Comparison of the distributions of substitution types among cases of binding site loss and those of benign effect.

The counts of the different substitution types are displayed for the inferred cases of binding site losses and benign effect (‘benign’ are defined as cases where both the original motif and the substituted variant share at least one condition in which both display significant EC scores). The distributions of counts differ significantly among the two types of substitution outcomes (chi-square test; chi-square deviation = 82.89; 11 degrees of freedom; p-value = 4.07e-13). For each substitution type we also tested whether it is distributed differently among the two categories of substitution outcomes, using a chi-square test with one degree of freedom (for the purpose of this test all other substitution types were collapsed into one category of substitution types). Statistically significant p-values, using a FDR [23] of 0.1, are shown in bold. The table also specifies whether the substitution type is found more/less than expected among the cases of binding site loss. Using the direction and size of the deviation from expectation the substitution types are sorted in ascending order of predicted severity.
aFor cases where the substitution results in a benign outcome, i.e. both original motif and the substituted variant are within the filtered motif dataset, the identity of the original motif and the mutant was decided according to the order of traversal of the motif dataset (in order to avoid double-counting of substitutions). Therefore, in order to avoid the dependence of counts on the order of processing of motifs, counts for complementary substitution types (e.g. C->G and G->C) were averaged.