In order to illustrate the utility of LD Hub, we conduct an analysis using summary results data from a large GWAS of atopic dermatitis (AD) for 40,835 (10,788 cases and 30,047 controls, sample prevalence: 0.264) individuals of European ancestry (i.e. the whole discovery set except 23andMe results) (Paternoster et al, 2015 (link)). In total, 11,059,640 SNPs were included in this meta-analysis. Since AD is influenced by a gene of major effect, i.e. filaggrin- variants in this region have allelic odds ratios > 7 (Sandilands et al, 2007 (link)), which could bias estimates from LD Hub, we excluded this region from the uploaded results file. For traits/diseases that have a single locus of disproportionately large effect (i.e. χ2 > 80) compared to the rest of the genome, we recommend the exclusion of SNPs in these regions as good practice when using LD Hub (and LD score regression in general), since the inclusion of these SNPs could unduly leverage the regressions and consequently the estimates of genetic correlations and SNP heritability. However, with the exception of autoimmune diseases (SNPs in the MHC can have large effects on certain autoimmune diseases), it is unusual for common traits/diseases to exhibit a single locus of large effect, and thus this potential source of bias should not be an issue for a majority of diseases/traits. For traits that exhibit a single locus of disproportionately large effect (χ2>80), we recommend fine-mapping and direct evaluation of overlap in the particular region to assess whether genetic effects are shared, and LD score regression of the rest of the genome with this particular region excluded from analyses. After the abovementioned quality control steps, 1,215,002 SNPs were selected for upload.