A novel approach to significant pathway identification using pathway interaction network from PPI data

Discovering and understanding a variety of genetic markers (e.g., SNPs, genes, pathways) related to a certain phenotype of interest is one of the fundamental challenges in recent genetic studies. For this purpose, conventional methods have usually done by detecting significantly differentially expressed genes or SNPs between case and control samples. However, such approaches often produce a large list of potential markers which contain only a few genetic markers truly associated with a given phenotype. That is, their results often include too many false positives about phenotype relevant markers. As an alternative, lately, several studies have attempted to identify significant functional modules (or pathways) each of which contains a set of genes involved in a particular biological function or process. These pathway marker findings could be better in uncovering complex disease mechanism than individual gene marker findings. This paper investigates a novel approach to significant pathway identification that exploits pathway interaction network (PIN) derived from protein-protein interaction (PPI) data. Specifically, we first construct PIN which indicates the hidden associations between biological pathways, by exploring PPI data and then prioritize pathway nodes over PIN with PIN-PageRank algorithm to identify significant pathways. In this procedure, we employ differentially expressed gene profiles for PIN node initialization. To evaluate efficacy and usability of our proposed approach, we performed experiments for the identification of breast cancer relevant pathways and compared these results with existing approaches like GSEA and DAVID. Overall, it was observed that our PIN-PageRank approach outperforms existing approaches in finding significant pathways.