Abstract:
|
With the advance in next-generation sequencing technology, massive genetic and genomic data have been produced. Statistical methods for testing genetic association with rare genetic variants have been well established and widely applied to unrelated samples. These methods are also known as gene-based or variant set association tests, since rare variants are often grouped by genes, functional units or genomic regions in the analysis. In recent years, large-scale sequencing projects have included correlated study samples from family studies, or with cryptic relatedness, and there is a pressing need of developing efficient statistical methods to analyze these data. Here we propose and implement SMMAT, a powerful and efficient variant set association test for correlated study samples in the generalized linear mixed model framework. SMMAT is a hybrid test that aggregates association evidence from the burden test and the sequence kernel association test. We show in simulation studies that SMMAT controls correct type I error rates and maintains good power in both single-cohort studies and meta-analysis. We also illustrate SMMAT in a real data example from a large-scale sequencing study.
|