Kernel Machine--Based Testing for Gene-Gene Interactions in Genetic Association Studies

• Activity Number: 248
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Jennifer Clark*+ and Michael C Wu and Arnab Maity
• Companies: and The University of North Carolina at Chapel Hill and North Carolina State University
• Keywords: Epistasis ; kernel machine regression

Abstract

Current analysis strategies for genome-wide association studies have thus far failed to unearth new variants explaining a large proportion of genetic heritability. Across all the studies in the literature there are few replicable SNPs with odds ratios going above 1.5 for common human diseases. The lack of success in GWAS studies thus far has to do with the limitations of only testing for single variant associations which places considerable facile assumptions on the disease complexity. It is thought that epistasis, or gene-gene interaction, plays a key role in the genetic architecture of common diseases and can explain a large proportion of the missing heritability. We present a new method which utilizes additive least squares kernel machine regression to construct a score test to test for epistatic interactions across multiple gene regions, allowing for complex, high-order interactions. Testing at the gene level allows for capture of un-genotyped variants and simultaneously reduces multiple comparisons. Simulations and real data analyses demonstrate that our approach correctly controls type I error and offers improved power over competing approaches.