PWMs are an established way of representing the sequence preferences of TFBSs, with PWM match scores reflecting the similarity of a given sequence to the hypothetical 'ideal' binding site for a given TF [33]. To study the relationship between PWM scores and variation, we compared the variation properties of Twi, Bin and Tin motifs at three score ranges ('strong', 'medium' and 'weak' scoring). Weaker (that is, potentially 'less optimal') motifs generally showed higher levels of individual variation (Figure 1c), as further confirmed using only the strongest scoring sites from each bound region to reduce the contribution of non-functional motif matches (Figure S1B in Additional file 1). This result is consistent with the expectation that selection would predominantly work towards increasing TFBSs' match to consensus [34]. We revisit this question more formally later in the study.
