We performed systematic fine-mapping using three commonly used tools and observed highly concordant results in identifying credible set variants in each causal block (Supplementary Figure S3). Among all identified causal blocks in the CAUSALdb, only five of them show relatively low correlation between FINEMAP and other two tools (Spearman's Rank correlation coefficient <0.8, Supplementary Figure S3A and Table S4), which may due to the Shotgun Stochastic Search algorithm used in FINEMAP. Also, the credible set size of each causal block is largely similar across fine-mapping tools, especially between CAVIARBF and FINEMAP, probably due to they used similar statistical models (Supplementary Figure S3B). By pooling fine-mapped variants in the 95% credible set (α = 0.95), we built a dataset composed of 962 176 potential causal variants corresponding to 5 097 732 genotype–phenotype associations across the entire human genome (Supplementary Figure S4). The genomic distribution of these variants showed that 98.9% of them were located in non-coding genomic regions (Supplementary Figure S5), emphasizing the pivotal role of regulatory variants in the development of complex traits. Among these variants, ∼55% were identified by
