Statistical Methods For Gene/environment Interaction And Genetic Susceptibility

National Institute Of Environmental Health Sciences

Investigators

Linked publications, trials & patents

Abstract

Identification of causative SNPs in a genome-wide study can be challenging when individual SNPs have small marginal effects because, to avoid excessive false positive conclusions, testing thresholds must reflect the large number of SNPs under study. For complex diseases, particular combinations of SNPs may dramatically increase disease risk through epistasis or gene-gene interactions. We are currently investigating the use of a machine learning technique with case-parents data for the discovery of sets of SNPs that together cause disease. We implemented an existing stochastic search algorithm, an evolutionary or genetic algorithm (called GA), to find multiple sets of d SNPs that are predictive of disease (here d is a small number, say 2 to 6). By cataloging those SNPs which appear most frequently among the sets that are predictive of disease, we hope to uncover sets of epistatic causative SNPS. Our first paper describing our algorithm, called GADGETS, was published last year; we showed through simulations that GADGETS can uncover epistatic sets of 3-6 SNPs from data on 10,000 SNPs in 1000 nuclear families. We have also developed and evaluated a permutation test procedure to probe whether or not the risk increase attributed to a nominated set of SNPs arises from epistatic interactions rather than solely from marginal effects of the individual SNPs. In ongoing work, we are modifying and extending our algorithm to uncover sets of SNPs whose epistatic effects may differ between levels of an environmental factor and sets where maternal and offspring SNPs may together influence risk. (see also Z01 ES040007; PI Clare Weinberg; Min Shi and Michael Nodzenski are also within-lab collaborators on this project; there time is allocated in Weinberg's project but not in this one.)