Need for substitution model selection in phylogenetics

The selection of a best-fitting substitution model of molecular evolution has been traditionally considered as a mandatory step in the pipeline of probabilistic phylogenetic inference. However, some recent studies suggested that substitution model selection is unnecessary because they found very small differences among phylogenetic trees reconstructed under different substitution models of protein evolution. This controversy motivated us to explore the influence of selecting a substitution model of protein evolution on the accuracy of phylogenetic tree reconstruction with a clear study based on real and simulated data. Our results indicate that selecting the best-fitting substitution model of protein evolution leads to more accurate phylogenetic tree reconstructions, considering both topology and branch lengths, especially if the input genetic data presents large genetic diversity (i.e., sequence identity below 0.7). We conclude that selecting the best-fitting substitution model of molecular evolution is crucial for data with large genetic diversity and recommended for data with small genetic diversity