Correction for Sampling Structure Using Generalized Linear Mixed Models for Discrete and Continuous Phenotypes in Genome-Wide Association Studies

• Activity Number: 84
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Han Chen*+ and Chaolong Wang and Xihong Lin
• Companies: Harvard School of Public Health and Harvard School of Public Health and Harvard School of Public Health
• Keywords: generalized linear mixed models ; genome-wide association studies ; population stratification ; cryptic relatedness ; score test

Abstract

In genome-wide association studies (GWAS), population stratification and cryptic relatedness can result in identifying spurious association signals. Principal component analysis (PCA) has been widely used to adjust for population structure. However, PCA does not work well in the presence of cryptic relatedness in addition to population stratification. Linear mixed models (LMMs) have been proposed as an alternative method for normally distributed continuous phenotypes. However, they are not applicable for discrete traits, such as binary phenotypes in case-control studies. We propose in this paper to use generalized linear mixed models (GLMMs) to control for population stratification and cryptic relatedness in GWAS for both discrete and continuous phenotypes. We develop a computationally efficient score test for association analysis by fitting GLMMs. We show in our simulation studies that our method performs well in controlling type I errors in the presence of population stratification and cryptic relatedness, and illustrate our method in a real data example.