This project seeks to develop new statistical tools and to apply existing ones in evaluating gene-environment interactions and genetic susceptibility. Work proceeded in two areas: (1) improving epidemiologic study designs, and (2) devising a statistical modeling approach to enhance our ability to detect genotype-exposure interactions. Geneticists have proposed using cases and their parents (case-parent triads) to study genetic effects on disease risk while overcoming the problem of population stratification or admixture. If a population consists of a number of distinct subpopulations that differ in baseline disease rates and in the frequency of a genetic variant, case-control studies may find associations between the variant and disease that lack etiologic import that arise simply because of the differing characteristics of the subpopulations. Case-parent triad designs can eliminate such potentially misleading associations. We are examining designs that augment case-parent triads with control-parent triads. Such designs should extend protection against admixture to studies that examine genotype and exposure jointly. Genotype-environment interactions become difficult to detect when genes have more than two alleles, in part because describing interactions requires many parameters. We have proposed a way to parameterize interactions parsimoniously that allows investigators to detect certain interactions that traditional methods could miss.
Showing the most recent 10 out of 25 publications
