We analyzed data from the UK Biobank consisting of 152,249 samples genotyped on ~800,000 SNPs and imputed to ~73 million SNPs.
We selected 15 highly heritable phenotypes with large sample size.
For each phenotype, we computed mixed model association statistics using version 2.2 of BOLT-LMM software (Loh et al. Nat Genet 2015) with genotyping array (UK BiLEVE / UK Biobank) and assessment center as covariates. We included 607,518 directly genotyped SNPs in the mixed model (specifically, all autosomal biallelic SNPs with missingness < 2 % and consistent allele frequencies between the UK BiLEVE array and the UK Biobank arrays). Association statistics were computed from dosage data on imputed SNPs in HapMap3 (1,211,182 SNPs).
In this study we determined that SNPs with low level of LD (LLD) have significantly larger per-SNP heritability. Roughly half of the LLD signal can be explained by functional annotations that are negatively correlated with LLD, such as DNase I hypersensitivity sites (DHS) and histone marks. The remaining signal is driven by annotations related on negative selection, highlighting the effect of negative on all polygenic complex traits.
Steven Gazal, Hilary K. Finucane, Po-Ru Loh, Pier Francesco Palamara, Xuanyao Liu, Armin Schoech, Brendan Bulik-Sullivan, Benjamin M Neale, Alexander Gusev, Alkes L. Price1. Linkage disequilibrium dependent architecture of human complex traits reveals action of negative selection Nature Genetics volume 49, pages 1421 1427 (2017) doi:10.1038/ng.3954