To quantitatively measure the difference on the binding affinity caused by different alleles of candidate variants with GWAS3D signals and its significance, we used a comprehensive TF motif set to evaluate the possible reduced or enhanced binding. We first computed the position weight matrices (PWMs) from position frequency matrices of all ENCODE motifs by converting normalized frequency value to log-scale value using the method described previously (33,34). Given a variant (V) with GWAS3D signal, we took 30 bp of surrounding sequence and constructed the mutated sequences between the reference alleles (Ar) and the alternative alleles (Aa1, … Aan). For user-selected motifs of TFs, we scanned these sequences using PWMSCAN (35) and fetched P-values represent the significance of each putative TF-binding site. We set a PWMSCAN P-value threshold (1E-3) to reduce the number of false positive bindings. We then measured the score of binding affinity change by calculating the log-odds (LOD) of probabilities of paired binding sites for each motif (m): To estimate the statistical significance of binding affinity change, we performed permutations of on 52 054 804 SNPs in dbSNP137
