Aberrant promoter methylation and ensuing abnormal gene expression are important epigenetic mechanisms that contribute to colorectal oncogenesis. Yet, the prognostic significance of such methylation-driven genes in colorectal cancer (CRC) remains obscure. Herein, a total of 181 genes were identified as the methylation-driven molecular features of CRC by integrated analysis of the expression profiles and the matched DNA methylation data from The Cancer Genome Atlas (TCGA) database. Among them, a five-gene signature (POU4F1, NOVA1, MAGEA1, SLCO4C1, and IZUMO2) was developed as a risk assessment model for predicting the clinical outcomes in CRC. The Kaplan–Meier analysis and Harrell’s C index demonstrated that the risk assessment model significantly distinguished the patients in high or low-risk groups (p-value < 0.0001 log-rank test, HR: 2.034, 95% CI: 1.419-2.916, C index: 0.655). The sensitivity and specificity were validated by the receiver operating characteristic (ROC) analysis. Furthermore, different pharmaceutical treatment responses were observed between the high-risk and low-risk groups. Indeed, the methylation-driven gene signature could act as an independent prognostic evaluation biomarker for assessing the OS of CRC patients and guiding the pharmaceutical treatment. Compared with known biomarkers, the methylation-driven gene signature could reveal cross-omics molecular features for improving clinical stratification and prognosis.