Analysis of a large cohort of pancreatic cancer transcriptomic profiles to reveal the strongest prognostic factors

Abstract

AbstractPancreatic adenocarcinoma remains a leading cause of cancer‐related deaths. In order to develop appropriate therapeutic and prognostic tools, a comprehensive mapping of the tumor's molecular abnormalities is essential. Here, our aim was to integrate available transcriptomic data to uncover genes whose elevated expression is simultaneously linked to cancer pathogenesis and inferior survival. A comprehensive search was performed in GEO to identify clinical studies with transcriptome‐level gene expression data of pancreatic carcinoma with overall survival data and normal pancreatic tissues. After quantile normalization, the entire database was used to identify genes with altered expression. Cox proportional hazard regression was employed to uncover genes most strongly correlated with survival with a Bonferroni corrected p < 0.01. Perturbed biological processes and molecular pathways were identified to enable the understanding of underlying processes. A total of 16 available datasets were combined. The aggregated database comprised data of 1640 samples for 20,443 genes. When comparing with normal pancreatic tissues, a total of 2612 upregulated and 1977 downregulated genes were uncovered in pancreatic carcinoma. Among these, we found 24 genes with higher expression which significantly correlated with overall survival length also. The most significant genes were ANXA8, FAM83A, KRT6A, MET, MUC16, NT5E, and SLC2A1. These genes remained significant after a multivariate analysis also including grade and stage. Here, we assembled a large‐scale database of pancreatic carcinoma samples and used this cohort to identify carcinoma‐specific genes linked to altered survival outcomes. As our analysis focused on genes with higher expression, these could serve as future therapy targets.