Finding G-E Interactions in Quantitative Trait Analysis Using Two-Step Methods

Abstract

In recent years, screening approaches known as two-step methods have been proposed to detect gene-environment interactions for genome-wide association studies (GWAS). Genetic and environmental factors are believed to affect disease outcome as well as various quantitative traits such as height and blood pressure. The performance of the two-step methods has not been demonstrated in the quantitative trait setting. This thesis examines the method proposed by Wang and Abbott (2008) for generating genotyped markers in linkage disequilibrium (LD) and takes this approach in simulating data pertaining to a quantitative trait. The simulation results demonstrate that the two-step methods maintain type I error and have power to detect the quantitative trait locus. In this setting, the EG method (Murcray et al., 2009) is influenced by the strength and structure of the gene-environment dependency, the sample type, and the disease model. As such, the power of the EG method can fluctuate depending on the type of data while the DG method (Kooperberg and LeBlanc, 2008) remains fairly robust across a wide range of scenarios. The performance of the combined two-step approaches (EDGE (Gauderman et al., 2013) and H2 (Murcray et al., 2011) methods) tends to favour the more powerful underlying method. The power of the EDGE method can be improved if DG and EG demonstrates similar power while the H2 method can be made more powerful by choosing the appropriate parameters.