A Haplotype-Similarity Based Approach for Detecting Genetic Association

Abstract

One way toward identifying the mutations that increase the liability of contracting a disease involves population-based association analysis, i.e., case-control studies. To detect the association between the genetic variants and the disease, we use the fact that disease mutations can be identified by the excess of haplotype sharing among cases, and introduce a class of one- degree-of-freedom statistics that are quadratic functions of the haplotype frequencies to measure haplotype similarity. We then present a testing procedure built upon the class of quadratic statistics for initial genome screening. Two main issues are involved in constructing this haplotype-based approach: (1) population substructure and correlated samples, and (2) unphased genotypic data. We tackle the first problem by generalizing the Genomic Control principle to haplotype setting. We show that this procedure extends naturally to genotypic data and can be robust to phase uncertainty. Finally via simulation, we evaluate the performance of the similarity-based approach through power comparisons with the commonly used alternative, the Pearson's Chi-square test. Our results show that each method is powerful under a different background of mutations, depending on whether the mutations occurred on rare or on common variants.