A Graph Community Approach for Constructing microRNA Networks

Abstract

Network integration methods are critical in understanding the underlying mechanisms of genetic perturbations and susceptibility to disease. Often, expression quantitative trait loci (eQTL) mapping is used to integrate two layers of genomic data. However, eQTL associations only represent the direct associations among eQTLs and affected genes. To understand the downstream effects of eQTLs on gene expression, we propose a network community approach to construct eQTL networks that integrates multiple data sources. By using this approach, we can view the genetic networks consisting of genes affected directly or indirectly by genetic variants. To extend the eQTL network, we use a protein-protein interaction network as a base network and a spin glass community detection algorithm to find hubs of genes that are indirectly affected by eQTLs. This method contributes a novel approach to identifying indirect targets that may be affected by variant perturbations. To demonstrate its application, we apply this approach to study how microRNAs affect the expression of target genes and their indirect downstream targets in ovarian cancer.