Analysis of Methylation-driven Genes in Pancreatic Ductal Adenocarcinoma for Predicting Prognosis

Purpose: Considerable variations in methylation profile have been found in various cancers to modulate tumorigenesis and affect prognosis. To provide a theoretical basis for early detection, prognosis evaluation and targeted treatment for patients with pancreatic ductal adenocarcinoma: PDAC, this study identified methylation-driven genes in PDAC and explored their prognostic performance. Methods: The methylation, expression and clinical data of PDAC patients were extracted from TCGA database. Based on the beta-mixture model of the MethylMix R package, the differential methylation status and connection between methylation and expression degree were examined to screen out methylation-driven genes in PDAC. COX analyses and lasso regressions were applied to construct a linear risk model based on methylation-driven genes. Univariate and multivariate analyses were performed to ensure the risk model was an independent prognostic factor. Joint survival analyses of methylation and gene expression were conducted to explore the prognostic value of component genes. The methylation sites in the key genes were also investigated. Results: A total of 118 methylation-driven genes in PDAC were identified, and two genes (FOXI2, MYEOV) constituted the risk model whose AUC was 0.722 at one year of overall survival rate, displaying a better performance on survival prediction than other clinical features. Further survival analyses demonstrated that the expression of MYEOV and combined methylation and expression levels of the genes MYEOV and FOXI2 can be potential biomarkers for survival prediction and targets of drug manipulation of PDAC patients. Close relationships were discovered between two sites in MYEOV and one site in FOXI2 and the prognosis of PDAC patients. Conclusion: Concentrating on DNA methylation, our study identified potential biomarkers and developed a reliable short-term predictive model for prognosis of PDAC patients.