Extensive research in plants, model organisms, and humans has demonstrated the near ubiquity of gene-environment (GxE) interactions in complex trait variation. Human geneticists have a long list of genetic variants that define an elevated disease risk whose manifestation is only triggered under particular environmental insults. What is lacking is the application of a sound statistical framework for uncovering, at a genome-wide scale, the role of environmental context in shaping the manifestation of genetic variation. As popular as mixed linear models are for problems of this sort, many of the assumptions of these models are violated in a way that can both reduce the power of the tests and generate false positives. I propose to develop and apply strategies for detecting SNP-disease associations in an environmental context using the Patient Rule Induction Method (PRIM). The PRIM approach not only offers a way to discover context-dependent genotype effects, but with a view towards improving prediction of the phenotype through assessment of the impact of context dependence on disease risk. I will develop extensions of PRIM for performing genome-wide scans for context dependence, and through extensive simulations I will test and validate the approach. This work will aim to identify both common and rare genetic variants that play a role in ischemic heart disease and high density lipoprotein levels in different environmental contexts (such as smoking, high blood pressure, obesity). In instances of strong context-dependence, targeted modification of environmental risks for individuals who harbor risk alleles may be especially efficacious to reduce the risk of disease.
By considering the effect of each SNP in the genome on ischemic heart disease risk and high density lipoprotein levels in the context of contributing environmental factors we will generate genome-wide scans to detect context dependent effects. The contexts of interest are environmental factors such as smoking, diet and exercise, which are known to impact cardiovascular disease risk. This method will also accommodate etiological heterogeneity, where a given disease phenotype may be caused by more than one combination of genes and environments. Specific Aims The primary goal of this project is to improve our understanding of the contribution of genotype to disease risk in the context of environmental risk factors. I will use extensions of the Patient-Rule Induction (PRIM) method developed by Dyson et al. (2009) to detect context-dependence on a genome-wide scale, which I refer to as 'PRIMWAS' in this proposal.
