hub genes closely related to fibroblast infiltration.

8. A colon cancer prognosis and treatment resistance prediction model based on fibroblast-related genes constructed according to the construction method of any one of claims 3 to 7.

9. The use of the colon cancer prognosis and treatment resistance prediction model based on fibroblast-related genes according to claim 8 in establishing a Nomogram model for predicting the probability of overall survival of colon cancer, characterized in that: The specific application method is: using CAFscore, gender, age, and TNM stage as input factors of the Nomogram model, and outputting the probability value of the overall survival of patients with colon cancer after surgery; the method for obtaining the probability value is to correspond the sum of the scores corresponding to the four input factors to the value on the probability axis.

10. A colon cancer prognosis and treatment resistance prediction system based on fibroblast-related genes, characterized in that: The system comprises: A data acquisition module, used to obtain transcriptome sequencing data of RNA from tumor tissues of patients with colon cancer; Model scoring module, used for data acquisition module CRIP2、FSTL3 and SLC2A3 The mRNA expression level of is provided as input data to the prediction model of claim 8 to obtain CAFscore; The result output module is used to obtain the CAFscore in the model scoring module and output the prediction result, compare the CAFscore with the threshold, and if the risk value is greater than the threshold, the prediction result is output as high risk, and if the risk value is less than or equal to the threshold, the prediction result is output as low risk.

Images