Background: Abnormal methylation is associated with the survival of colon cancer. This study intended to discover a significant model based on methylation-driven genes (MDGs) and screen relative risk loci to assist with determining the prognoses of colon cancer patients.
Methods:We downloaded transcriptome expression profiles and 450K methylation data from the TCGA database. We then collected the two normalized profiles and utilized the MethylMix package to identify a significant signature showing the aberrantly methylated events highly correlated with expression levels. Also, functional enriched pathway analysis based on the ConsensusPathDB database was conducted to further explore the underlying cancer-related crosstalk among the identified MDGs. To find the significant MDGs for prognosis, we applied a univariate Cox regression model, and the hub signature was identified based on the stepwise regression method. A risk model based on MDGs was constructed from the multivariate Cox analysis, and a receiver operating characteristic (ROC) curve was drawn to assess the predictive value of the MDG signature. Additionally, the Kruskal-Wallis (K-W) test was conducted to compare differential distributions of risk scores across groups of clinical variables. Furthermore, the methylation sites relating to the hub genes were screened out and the prognostic genes were searched using the Cox regression method.Last, we carried out gene set enrichment analysis (GSEA) with the risk score levels serving as the phenotype base on the JAVA platform.Results: A total of 514 colon cancer samples with transcriptome profiles, including 473 tumor samples and 41 matched normal samples, were downloaded. We also obtained 351 methylation profiles comprising 314 tumor samples and 37 normal samples. The 320 MDGs identified by MethylMix were enriched in the generic transcription pathway, RNA polymerase II transcription, activation of SMO, or glutathione metabolism. Furthermore, a 10-MDGs signature was selected as the hub prognostic marker, and the risk model was constructed from the multivariate Cox regression results. We also discovered multiple specific methylated sites that were highly associated with survival. Finally, the GSEA results suggested that several enriched pathways were associated with the identified risk drivers, including extracellular matrix (ECM) receptor interaction, chemokine receptor interaction, and pathways in cancer, as well as calcium signaling pathways.
Conclusions:We conducted a comprehensive investigation of the molecular mechanisms in colon cancer by discovering the risk methylation-driven signature combined with relative methylated sites and constructing a risk model to predict prognosis.