For each risk locus, we selected the target genes based on significant associations (P < 0.05). Then for each of the target genes, we selected 1-50kb regions on both sides of the transcription start site as putative enhancer regions. We next retrieved all the sites overlapping the DNaseI hypersensitivity (DHS) peaks in the ENCODE MCF-7 DNaseI-seq data generated by the Stamatoyannopoulos group (Birney et al., 2007) (http://hgdownload.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeUwDnase/). Thus for each risk locus, we obtained a set of DNA sequences from the putative enhancer regions of the target genes. To control for directional regulatory effects, we considered up-regulated and down-regulated target genes separately. Then we identified the binding motifs of known transcription factors (TFs) from databases embedded in cistrome, including Transfac, JASPAR, UniPROBE (pbm) and hPDI (Liu et al., 2011) which are significantly overrepresented in these sites (P <1×10-6). We further mandated that the TF whose motif was enriched in the target genes to be located within 1 MB range of the corresponding risk locus. If a transcription factor satisfies these criteria, we considered it as a candidate for empirical 3C validation.
