The Generation and Validation of a 20‐Genes Model Influencing the Prognosis of Colorectal Cancer

ABSTRACT Colorectal cancer is a common malignant tumor with high incidence affecting the digestive system. This study aimed to identify the key genes relating to prognosis of colorectal cancer and to construct a prognostic model for its risk evaluation. Gene expression profiling of colorectal cancer patients, GSE17537, was downloaded from Gene Expression Omnibus database (GEO). A total of 55 samples from patients ranging from stages 1 to 4 were available. Differentially expressed genes were screened, with which single factor survival analysis was performed to identify the response genes. Interacting network and KEGG enrichment analysis of responsive genes were performed to identify key genes. In return, Fisher enrichment analysis, literature mining, and Kaplan–Meier analysis were used to verify the effectiveness of the prognostic model. The 20‐gene model generated in this study posed significant influences on the prognoses (P = 9.691065e‐09). Significance was verified via independent dataset GSE38832 (P = 9.86581e‐07) and GSE17536 (P = 2.741e‐08). The verified effective 20‐gene model could be utilized to predict prognosis of patients with colorectal cancer and would contribute to post‐operational treatment and follow‐up strategies. J. Cell. Biochem. 118: 3675–3685, 2017. © 2017 Wiley Periodicals, Inc. The 20‐gene model generated in this study posed significant influences on the prognoses (P = 9.691065e‐09). Significance was verified via independent dataset GSE38832 (P = 9.86581e‐07) and GSE17536 (P = 2.741e‐08). The verified effective 20‐gene model could be utilized to predict prognosis of patients with colorectal cancer and would contribute to post‐operational treatment and follow‐up strategies.