Of the 40,000 trees inferred under ΔJC, 83.1% showed an equal or lower RF distance than those reconstructed by the BIONJ-GTR method.
A phylogenetic tree was constructed from MLSA, with the maximum likelihood method and GTR as substitution model (with I: 0.01 and G: 0.52).
To determine the evolutionary relationship between the repeats, we resorted to a pairwise similarity matrix approach (Björklund et al. 2006), as the short and divergent nature of the repeats did not allow for the use of common model-guided phylogenetic methods (e.g., GTR using RaxML; see Materials and Methods).
The NeighborNet method and the GTR distance model were used to create the network.
The level of GTR was determined by the method of Ellman [ 24].
This method (termed BIONJ-GTR) used the BIONJ reconstruction algorithm [ 43] on distances obtained under the general time-reversible model with invariant sites and Gamma distribution of rates across variant sites (GTR+ Γ +I) [ 8, 44].
As for the Bayesian phylogenetic reconstruction method, we chose the GTR + I + G model of nucleotide evolution for all 8 data sets, where the proportion of invariant sites and the shape of the gamma parameter were estimated for each gene or partition employed.
We selected the model that yielded the best AICM value [ 27]: the Gaussian Markov random field skyride model, assuming a lognormal relaxed clock and the GTR substitution model [ 28] (Methods, Additional file 10).
Although not shown here, the 157-protein tree inferred by Bayesian method with the CAT-GTR model essentially agreed with the corresponding ML tree reconstructed under the LG + Γ + F model.
Maximum likelihood (ML) and Bayesian methods using the General Time Reversible (GTR) model of nucleotide substitutions were performed on all four alignments, this model was selected with MODELTEST version 3.06 [ 66].
