Kernel Approaches for Detecting Interaction Effects in Complex Human Traits

• Activity Number: 277
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Michael Epstein*+
• Companies: Emory University
• Keywords: genomewide association study ; kernel ; single-nucleotidy polymorphism ; interaction

Abstract

Gene-gene and gene-environment interactions likely have a substantial role in the genetic origins of complex human traits. To facilitate investigation of these topics, we have developed mixed models to test interactions by incorporating appropriate genotype and environmental predictors in a covariance matrix based on an appropriate kernel function, which models both main and high-order interaction effects of the predictors within a small parameter space. Using a modified kernel function, we can test for the effect of a genetic variant (or multiple genetic variants in a region) in the presence of potential interactions with other genetic or environmental factors. Compared with standard approaches for interaction testing, our proposed method allows more flexible modeling of the genetic effect as a function of environment. Compared with existing multivariate interaction tests using haplotypes or genotypes, our method has fewer degrees of freedom, which should yield a more powerful test. We illustrate the approach using both simulated data as well as real data from existing genetic studies of atherosclerosis and coronary artery calcification.