MrBayes 3.2: efficient Bayesian phylogenetic inference and model choice across a large model space

Area of Science:

• Computational Biology
• Evolutionary Biology
• Bioinformatics

Background:

• MrBayes is a popular software for Bayesian phylogenetic inference using Markov chain Monte Carlo (MCMC).
• The previous official release was in 2003, necessitating an update to incorporate recent advancements.

Purpose of the Study:

• Announce the release of MrBayes version 3.2, a significant upgrade.
• Introduce new features and performance improvements for Bayesian phylogenetic analysis.

Main Methods:

• Implemented convergence diagnostics and parallel analysis with on-the-fly monitoring.
• Introduced new proposals and automatic tuning parameter optimization.
• Integrated streaming single-instruction-multiple-data (SSE) extensions and the BEAGLE library for faster likelihood calculations, including GPU support.

Main Results:

• Achieved speedup factors of up to 2x with SSE and over 50x with BEAGLE for codon models.
• Introduced new models for relaxed clocks, dating, model averaging, and tree constraints.
• Enabled accurate marginal model likelihood estimation using the stepping stone method for Bayes factor tests.

Conclusions:

• MrBayes 3.2 offers substantial performance improvements and expanded analytical capabilities for phylogenetic inference.
• The new version facilitates more robust and efficient evolutionary analyses, including species tree inference and complex model testing.