bModelTest: Bayesian phylogenetic site model averaging and model comparison.

Area of Science:

• Computational Biology
• Evolutionary Biology
• Bioinformatics

Background:

• Bayesian methods offer a robust framework for phylogenetic reconstruction, providing uncertainty estimates and integrating diverse data.
• Phylogenetic analyses of nucleotide sequences require specifying site and substitution models.
• Current methods often rely on pre-determining site models using likelihood-based approaches, which can be suboptimal.

Purpose of the Study:

• To develop a Bayesian approach for inferring and marginalizing site models in phylogenetic analyses.
• To integrate site model selection directly into the phylogenetic inference process.
• To offer a more flexible and statistically sound alternative to pre-determined model selection.

Main Methods:

• Implemented a trans-dimensional Markov chain Monte Carlo (MCMC) framework within the bModelTest package for BEAST 2.
• Enabled switching between various nucleotide substitution models.
• Estimated posterior probabilities for rate heterogeneity (gamma distribution), proportion of invariable sites, and unequal base frequencies.

Main Results:

• Successfully inferred and marginalized site models during the MCMC analysis.
• Demonstrated the utility of bModelTest with the full set of time-reversible nucleotide models and introduced two subsets.
• The method allows joint inference of site models and phylogenetic trees.

Conclusions:

• bModelTest automates site model selection within a Bayesian framework, eliminating the need for separate, pre-determined analyses.
• This approach enhances the accuracy and efficiency of phylogenetic reconstruction using nucleotide sequence data.
• The bModelTest package is integrated into BEAST 2, providing an open-source solution for advanced phylogenetic analyses.