We then quantitatively evaluated the performance of our method. We first collected 118 known regulatory variants from OregAnno database (46). We randomly selected three data sets from dbSNP with same number of genetic variants in each of the regulatory regions (promoter, intergenic and genome-wide). For each of aformentioned four SNVs list, we performed GWAS3D pipeline without considering GWAS P-value, cell type restriction and population LD. Wilcoxon rank-sum test showed significant differences between OregAnno and each random sets, with P-values of 0.0344, 0.0011 and 0.0052 for promoter, intergenic and genome-wide data set, respectively, whereas there are no differences among the three random data sets. The experiment demonstrated that GWAS3D pipeline gives higher scores to regulatory variants and thus differentiates them from random variants.
