Gene Selection and Enrichment for Microarray Data—A Comparative Network Based Approach

Abstract

Gene selection plays a vital role in understanding the disease progression and further it helps in understanding the therapeutic targets. Most of the genes available in micro array data are not informative for a particular disease of interest. Study of functional analysis and interaction structure of genes plays a vital role in selecting genes associated to complex diseases. This work uses two different network based approaches for gene selection and compares the biological and statistical enrichment of selected genes. Functional modules in the gene expression data are obtained using Gene Correlation Network (GCN) and marker genes in the modules are identified using R package Weighted Gene Co- expression Analysis (WGCNA). WGCNA is considered to be one of the best methods for analysis of global GCN using a suitable threshold that leads to a network with scale free topology. The differentially co-expressed genes are then compared with the existing gene selection approach which integrates the selected co-expressed gene modules with protein-protein interaction (PPI) network. Observation shows that using PPI network which is generated using multitude of high throughput experiments and available in public data bases selects more disease specific genes in comparison to constructed GCN. The study shows that integrative network analysis to find genes may provide greater insight in underlying biological response.