Objective：To investigate the DNA methylation sites related to the prognosis of patients with pancreatic ductal adenocarcinoma（PDAC）at the whole-genome level by utilizing The Cancer Genome Atlas（TCGA）database. Methods：The clinical data，DNA methylation data detected on Illumina Humanmethylation450k beadchip and transcriptome data produced by Illumina Hiseq of PDAC patients were downloaded from TCGA database（version 2016_01_28）. We finally chose 179 cases containing both of clinical data and methylation expression profiles to analyze the effect of methylation level on survival by Cox’s proportional hazards model. Five factors including age，sex，location，histological grade and pathological stage were used to correct hazard ratio and P value. Next we chose 173 cases containing both of gene expression and methylation expression profiles to explore correlation between methylation level and mRNA expression level. Furthermore，we also evaluated the expression level of mRNAs with the prognosis. Results：The age of 179 patients was 64.64±10.96 years and that of 173 patients was 64.46 ± 10.91 years. The median survival times were both 20.1 months. Among the top 20 methylation sites，hypermethylation of 11 sites and hypomethylation of other 9 sites were associated with worse prognosis. The strongest site influencing the survival for PDAC in this study was cg01656216，which was located in the 5′UTR region of ZNF438（P=4.11×10-6，HR=0.37，95%CI：0.24-0.56）. The methylation levels of 8 sites showed significant inverse correlation with mRNA expression levels（r<-0.3，P < 0.05）. In addition，the mRNA expression level of PKP3（cg20268054）and EFNB2（cg22179913）were also related to the survival of PDAC（HR=1.66，95%CI：1.06-2.61，P=0.027；HR=1.86，95%CI：1.21-2.88，P=0.005）. Conclusion：Based on the analysis of TCGA database，methylation sites in PKP3 and EFNB2 genes regions are significantly associated with PDAC prognosis，whose potential for predicting prognosis of PDAC can be further studied.