Large-scale initiatives such as The Encyclopedia of DNA Elements (ENCODE) [18] have ascribed functional importance to more than 80% of the human genome and have provided a genome-wide catalogue of regulatory regions. This functional annotation data can be used jointly with the standard association signal to gain insights into the genetic basis of common traits. Indeed, variants associated with certain ENCODE genomic functional annotations such as DNase I Hypersensitive Sites, transcription factor binding sites and expression quantitative loci are enriched among GWAS hits [19], [20], [21], [22], [23], with recent work demonstrating that it is possible to integrate such data with the GWAS association signal to identify novel risk loci [10]. However, existing integrative frameworks typically either assume a single causal variant per risk locus [10] that is likely to be incorrect at many risk loci [10], [24], [2], [7], [25], [26], [27], [28] or do not make use of functional data [29], [30]. Although ENCODE functional annotation data are clearly beneficial for fine-mapping [22], a rigorous statistical framework for integrating the different types of information for the purpose of prioritizing plausible causal variants is currently lacking.
