Integrative analysis of functional genomic annotations and sequencing data to identify rare causal variants via hierarchical modeling

Identifying the small number of rare causal variants contributing to disease has been a major focus of investigation in recent years, but represents a formidable statistical challenge due to the rare frequencies with which these variants are observed. In this commentary we draw attention to a formal statistical framework, namely hierarchical modeling, to combine functional genomic annotations with sequencing data with the objective of enhancing our ability to identify rare causal variants. Using simulations we show that in all configurations studied, the hierarchical modeling approach has superior discriminatory ability compared to a recently proposed aggregate measure of deleteriousness, the Combined Annotation-Dependent Depletion (CADD) score, supporting our premise that aggregate functional genomic measures can more accurately identify causal variants when used in conjunction with sequencing data through a hierarchical modeling approach.