As well as looking across the entire PWM, we can consider each motif position in turn. Consistent with previous findings for other TFs in yeast [35] and Drosophila [36], cross-species variation at Twi, Bin and Tin motif positions strongly anti-correlated with their information content (Figure 1a; Figure S2 in Additional file 1). Variation across individuals also anti-correlated with positional information content (Figure 1d), confirming the general link between evolutionary conservation and population diversity [37]. There are, however, some interesting exceptions. For example, positions 6, 7 and 12 of the Twi motif are less varied in the population than would be expected from their information content (Figure 1d, left panel). These positions correspond to the 'spacer' region of the CANNTG E-box consensus motif recruiting basic helix-loop-helix (bHLH) proteins, for which specific sequence preferences were documented depending on specific dimerization partners [38]. Similarly, we found the first two positions of the Bin motif to be highly constrained despite their very low information content (Figure 1d, middle panel), suggesting that these positions may also be subject to specific restrictions depending on the cis-regulatory
