Using the final alignment, GTR + Gamma + Proportion Invariant (GTR+G+I) model of DNA evolution was determined by the hierarchical likelihood ratio test implemented in MrModeltest [65] as the best model for the data.
Substitution models for DNA evolution were selected using AIC in ModelTest3.7/ModelTestJ [69]; GTR+G (for original Alignment 1) and GTR+G+I (and moderately stringent Alignment 2 and the stringent Alignment 3).
The substitution model of DNA evolution was selected based on AIC using Modeltest ver. 3.06 [ 69].
A general time-reversible model (GTR [ 43]) of DNA evolution was used, with a gamma-shaped rate variation with a proportion of invariable sites.
For Bayesian and ML analyses, the most appropriate model of DNA evolution was selected using Akaike Information Criteria using ModelTest v 3.7 [ 34].
The best fitting model of protein and DNA evolution were obtained based on the AIC Akaike Information Criterionn) using ProtTest 2.4 server [ 31] and jModelTest v. 2 [ 32], respectively.
Because the sequences differ in lengths and comprise regions of divergent variability, methods that take into account specific models of DNA evolution were considered as the most suitable for the construction of phylogenetic trees, ML and Bayesian (Markov chain Monte Carlo).
The Tamura-Nei [ 40] model of DNA evolution was used, with complete deletion and a homogeneous pattern of nucleotide substitution among lineages and uniform rates of nucleotide substitution across all sites.
On the other hand, to generate a compositionally heterogeneous region (i.e. from 0%to50%0% of the total length, respectively), DNA evolution was simulated with 80% GC-content for the external branches corresponding to the taxa u and x, and 20% GC-content for the two other external branches (i.e. taxa v and y).
For this study we chose to focus on nuclear markers, since rates of mitochondrial DNA evolution have been shown to be very slow in corals, limiting the ability to detect more recent speciation events [ 60].
