Despite extensive biological and clinical studies, including comprehensive genomics and transcriptomics analysis, pancreatic ductal adenocarcinoma (PDAC) remains a devastating disease, with poor survival and no effective therapies to date. Correlation networks are emerging as a powerful approach to infer tumor biology and to prioritize candidate genes as biomarkers or drug targets. In this study we applied a weighted co-expression analysis to the functionally relevant proteome of 20 surgically resected patients with PDAC. We obtained twelve modules with overlapping yet distinct biology, which implicated metabolism and ECM complexes in several modules. Notably, one module enriched for metabolic processes and epithelial-mesenchymal-transition (EMT) was significantly associated with overall survival (p=0.01) and was validated in public RNA data (p=0.02). The prognostic value of three proteins (SPTBN1, KHSRP and PYGL) belonging to this module was confirmed using immunohistochemistry in a cohort of 82 radically resected patients.