Figure 2: Example of analysis of a single protein, the 894AA long human zinc-finger PRDM9.

The 13 tandemly 28AA long repeats (ID) are identified by the algorithm at the end of the protein, ordered by their location on the protein. Underlined letters correspond to the Zinc fingers annotated in SwissProt (first finger starts 7AA before the beginning of the first repeat). The order by which the method accumulates the repeats (Order), reveals clusters of identical repeats: 1 (ID=3,6,7), 2 (ID=4,5) and 3 (ID=8,11). Black coloured repeats represent the seed identified in the second step, and red coloured repeats are those identified by the PPM-based predictor in the third step. Repeats are highly similar, with high IC, as shown by the sequence logo. A maximum-likelihood tree of the repeats is shown in the right panel, where the repeats IDs are given on the y axis. The plot beneath the tree shows the positive correlation between evolutionary distance (in substitutions per site) and physical distance (in amino acids), obtained by comparing all repeat pairs, where the red line represents a linear regression fit (Spearman correlation=0.53, P value=6.7e−7). The mean dN/dS for all pair comparisons (n=78): <dN/dS>=2.7146±0.23, that is, significant positive selection.