We cleaned and harmonized 963 publicly available GWAS summary-level datasets from 36 consortia, which included 82 diseases, 154 complex traits, 576 metabolites and 151 immune markers (Hemani
et al, in preparation).
From this database pool, we chose datasets that fit the following selection criteria:
Non-sex-stratified
Meta-analyses of predominantly European populations. We include a few GWAS meta-analyses that contain a small proportion of non-European individuals in them in the LD Hub database. Whilst we believe the effect of these small numbers of non-European individuals on the LD Score regression analyses will be relatively minor, users should be aware that results from these meta-analyses may be less robust because of inconsistent patterns of linkage disequilibrium between individuals of different ancestry. In order to flag these studies to the user, we have included an additional field in the Test Center and the GWAShare Center (last column) that indicates the population ancestry of individuals in the corresponding meta-analysis, as well as a similar field in the LD Score regression results file (see also Table S1).
Meta-analyses using a GWAS backbone chip only (i.e. exclude meta-analyses involving immuno | metabo | psych | exome chip or GWAS + custom chip)
Number of SNPs is large (N > 450 000)
Number of individuals is large (N > 5000)
Mean Chi-square of the test statistics is larger than 1
As shown in
Figure 2, after filtering on the selection criteria, genome-wide results for 173 traits were included in LD Hub, of which 18 are GWAS of diseases (Boraska
et al., 2014 (
link); Cross-Disorder Group of the Psychiatric Genomics Consortium, 2013 (
link); Lambert
et al., 2013 (
link); Liu
et al 2015 (
link); Moffatt
et al., 2007 (
link); Morris
et al., 2012 (
link); Neale
et al., 2010 (
link); Nikpay
et al., 2015 (
link); Okada
et al., 2013 (
link); Paternoster
et al., 2015 (
link); Ripke
et al., 2012 (
link); Ripke
et al., 2014 (
link); Simon-Sanchez
et al., 2009 (
link); Sklar
et al., 2011 (
link)), 48 are medically relevant risk factors/complex traits (Benyamin
et al., 2013 (
link); Berndt
et al., 2013 (
link); Bradfield
et al., 2012 (
link); Dastani
et al., 2012 (
link); de Moor
et al., 2010 (
link); Dupuis
et al., 2010 (
link); Estrada
et al., 2012 (
link); Furberg
et al., 2010 (
link); Horikoshi
et al., 2012 (
link); Huffman
et al., 2015 (
link); Lango Allen
et al., 2010 (
link); Manning
et al. 2012 (
link); Moffatt
et al., 2007 (
link); Pattaro
et al., 2016 ; Perry
et al., 2014 (
link); Rietveld
et al., 2014 (
link); Rietveld
et al., 2013 (
link); Saxena
et al., 2010 (
link); Shungin
et al., 2015 (
link); Soranzo
et al., 2010 (
link); Speliotes
et al., 2010 (
link); Taal
et al., 2012 (
link); Teslovich
et al., 2010 (
link); Teumer
et al., 2016 (
link); van den Berg
et al., 2014 (
link); van der Valk
et al., 2014 (
link)) and 107 are metabolites (Kettunen
et al., 2016 ). Table S1, displays descriptive information for each of the GWAS in LD Hub, including, trait name, consortium name, ethnicity, gender, sample size, PubMed ID, year of publication and other relevant information.
