Most of the analyzed TFBSs have no detected variation, in particular in human, and therefore a zero load. This affects the statistical power, making it challenging to examine many TFs one-by-one. However, analyzing the data globally for all included TFs in each organism has allowed us to identify a number of significant trends, as presented below. Technically, the high proportion of sites with no detected variation also leads to a considerable zero-inflation of TFBS load distributions, which violates the assumptions of conventional significance tests. Therefore, instead we estimate significance by using permutation tests, as further described in Materials and methods. For the same reason, we also chose to present average (more precisely, trimmed mean) TFBS load values in many comparative analyses as a metric that reflects both the frequency of variation (that is, zero versus non-zero load) and the intensity of its effect (that is, the distribution of non-zero load).
