A robust eleven-genes prognostic model can predict overall survival in bladder cancer patients based on five cohorts
Abstract Background: Bladder cancer is the tenth most common cancer in the world, but existing biomarkers and prognostic models are limited.Method: In this study, we used four bladder cancer cohorts from The Cancer Genome Atlas and Gene Expression Omnibus databases to perform univariate Cox regression analysis to identify common prognostic genes. We used selected genes to construct a prognostic model. Kaplan-Meier analysis, Receiver Operating Characteristic curve, and univariate and multivariate Cox analysis were used to evaluate the prognostic model for the four cohorts. Finally, a co-expression network, CIBERSORT, and ESTIMATE algorithm were used to explore the mechanism related to the model.Results: A total of 11 genes were identified from the four cohorts to construct the prognostic model, including eight risk genes (SERPINE2, PRR11, DSEL, DNM1, COMP, ELOVL4, RTKN, and MAPK12) and three protective genes (FABP6, C16orf74, and TNK1). The model and the 11 genes have excellent performance in predicting overall survival and have been confirmed in 5 cohorts. The model's predictive ability is stronger than other clinical features and has practical significance in clinical application.Through the analysis of the weighted co-expression network, the gene module related to the model was found, and the key genes in this module were mainly enriched in the items related to the tumor microenvironment. When comparing the level of immune cell infiltration in high-risk samples, B cell memory showed low infiltration in high-risk patients. Furthermore, in the case of low B cell memory infiltration and high-risk score, the prognosis of the patients was the worst.Conclusion: The model we developed has strong stability and good performance and can stratify the risk of bladder cancer patients, to achieve individualized treatment.