Estimating amino acid substitution models and rooting bacterial trees

Abstract

Reconstructing phylogenetic trees from protein sequences normally requires empirical amino acid substitution models to calculate the likelihood of trees or genetic distances between species. The tree of life is classified into three domains of Eukaryotes, Archaea, and Bacteria. The amino acid substitution models have been intensively studied for decades, but few are related to Bacteria. Rooting bacterial trees remains a challenging problem in the phylogenetic analysis due to the long branch separating Bacteria and other domains. The two main objectives of this paper are estimating amino acid substitution models Q.bac and NQ.bac for bacterial evolutionary studies and assessing the capability of the time non-reversible model NQ.bac in rooting bacterial trees. Experiments showed that both the time-reversible model (Q.bac) and the time-non-reversible model (NQ.bac) were significantly better than the existing models in analyzing bacterial protein sequences. Interestingly, the time non-reversible model NQ.bac helped reconstruct maximum likelihood bacterial trees with reliable roots for 177 (23.7\%) out of 748 testing alignments without requiring predefined outgroups. This outgroup-free rooting method enhances the studies of bacterial evolution. We recommend researchers employ both Q.bac and NQ.bac models in analyzing bacterial protein sequences. The datasets and scripts used in this manuscript are available at https://doi.org/10.6084/m9.figshare.20457264.