Statistical advances in gene by gene interaction and gene by environment interaction in the era of genome-wide association studies

Genome-wide association scans (GWAS) and meta-analyses within consortia have been used to detect novel genetic variants associated with disease. While many variants have been found, methods to detect genetic variants in the presence of gene by gene (G by G) and gene by environment (G by E) interaction are needed to identify additional loci. This thesis is comprised of projects related to the statistical analysis of G by G interactions and G by E interactions in the context of GWAS and consortia formed to analyze disease or quantitative outcomes. First, for the detection of interacting genetic loci, we compare a screening approach based on biological knowledge to one where the screening is by marginal association effects. Next, we describe the joint meta-analysis (JMA) approach, a novel application of multivariate meta-analysis methods involving simultaneous meta-analysis of both the gene and G by E interaction effects. We show that the JMA has equal or greater power to other methods in comparative simulation studies. Finally, we explore two different sampling designs for the meta-analysis of G by E interaction effects: one involving combining case-control samples with case-only samples for dichotomous outcomes and the other that includes additional samples measured on one level of a dichotomous environment variable.;The methods presented in this thesis are applied to three data sets. They are (1) a G by G interaction study of rheumatoid arthritis on the North American Rheumatoid Arthritis Consortium data set, (2) GWAS of fasting glucose with interaction by body mass index (BMI) in a meta-analysis of five cohorts and (3) a meta-analysis of eight cohorts of the interaction effects of BMI with the ENPP1 gene in studies of type 2 diabetes.;These projects are timely and relevant; researchers are joining consortia to conduct meta-analyses and are looking beyond simple regression models and desire methods for improving their ability to detect genetic loci in the presence of interaction.