The Performance of Conditional Tests for Family Data in Associated Regions Derived from GWAS

Abstract

Genome-wide association studies (GWAS) have been used successfully to identify genetic loci associated with complex diseases and phenotypes. Often this association takes the form of several significant signals (such as small p-values) in a univariate analysis at various markers within a single genetic region. Once confirmed, these associations lead to the question if a single marker tags the association signal of another, functionally relevant variant or if the single marker tags a functionally relevant haplotype. To deal with this question, methods for family data based on logistic regression, adaptations of the transmission/disequilibrium test (TDT) or weighted haplotype likelihood (WHL) methods have been proposed in the literature. Objectives were to examine the effect of parameters such as sample size, inheritance model, and the effects of linkage disequilibrium (LD) in the region on the ability of a selection of methods to detect an independent effect from an additional locus. All methods tested were applied to simulated genetic data of trios comprising a single affected offspring and two parents. While regression-based methods have advantages such as model flexibility, potentially increasing power, the WHL method was more robust against increasing LD in the scenarios analyzed. Simulation results suggest that the regression and WHL methods are better able with regard to statistical power than the adaptation of the TDT analyzed here to detect genetic effects at an additional locus while controlling for confounding at another locus.