Robust analysis of related samples under the presence of population substructure

Abstract

We propose a new method for genome-wide association analysis with a family-based design. The proposed method is robust against population substructure while it is more efficient than the traditional method such as transmission disequilibrium test for related samples. The proposed method estimates the correlation matrix between individuals and then the principal component analysis is applied. To maximize the statistical power, we consider the additive polygenic model and a best linear unbiased predictor is used as offset. We confirmed that the proposed method is always efficient by simulation studies. The method will be applied to Framingham Heart study.