Next-Generation Sequencing Studies: Optimal Design and Analysis, Missing Heritability and Rare Variants

Abstract

As we enter the ``Post Genome-wide Association Study'' era, there is a widespread desire to move towards sequencing study samples, often after an initial genome-wide association analysis that has found some loci associated with a phenotype of interest in a given population. The rationale is clear: the initial study is highly unlikely to have found all, or even most, associated loci. Indeed, in terms of heritable variation, for virtually every phenotype, much remains to be explained. The details of how such a sequence-based study should be designed are unclear. In this article we review the rationale for such a sequence-based approach, such as the desire to find new (likely rare) variants that are also associated with the phenotype, and the wish to explain the ``missing heritability''. We go on to review simulation-based and data-based ways of developing guidelines for the design of such next-generation sequence-based studies. Finally, we discuss the desirability of using biological prior information to help lessen the problems caused by ``the curse of dimensionality'' in this context.