Earlier in the study we have shown evidence that 'weaker' motifs (that is, those with a poorer PWM match) are more prone to variation, suggesting that they are less functionally constrained. Weaker sites have many more possible variants with similar match scores, while mutations at stronger sites are less likely to preserve their match. Motif load is based on the decrease in PWM score associated with mutations and not sequence variation per se and is therefore more 'protected' from this bias. Using this metric, we confirmed our original findings, suggesting that TFBSs with higher PWM scores are generally more functionally constrained compared to 'weaker' sites (Figure 3c). The fraction of detected sites mapping to bound regions remained similar across the whole analyzed score range, suggesting that this relationship is unlikely to be an artifact of higher false-positive rates at 'weaker' sites (Figure S4A in Additional file 1). This global observation, however, does not rule out the possibility that a weaker match at some sites is specifically preserved to ensure dose-specific TF binding. This may be the case, for example, for
