We aim to analyze TFBS functional constraints using the binding data generated by the ENCODE, modENCODE and published sources. Prior to these global analyses, however, we first examined the relationship between binding sites' match to consensus, their conservation and variation using three well-characterized Drosophila TFs, Twist (Twi), Biniou (Bin) and Tinman (Tin), which have large numbers of TFBSs whose general occupancy is predictive of specific spatio-temporal activity [2]. The discovered PWMs for these TFs from both in vitro and in vivo studies are good predictors for their binding [2] and their binding sites show an appreciable level of variation, presumably much of which is deleterious but not lethal. For these TFs, 24 to 28% of the bound sites overlapped with SNPs identified by the DGRP [22] in 162 isogenic lines of Drosophila melanogaster (hereafter we refer to variation across these lines as 'individual variation'). As expected, variation at the same sequences detected outside of TF-bound regions (that is, at potentially random motif matches) was even higher, with 35% of them containing known SNPs (Fisher test, P < 1e-50 compared to the bound sites).
