MrBayes

MrBayes performs Bayesian inference of phylogeny using Markov chain Monte Carlo (MCMC) to estimate posterior distributions of phylogenetic trees and associated model parameters.

Key Features:

• Phylogenetic inference: Combines information across data partitions and heterogeneous data types, including morphological, nucleotide, and protein data, under stochastic evolutionary models.
• Parallel processing and convergence diagnostics: Implements Metropolis coupling parallelized via MPI (Message Passing Interface) with convergence diagnostics and the ability to run multiple analyses in parallel with real-time convergence monitoring.
• Performance enhancements: Uses streaming single-instruction-multiple-data extensions (SSE) and supports the BEAGLE library to delegate likelihood calculations to GPUs, yielding speedups (≈2× with SSE and >50× with BEAGLE for codon problems).
• Model flexibility and extensions: Supports relaxed clocks, dating, model averaging across time-reversible substitution models, tree constraints (hard, negative, partial), and Bayesian estimation of species trees via BEST algorithms.
• Advanced statistical outputs: Outputs samples of ancestral states, site rates, site d(N)/d(S) ratios, branch rates, node dates, and statistics on tree parameters for visualization (e.g., FigTree).
• Robustness features: Provides checkpointing across all models to resume long analyses after premature termination.
• Model likelihood estimation: Estimates marginal model likelihoods using the stepping stone method to facilitate Bayes factor tests.

Scientific Applications:

• Bayesian phylogenetic reconstruction: Estimation of posterior distributions of trees and parameters from morphological, nucleotide, and protein datasets.
• Divergence dating and relaxed-clock analyses: Inference of node dates and branch-rate variation under relaxed-clock models.
• Species-tree inference: Estimation of species trees from gene trees using BEST algorithms.
• Model selection: Comparison of substitution and clock models via marginal likelihoods and Bayes factor tests using the stepping stone method.
• Ancestral state and selection analyses: Reconstruction of ancestral states and estimation of site-specific rates and d(N)/d(S) ratios.
• Rate heterogeneity assessment: Analysis of site and branch rate variation for evolutionary rate studies.

Methodology:

Implements Markov chain Monte Carlo (MCMC) with Metropolis coupling (parallelized via MPI), convergence diagnostics, stepping stone marginal likelihood estimation, BEST for species-tree inference, SSE and BEAGLE GPU-accelerated likelihood calculations, and checkpointing.