The NHGRI-EBI GWAS catalogue was downloaded on 2016-04-18 (release 2016-04-10). Entries with missing positional or P-value information were removed and the positions of the remaining entries were converted to hg19 with the UCSC liftOver function in the “rtracklayer” R/Bioconductor package 77 and the ‘hg38ToHg19’ chain file. This resulted in 24,861 catalogue entries, corresponding to 18,446 unique variants of which 6,681 were significantly associated to a trait (P < 5x10-08). However, many studies included in the complete catalogue are not reliable, so we parsed the sample size information in the catalogue and excluded studies that did not report effect sizes (odds ratio or regression coefficient), had a sample size below 1,000, or assayed fewer than 100,000 variants. Then, following the approach taken in 78, we further filtered the set of remaining associations to retain only traits that had at least six significantly associated variants, and kept all associations with P < 1x10-06 for these traits. This approach yielded a filtered set of 9,562 associations for 6,059 variants with 358 diseases and traits. Each variant position was then parsed for trait overlap
