Fig. 3: Genetic architecture of enrichment for height (HT), body mass index (BMI), cardiovascular disease (CAD) and type-2 diabetes (T2D) for 382,466 unrelated European ancestry UK Biobank individuals genotyped at 8,430,446 SNP markers.

a We partition SNP markers into seven location annotations (coding regions, intronic regions, and windows 1, 1–10, 10–500 kb and 500 kb–1 Mb upstream of genes, with other SNPs grouped in a category labelled “others"). Windows 1–10 kb, 10–500 kb and 500 kb–1 Mb upstream of genes are further split into SNPs mapped to enhancers (enh), transcription factor binding sites (tfbs) and others. Within each of the 13 annotations, we have three minor allele frequency groups (MAF ≤ 0.01 annotated as rare, 0.01 < MAF ≤ 0.05 annotated as low, and MAF > 0.05 annotated as common), and then each MAF group is further split into two based on median LD score. This gives 78 groups for which our BayesRR-RC model jointly estimates the phenotypic variation attributable to, and the SNP marker effects within, each group. For each of the 78 groups, SNPs were modelled using five mixture groups with variance equal to the phenotypic variance attributable to the group multiplied by constants (mixture 0 = 0, mixture 1 = 0.0001, 2 = 0.001, 3 = 0.01, 4 = 0.1). b Posterior distribution of the proportion of the total phenotypic variance attributable to the SNP markers that is contributed by each of the four non-zero mixtures within each MAF-annotation group for HT, BMI, CAD and T2D. Within these, are boxplots of the posterior mean and 95% credible intervals. Values are summed over LD groups. c Bar plots with error bars giving the 95% credible intervals for the average effect size of markers in the model for each MAF-annotation group, split by mixture.