Construction Of A Prognostic Model For Head And Neck Squamous Cell Carcinoma Based On Bioinformatics Analysis Of Immune-related Genes

Objective:Head and neck squamous cell carcinoma,as the sixth most common tumor in the world,has always threatened human life and health.An accurate clinical decision will affect the clinical outcome of the patient.The prognosis is one of the main factors that promote clinical decision-making.Therefore,accurately predicting the patient’s prognosis not only means more accurate clinical decision-making,but also means better clinical outcomes.However,common clinical staging systems have problems such as inaccurate predictions in predicting the prognosis of patients.At the same time,with the rapid development of immunotherapy,although the results of its clinical research are exciting,certain problems of immunotherapy such as low response rate are inevitably exposed.In order to avoid patients blindly following immunotherapy,it is necessary to develop biomarkers to predict the response of immunotherapy.Although the FDA has approved certain biomarkers to predict the status of immunotherapy,the results are not optimistic.A large number of studies have reported that certain immune-related genes are not only related to the prognosis of patients but also closely related to immunotherapy.Therefore,the scientific problem to be solved in this study is whether to establish a prognostic model constructed by immune-related genes to predict patient prognosis more accurately,and predict the immunotherapy of the patients.Methods:Data download and processing of head and neck squamous cell carcinoma: 1.Download the dataset of TCGA-HNSC,GSE41613,GSE42743,GSE65858;2.Use the chi-square test to explore the relationship between the clinical variables of the four data and the clinical outcome.Construction of a prognostic model of immune-related genes: 1.Analyze the TCGAHNSC by univariate Cox to determine immune genes related to prognosis.To further reduce genes,determine the optimal gene set through LASSO and cross-validation;2.Use multivariate Cox analysis and further reduce based on Akaike information criterion.The number of genes is the final gene,and the random survival forest is used to determine the importance of the gene;3.Perform differential analysis of model genes to explore the behavioral patterns of genes;4.Build a model and distinguish between high and low risk groups;5.Use box plots to determine the relationship between risk scores and clinical staging.Use a variety of methods to evaluate and verify the predictive ability of the model: 1.Internal and external verification of the model that is to use K-M survival analysis and ROC evaluation model predictive ability.At the same time,multivariate Cox analysis is used to determine whether the model can be used as an independent predictive risk factor.Use ROC to evaluate the ability of prognostic models and clinical staging systems to predict prognosis;2.Construct a nomogram to quantify each the risk value of clinical variables and risk scores,so as to more accurately assess the prognosis.Use the degree of discrimination and the degree of calibration to evaluate the distinguishing ability and predictive ability of the model;3.Use the decision curve to evaluate the clinical net benefit of the patient;4.Use the HPA database immunohistochemical results to assist in verifying the protein expression of model genes;5.q PCR verification of model gene expression.The prognostic model predicts immunotherapy: 1.Explore the expression of immune checkpoint-related genes in high and low risk groups for exploration Immunotherapy as a pavement;2.Use CIBERSOFT and ss GSEA methods to evaluate the difference between immune cells and immune functions in the high and low risk groups to explore the immunotherapy conditions of the high-and low-risk groups;3.Use the GSEA method to explore differential pathways between high-and low-risk groups;4.Assess the relationship between risk scores and TMB;5.Immunophenoscores to determine which immunotherapy is better for high-and low-risk groups.Results:Data processing: 1.The basic clinical characteristics table shows that certain clinical variables are related to the clinical outcome of the patient.Successfully constructed a prognostic model based on immune-related genes: 1.Univariate Cox analysis confirmed that 162 immune-related genes are related to prognosis.When 24 genes are confirmed by LASSO regression and cross-validation,the error is minimal;2.The multivariate Cox analysis based on the Akaike information criterion will eventually reduce the 24 genes to 11 genes.Random survival forest analysis found that the importance of 11 genes were all non-negative;3.The final model was constructed using 11 genes and the patients were divided into high and low risk groups;4.TNFSRF25,TNFSRF40,OLR1,AIMP1,MAP2K1,PDGFA are highly expressed in tumor tissues,while CTSG,SEMA5 A,PTX3 are low-expressed genes in tumor tissues.There is no significant difference in the expression of CHGB and IKBKB in normal tissues and tumor tissues;5.Risk scores are significantly different in some clinical stages,such as T1 and T2.The prognostic model has good predictive ability and reliability: 1.K-M survival analysis of 4 data sets found that the survival period and overall survival rate of the lowrisk group were longer than those of the high-risk group,and the ROC evaluation found that the model was in 4 data sets,the AUC of the sets are all greater than 0.6.Multivariate Cox regression analysis verified that the model can be used as an independent predictor(HR = 1.394,95% CI = 1.289-1.508,P <0.001).At the same time,the AUC of other clinical variables is less than 0.6;2.The nomogram is successfully constructed,and the model has a certain degree of discrimination and calibration;3.Through the decision curve,it is found that the net benefit of patients using the model is higher than that of patients using other variables alone;4.The immunohistochemical results of IKBKB,AIMP1,MAP2K1,PDGFA,PTX3,SEMA5 A show Its expression trend in normal tissues and tumor tissues is basically the same as the trend presented by the different analysis results.OLR1 and TNFRSF25 lack corresponding results in the HPA database.The staining intensity of CHGB,CTSG,and TNFRSF4 in normal tissues and tumor tissues in the HPA database were all negative;5.q PCR results show TNFRSF4 has no significant difference between normal cells and tumor cells,MAP2K1 has no significant difference between PJ15 tumor cells and normal cells,and other genes have significant differences between normal cells and tumor cells.The prognostic model has the ability to predict immunotherapy: 1.The expression of PD-1/PD-L1 in the low-risk group is higher than that of the high-risk group Group;2.CIBERSOFT and ss GSEA showed that the immune cells and immune functions of the low-risk group were stronger than those of the high-risk group;3.GSEA revealed that the immune-related pathways of the low-risk group were more active,and the carbohydrate metabolism pathway of the high-risk group was more active;4.The TMB of the high-risk group is significantly higher than that of the low-risk group;5.The low-risk group and the higher-risk group are more suitable to receive anti-PD-1 and anti-CTLA4.Conclusion:This study successfully constructed a prognostic model based on immune-related genes,which improved the accuracy of predicting the prognosis of patients.At the same time,the model can better predict the situation of immunotherapy,that is,the response rate of patients in the low-risk group will be higher when receiving immunotherapy.