An Outranking Approach for Gene Prioritization Using Multinetworks

High-throughput experimental techniques such as genome-wide association studies have been instrumental in the identification of disease-associated genes. These methods often produce large lists of disease candidate genes which are time-consuming and expensive to experimentally validate. Computational gene prioritization methods are required to identify relevant genes from a larger pool of candidates. Research has shown that the integration of diverse “omic” evidence can reduce the candidate-gene search space. In this paper we present a general framework that integrates “omic” data using a multinetwork approach and topological analysis to prioritize disease-candidate genes. Specifically, we propose a data integration method within a multicriteria decision analysis context using aggregation mechanisms based on decision rules identifying positive and negative criteria for judging gene-candidates ranks. The proposed multinetwork disease gene prioritization method is applied to the prioritization of disease genes in ovarian cancer progression. Using this approach we uncovered known ovarian cancer genes GSTA1, ERBB2, IL1A, MAGEB2, along with significantly enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways ErbB signaling and pathways in cancer. Relatively high predictive performance (area under Receiver Operating Characteristic [ROC] curve 0.704) was observed when classifying epithelial ovarian high-grade serous carcinoma cancer early and late stage RNA-Seq expression profiles from individuals using 10-fold cross-validation.