Presenter: Boxi Lin

Supervisory Committee: Lei Sun (Supervisor), Lisa Strug, and Shelley Bull

Date and Time: Tuesday, December 20, 2022, 10am-12noon EST

Location: 155 College Street, Health Science Building, Room 790

Abstract: The genetic architecture of a complex trait may differ among subpopulations of a studied population. In genetic association studies, stratified and joint analysis strategies have been applied to incorporate heterogeneity. However, there is no uniformly most powerful test, and their performance will depend on the underlying genetic model. We aim to develop robust association testing methods that remain powerful under various patterns of heterogeneity by leveraging heterogeneity through the lens of interaction. We first propose second-order meta-analysis combing evidence from each subpopulation by taking the sum of Z-scores. We show the analytical connection between this meta-analysis is equivalent to mega-analysis approach based on weighted regression. Then we will conduct large-scale real data application to discover genetic association hidden in previous GWAS with stratified design or ignoring the potential heterogeneity and heteroskedasticity. Apart from identifying genetic effect in any of the groups, we also aim to develop meta-analysis for detecting the presence of genetic effect heterogeneity between groups. Lastly, we plan to apply the joint location-scale (JLS) testing framework detecting genetic association under heterogeneity due to unknown environmental factors.

All are welcome!