The ability to inexpensively genotype individuals at high density across the human genome has enabled hundreds of genome-wide association studies that search for specific variants linked with diseases of interest. These efforts have been extremely productive in that they have produced large numbers of new candidate loci that were previously uncharacterized. However, for most diseases, the cumulative risk explained by the discovered loci explains only a small fraction of the estimated total disease heritability. One possible explanation is the potential for genetic interactions among variants. Most statistical geneticists agree that genetic interactions contribute to complex diseases, but traditional methods for identifying genetic interactions, which focus on testing individual locus pairs, lack statistical power due to the large number of possible variant combinations present within a human genome. Thus, if we hope to fully understand the complex genetics underlying human diseases, there is a critical need for new paradigms for identifying genetic interactions from population genetic data. A new perspective for approaching the problem of identifying genetic interactions emerges from extensive work in model organisms, where systematic efforts have constructed millions of double mutants using reverse genetic approaches over the past decade. One striking outcome of these studies is that genetic interactions are organized into highly structured sets, bridging or connecting all possible gene pairs that belong to functionally compensatory but distinct biological pathways. This exquisite structure can be exploited to help identify genetic interactions in humans. Based on this principle, we developed a novel computational approach, called BridGE, for discovering genetic interactions between pathways from human GWAS data. The specific objective of this application is to extend the BridGE algorithm to increase the utility of this promising approach for studying the genetic basis of cancer risk. We will accomplish this through two specific aims: (1) developing a context-sensitive BridGE algorithm that integrates tissue/cell-type specific functional genomic data to improve discovery of novel genetic interactions relevant to specific cancers; and (2) extend the BridGE algorithm to enable the analysis of genetic interactions involving somatic mutations. The proposed research is innovative because the completed method will provide a strong alternative to the prevailing paradigm for analysis of genome-wide association study data focused on single or sets of loci that are individually associated with disease risk. Our proposed approach offers a means to overcome a major statistical barrier and extract previously hidden genetic interactions contributing to cancer risk.
Although genomic technology enables efficient sequencing of human genomes, we are still not able to accurately predict disease states from genome sequence. One reason for this gap is that we lack methods for finding combinations of variants that lead to disease. This project focuses on developing a new computational approach for identifying combinations of genetic variants that determine risk of developing specific cancers.