Fig. 2: Genetic architecture for time-to-diagnosis of CAD, HBP, T2D and age-at-menarche and age-at-menopause, estimated using 2,975,268 markers and unrelated European ancestry individuals from the UK biobank.

BayesW models were executed with and without partitioning the markers into groups. Groups were created by splitting them first into MAF quintiles and each MAF quintile was further split into quartiles based on LD giving us a total of 20 groups. The used sample sizes were N = 360,715; 371,878; 372,280; 200,493 and 151,472 for time-to-diagnosis of CAD, HBP, T2D and age-at-menarche and age-at-menopause, respectively. The models without groups were run with five chains and the models with groups were run with three chains, each chain contained 10,000 iterations out of which 5000 first iterations were discarded as burn-in and a thinning step of five was applied to give 1000 samples from each chain. a Mean proportions of genetic variances explained by each of the mixtures with the groups model and without groups model, groups model and no group model are yielding rather similar results; (b) distribution of proportion of genetic variance between mixtures for the model without groups, time-to-menarche stands out with almost all of genetic the variance attributed to the small mixtures; (c) distribution of proportion of genetic variance between LD quartiles within each MAF quintile, LD bins do not have a large impact on genetic variance partitioning as the credibility intervals are large and medians across LD quartiles are rather stable; (d) distribution of proportion of genetic variance between mixtures within each MAF quintile, mixture allocations tend to be similar compared to no groups model; (e) enrichment (ratio of proportion of genetic variance and proportion of markers attributed to each MAF quintile group) value for each phenotype, enrichment of higher than 1 represents that the markers are explaining more of the genetic variance compared to their count proportion and vice versa. b–e Height of the bars represents median and error bars represent 95% credibility intervals, (c–e) are group models. For all of the traits, most of the genetic variance is coming from common SNPs (MAF quintile 5).