Novel kernel methods for detecting gene-environment. K. A. Broadaway1, R. Duncan1, L. M. Almli2, K. J. Ressler2, B. Bradley2,3, M. P. Epstein1 1) Department of Human Genetics, Emory University School of Medicine, Atlanta, GA; 2) Department of Psychiatry and Behavioral Sciences, Emory University, Atlanta, GA; 3) Atlanta VA Medical Center, Atlanta, GA.

   The etiology of complex traits likely involves the effects of genetic and environmental factors, along with complicated interaction effects between them. Consequently, there has been interest in applying genetic association tests of complex traits that account for potential modification of the genetic effect by the presence of an environmental exposure. One can perform such an analysis using a joint test of gene and gene-environment interaction (GxE). GxE testing is recommended when a study of interactions is expected to provide evidence of an association between genotype and phenotype that would not be found if only the main effects of exposures were examined; for example, in a situation of complete interaction, where a genotype has an effect on phenotype in the presence of an environmental exposure, but no effect in absence of the exposure. However, when the genotypic effect in the absence of environmental exposure is greater than zero, a main effect test is expected to rival or outperform the joint test. When GxE is suspected, an optimal association test would be one that remains powerful under a variety of models, ranging from those of strong GxE effect (complete interaction) to little or no GxE effect. To fill this demand, we have extended a kernel-machine based approach for association mapping to consider joint tests of gene and GxE by incorporating a garrote parameter into the kernel framework. The kernel-based approach to GxE testing is promising for several reasons. First, since multiple typed markers are likely to be in linkage disequilibrium with the causal variant, joint consideration of these variants will capture the effect of a true causal variant more effectively than independent testing. Second, grouping variants together into sets along the genome allows epistatic interactions within the gene to be implicitly considered in the association test. Third, the kernel approach readily allows for inclusion of covariates, such as principal components to account for population stratification. We illustrate the method using simulated data of continuous phenotypes. We show that our kernel-machine approach typically outperforms the traditional joint test under strong GxE models and further outperforms the traditional main-effect association test under less strict models of weak or no GxE effects. We also illustrate our test using genome-wide association data from the Grady Trauma Project.

You may contact the first author (during and after the meeting) at