Regression-based multi-marker analysis for genome-wide association studies using haplotype similarity

Although haplotype analyses have been prevalent in genetic association studies, the main statistical analyses in genome-wide association (GWA) studies still mostly focus on singleSNP analyses. Several practical issues hinder practitioners’ enthusiasm to perform haplotypebased analysis under a GWA setting, including the need for large degrees of freedom and the treatment of missing phase information. To avoid these pitfalls, we propose a similarity-based regression model. It captures genetic variants via haplotype similarity to reduce the degrees of freedom, and uses phase-independent similarity measures to bypass the needs to impute phase information. We construct the score test to detect association between trait similarity and genetic similarity, and identify the limiting distribution of the score statistic. We show that the gene-trait similarity regression is closely connected with the random effects haplotype analysis although commonly they are considered as separate modeling tools in haplotype analysis. The proposed method is computationally efficient, allows for covariates and is applicable to both quantitative and binary traits. It serves as an effective tool for multi-marker analysis in genomewide association studies.