Population stratification (PS) and misclassification (MC) of genotype may cause the usual tests in case-control studies producing spurious gene-disease associations. To control for the resulting bias, several powerful approaches have been proposed using unlinked markers. However, the effects of PS and MC of genotype and environmental exposure in studies of gene-environment interactions are still not clear. In my research project, I propose to investigate and control the effects of PS and MC in case-control and case-only studies under general conditions. My first plan is to identify some important conditions under which population stratification and/or misclassification do not introduce bias in case-control or case-only studies of gene-environment interactions. If the effects of PS and MC do exist, another interesting issue is the magnitude of the bias. I propose to study the bounds of the bias and use information related to bounds to control the effect of PS and MC. My final plan is to suggest a new study design so that PS and/or MC can be automatically accounted for. I shall also conduct some simulation studies, with the help from my students and using parameter values from real examples, to compare the power performance of different tests for gene-environment interactions under PS and MC. My preliminary results show that some interesting method do provide valid test for interactions. However, these methods are not powerful, especially when the genetic factors have strong main effects on the risk of disease.
