Genetic variants data sets were collected from dbSNP137 (28) and 1000 Genomes Project phase 1 release version 2 (29), which comprise 52 054 804 and 26 152 995 SNPs and Indels, respectively. We assigned reference allele and all alternative alleles to dbSNP137 variants and used biallelic variants for 1000 Genomes Project. The allele information was used to calculate the binding affinity of TFs. Linkage disequilibrium (LD) data for 11 populations were retrieved from the merged data of HapMap phases I + II + III. LD data for four 1000 Genomes Project super populations were computed and retrieved from MACH (30). Genomic coordinate of each locus was converted to GRCh37 hg19 by UCSC liftover tool. Variants with dbSNP ID were mapped to dbSNP137 using dbSNP merge file. In addition, annotations for genes and other DNA elements were downloaded from the UCSC Genome Browser. Furthermore, GWAS3D used position frequency matrices of 73 transcription factors motifs grouped by family (ENCODE motifs) from ENCODE web site, which provides a comprehensive resource of 245 known motifs curated from TRANSFAC, Jaspar and protein-binding microarray experiments, and
