Distinguishing a true causal variant from an extremely high LD is a challenging task, and fine-mapping usually generates a challenging credible set in which highly linked variants achieve a similar PP of causality. For example, at locus 15p22.3 in the CAD GWAS mentioned before, there were five variants in the credible set, with similar PP of ∼0.2 (Supplementary Figure S9C). CAUSALdb provides base-wise variant annotations and tissue/cell type-specific epigenome data that help researchers determine true causal variants. By inspecting the functional annotations from CAUSALdb, we found that although rs17293632 did not obtain the highest PP, it could be a causal CAD variant at 15p22.3, with substantial supporting evidence. First, according to our aggregated conservation scores, rs17293632 was noted to be more conserved than the other four variants (Supplementary Table S7). Second, this variant obtained significantly higher functional scores than the other variants according to our integrated non-coding variant functional prediction tool, namely regBase (34) (Supplementary Figure S9D). Third, rs17293632 was found to overlap with most epigenomics signals such as open chromatin, histone modification, and transcription factor binding (Supplementary Figure S9E),
