Statistical analysis strategies for association studies involving rare variants

Abstract

We review the motivation for exploring the role of rare variants in phenotypic expression. There are several problems with capturing the effects of rare variants in association studies using current statistical analysis methods. We discuss the concept and use of collapsing sets of rare variants into predictors of phenotypic expression, to aid statistical analyses of rare variant associations. Functional annotations of specific variants and genomic regions can be used to define collapsed sets of rare variants. A range of statistical analysis models and inference-making procedures could be exploited to assess the association between rare variants and phenotypic expression. We discuss the relative merits of these approaches. We compare moving window and defined region approaches to the analysis of rare variant effects. We discuss the importance for rare variant analysis of the flexibility of statistical analysis models and methods in accommodating factors, including common variants, interactions between variants, beneficial and deleterious effects of variants and environmental factors.