This project seeks to develop new statistical tools for evaluating gene- environment interactions and genetic susceptibility. Work has begun in two areas: (1) improved epidemiologic study designs that exploit the independence of genotype and environmental exposure, and (2) statistical modeling that parameterizes interactions of genotype and environmental factors parsimoniously to enhance a data analyst's ability to detect such interactions. Genotyping control subjects may not be feasible because they may be reluctant to contribute tissue or may have concerns about privacy with genetic information. Under the often-plausible assumption that genotype and exposure are independent in the study populations, we found that study designs where controls are not genotyped can use fewer subjects than earlier designs to achieve the same precision in estimating both exposure effects and genotype-exposure interaction effects. However, genotype effects cannot be assessed without external information about gene prevalence in the population. When studying genes with more than two alleles, genotype-environment interactions can be difficult to detect because describing them requires many parameters. We have proposed a way to parameterize interactions parsimoniously that will allow investigators to detect interaction effects of carcinogen metabolism genes that would be missed with traditional methods.
Showing the most recent 10 out of 25 publications
