Utilizing Gene Co-expression Network for Identifying Subnetwork Biomarkers for Cancer

Abstract

Cancer is one of the major causes of death worldwide. Gene biomarkers identification for diagnosis, prognosis or predictive cancer is challenging work. This work aims to study the applicability of gene co-expression network (GCN) to identify gene subnetwork biomarkers for cancer using a previously developed subnetwork-based method. Four lung cancer expression datasets, gene-set, protein-protein interaction (PPI) and gene-gene interaction (GGI) data from public databases were used. The GCN was constructed using two criteria. The GCN constructed by using whole genes in the expression data with minimum spanning of the interaction within the network termed MST-based Gene Co-expression Network (MST-GCN) and the GCN was constructed by using gene members of a certain gene-set termed a Gene-set-based Gene Co-expression Network (gGCN). The subnetworks that resulted from MST-GCN and gGCN were compared with subnetworks that resulted from PPI and GGI data. The identified subnetworks were evaluated by classification performance and the overlapped gene with cancer related genes retrieved from a public database. The gGCN resulted in subnetworks that improved classification performance when compared with other network data. The identified subnetworks results from GGI contained more lung cancer related genes while the results from GCN and PPI contained more well-known lung cancer related genes.