Figure 7

Similarity of topologies for different tree algorithms. Average level of similarity for topology comparisons (when rewiring using network shuffle) made for a fixed number of shared species. The average similarity for each method is different, and the trends seen in PAML show marked difference to those methods taken from Phylip. The variance of similarities for each tree construction method increases as more species are compared. This may potentially bias the score results for the topological comparisons, thus suggesting the need to be vigilant when comparing graph ensemble results with empirical data. The network shuffle results exhibit a distribution of comparisons over the species similar enough for this to not affect the overall conclusions (see Additional File 1).