For each modENCODE and ENCODE TF, a single combination of motif and cell type was chosen based on appreciable enrichments at TF-bound versus unbound regions, the total numbers of TF-bound motifs and a correlation between per-position evolutionary conservation and information content. Motif PWM score thresholds for human TFs were determined using TFM_PVALUE (P = 4e-8) [70], consistent with the thresholds used in ENCODE integrative analyses [23]. For Drosophila TFs, thresholds were defined based on balancing the number of detected instances and motif enrichment at bound compared to unbound regions. Near-identical PWMs were removed based on Pearson correlation analyzed with STAMP [71,72]. See Supplementary note on TF selection in Additional file 2 for more detail. The properties of selected motifs are listed in Tables S1 and S2 in Additional file 2. PWMs are listed in the data/motifs.txt files at [60] and [61], respectively. The positions, sequences, PWM scores and variation properties of all TFBSs included in this study are listed in Additional file 3 (Drosophila) and Additional file 4 (human).
