All SNVs with P < 1 × 10−5 from the discovery meta-analysis in European individuals were considered for follow-up. As a QC step after meta-analysis, we excluded four SNVs with unrealistically high beta values, large standard errors and results that were reported in less than four studies. We defined a novel locus as a genomic region (i) with SNVs not in LD (r2 < 0.2) with any well-established HR-associated SNVs from the previously reported GWAS (12) (Supplementary Material, Table S1), and (ii) mapping to more than 500 kb from either side of a previously reported HR-associated SNV. At the time of our study, there were 21 loci reported from GWAS analyses with HR-associated SNVs (12). A potential secondary signal within a previously reported locus was defined as being within a 1 Mb region centred around the published SNV, but not in LD (r2 < 0.2) with the published SNV in that region. LocusZoom plots were produced for all selected SNVs. Only the lead SNV was carried forward, for each signal being followed up. Specifically, the most significantly associated SNV was
