Abstract
Genome-wide association studies (GWAS) and next generation sequencing studies (NGSS) are often performed in family studies to improve power in identifying genetic variants that are associated with clinical phenotypes. Efficient analysis of genome-wide studies with familial data is challenging due to the difficulty in modeling shared but unmeasured genetic and/or environmental factors that cause dependencies among family members. Existing genetic association testing procedures for family studies largely rely on generalized estimating equations (GEE) or linear mixed-effects (LME) models. These procedures may fail to properly control for type I errors when the imposed model assumptions fail. In this article, we propose the Sequence Robust Association Test (SRAT), a fully rank-based, flexible approach that tests for association between a set of genetic variants and an outcome, while accounting for within-family correlation and adjusting for covariates. Comparing to existing methods, SRAT has the advantages of allowing for unknown correlation structures and weaker assumptions about the outcome distribution. We provide theoretical justifications for SRAT and show that SRAT includes the well-known Wilcoxon rank sum test as a special case. Extensive simulation studies suggest that SRAT provides better protection against type I error rate inflation, and could be much more powerful for settings with skewed outcome distribution than existing methods. For illustration, we also apply SRAT to the familial data from the Framingham Heart Study and Offspring Study to examine the association between an inflammatory marker and a few sets of genetic variants.