|
Abstract:
|
Population stratification is a major confounding factor in sequencing studies and can lead to spurious associations. It has recently been shown that rare variants tend to exhibit a stronger stratification than common variants, especially when the phenotype has a sharp spatial distribution. The resulting inflation in test statistics may not be corrected for by popular methods such as principal components (PCs) adjustment and linear mixed models (LMMs). In this talk, we propose two methods to address this problem: one based on a combination of LMM and PC adjustment, and another based on hidden Markov random field models. We show via simulations that both methods effectively control for stratification in spatially structured populations when the phenotype has a sharp spatial distribution.
|
