Inference of gene-environment interaction from heterogeneous case-parent trios

Population stratification is a major source of confounding in gene-by-environment (G × E) interaction studies of case-parent trios when the study sample consists of individuals with distinct ancestral backgrounds. This dissertation discusses different ways of controlling for population stratification in order to reduce the chance of false positive signals in G×E inferences. This work is organized in three parts. First, we investigate the impact of confounding on the results of a genome-wide association analysis by Beaty et al., which identified multiple single nucleotide polymorphisms that appeared to modify the effect of maternal smoking, alcohol consumption, or multivitamin supplementation on risk of cleft palate. The study sample of case-parent trios was primarily of European and East Asian ancestry, and the distribution of all three exposures differed by ancestral group. Such differences raise the possibility that confounders, rather than the exposures, are the risk modifiers and hence that the inference of G × E interaction may be spurious. Our analyses generally confirmed the result of Beaty et al. and suggest the interaction G × E is driven by the European trios, whereas the East Asian trios were less informative. Next, we show that current methods to reduce the bias in estimated G × E interactions from case-parent trio data can only account for simple population structure involving two strata and propose methods to overcome this limitation. Through simulations, we show that our proposed method maintains the nominal type-1 error rate and higher statistical power. The proposed approach was then applied to case-parent trios which consists of cleft-palate-affected children. Consistent with Beaty et al., our results suggest that the gene-environment interaction signal in these data is due to the self-reported European trios. Finally, we discuss methods to infer local ancestry of cleft-palate-affected children and propose methods to control for population stratification via local ancestry. Again, we apply our methods to the case-parent trio data of cleft-palate-affected children to investigate whether local ancestry rather than the genotype at a test locus modifies the association between disease and exposure.