We selected the HR-associated SNVs and proxies in LD (r2 ≥ 0.8; calculated using the UK Biobank full genetic dataset) that were identified in this study, and from the previous GWAS (16) and UK Biobank studies (17) for annotation. To identify the potential regulatory variants, we retrieve the functional confidence score for SNVs from the RegulomeDb database (21). RegulomeDb assigns a functional confidence score to each SNV by overlapping them with functional genomic data mainly from ENCODE (e.g. DNase I hypersensitivity, DNase I footprinting, ChIP-seq), with eQTL data and with computational prediction (e.g. TF-binding sites and their disruption). We considered any SNP with at least one functional annotation to have regulatory potential (this corresponds to functional confidence scores: 1a-6).
