Background Uncontrolled proliferation is the most prominent biological feature of tumors. To rapidly proliferate and maximize the use of available nutrients, tumor cells regulate their metabolic behavior and the expression of metabolism-related genes (MRGs). In this study, we aimed to construct prognosis models for colon and rectal cancers, using MRGs to indicate the prognoses of patients. MethodsWe first acquired the gene expression profiles of colon and rectal cancers from the TCGA and GEO database, and utilized univariate Cox analysis, lasso regression, and multivariable cox analysis to identify MRGs for risk models. Then GSEA and KEGG functional enrichment analysis were utilized to identify the metabolism pathway of MRGs in the risk models and analyzed these genes comprehensively using GSCALite.ResultsEight genes (CPT1C, PLCB2, PLA2G2D, GAMT, ENPP2, PIP4K2B, GPX3, and GSR) in the colon cancer risk model and six genes (TDO2, PKLR, GAMT, EARS2, ACO1, and WAS) in the rectal cancer risk model were identified successfully. Multivariate Cox analysis indicated that the models predicted overall survival accurately and independently for patients with colon or rectal cancer. Furthermore, we identified the metabolism pathway of MRGs in the risk models and analyzed these genes comprehensively. Then, we verified the prognosis model in independent colon cancer cohorts (GSE17538) and detected the correlations of the protein expression levels of GSR and ENPP2 with prognosis for colon or rectal cancer. ConclusionsOur findings demonstrated that 14 MRGs are potential prognostic biomarkers and therapeutic targets of colorectal cancer.