bioRxiv2018
Liver cancer is among the top deadly cancers worldwide with a very poor prognosis, and the liver is a particularly vulnerable site for metastasis of other cancers. In this study, we developed a novel computational framework for the stage-specific analysis of hepatocellular carcinoma initiation and progression. Using publicly available clinical and RNA-Seq data of cancer samples and controls, we annotated the gene expression matrix with sample stages. We performed a linear modelling analysis of gene expression across all stages and found significant genome-wide changes in gene expression in cancer samples relative to control. Using a contrast against the control, we were able to identify differentially expressed genes (log fold change >2) that were significant at an adjusted p-value < 10E-3. In order to identify genes that were specific to each stage without confounding differential expression in other stages, we developed a full set of pairwise stage contrasts and enforced a p-value threshold (<0.05) for each such contrast. Genes were specific for a stage if they passed all the significance filters for that stage. Our analysis yielded two stage-I specific genes (CA9, WNT7B), two stage-II specific genes (APOBEC3B, FAM186A), ten stage-III specific genes including DLG5, PARI and GNMT, and ten stage-IV specific genes including GABRD, PGAM2 and PECAM1. Of these, only APOBEC3B is an established cancer driver gene. DLG5 was found to be tumor-promoting contrary to the cancer literature on this gene. Further, GABRD, well studied in literature on other cancers, emerged as a stage-IV specific gene. Our findings could be validated using multiple sources of omics data as well as experimentally. The biomarkers identified herein could potentially underpin diagnosis as well as pinpoint drug targets.