Disease Gene Prioritization Based On PPI Network And Transcriptional Co-expression Profiles

We made efforts to integrate protein-protein interaction network and microarray co-expression data to do gene prioritization. We compared the performances in gene prioritization based on disease-specific and non-specific microarray data firstly, and the former performed better as results. Then we used three different protein-protein interaction(PPI) networks and constructed a co-expression network by integrating multiple microarray experiments data and evaluated the advantage and disadvantage among these networks. Finally, we developed a novel gene prioritization method by merging co-expression and differential expression information generated from microarray data as well as PPI network. Present gene prioritization methods based on integrating PPI network and differential expression are defective in some aspects as irrelevant genes could be scored high due to network diffusion since candidate genes are scored by single differentially-expressed gene. In order to solve this problem, we proposed a new method, named as Group Rank, in which candidate genes were ranked based on co-expressed gene groups generated by gene clustering instead of individual gene. Only if the candidate gene was related to all genes in one group, this group could rank it high. The results of applying Group Rank on validation of some cancers showed that Group Rank performed well. We ran functional annotation enrichment analysis in effective groups which made major contributions in disease gene rank and found that these groups were strongly related to cancer development. It indicated that Group Rank, a gene-cluster based method, could be helpful to better understand the possible mechanism in important physiological and pathological processes of disease.