Inferring Molecular Mechanisms of Complex Disease by Integrating Patterns of Epistasis with Functional Genomic Networks

Jackson Laboratory, Bar Harbor ME

Investigators

Linked publications & trials

Abstract

The goal of this proposal is to develop computational methods that will identify genes driving epistasis between quantitative trait loci. Genetic association studies in both humans and model organisms have been increasingly able to detect interactions among genetic variants that influence disease risk and pathogenesis. However, identifying which genes are represented by individual variants is a major challenge. Many variants influencing disease are situated between genes, and even those that are in genes may be representing an effect from a neighboring gene. In model organism studies the situation is even more problematic. In breeding experiments, genetic associations with phenotypes typically encompass large regions of DNA with many genes, and experimental follow-up to identify the genes responsible for the association is resource-intensive. To address this problem, we are developing computational tools to prioritize putative gene interactions in interacting genomic regions for biological follow-up. We use machine learning classifiers trained on functional gene-gene interaction networks and combinatorial optimization to identify likely candidate gene-gene interactions responsible for epistatic interactions between genomic regions. Our methods sift through enormous spaces of candidate interactions, retaining only those whose functional interactions identify as plausible for supporting the epistatic interaction. These tools hold promise to dramatically limit the resource-intensity of biological follow up of putative epistasis and to clarify the genetic architecture of complex disease.