Test results on different sets of genomes.

As benchmark data, we used five sets of genome sequences from AF-Project [47] and sets of genomes from Wolbachia [48], Piroplasmida [49] and Termites [50]. For each data set, we generated up to 1,000,000 pairs of quartet blocks as described in the main text. The table shows the number of informative block pairs (‘# inf bp’), i.e. the number of block pairs for which we obtained either strong or weak support for one of the three possible quartet topologies of the involved sequences. In addition we show the percentage of correct quartet topologies (with respect to the respective reference tree), out of all informative block pairs, as well as the ‘coverage’ by quartet blocks, i.e. the percentage of sequence quartets for which we found at least one informative block pair. Standard deviations are shown in parentheses. For each data set, we obtained 1,000,000 block pairs, except for the three sets of mitochondrial genomes, where it was not possible to find this number of block pairs.