OBAMA

OBAMA performs Bayesian model averaging over amino-acid substitution and site models to account for model uncertainty in phylogenetic analyses.

Key Features:

• Model averaging: Averages over empirical amino-acid substitution and site models to let the data inform model choice and to account for model uncertainty.
• Trans-dimensional MCMC proposals: Uses trans-dimensional Markov Chain Monte Carlo proposals to switch between empirical substitution models, including Dayhoff, WAG, and JTT.
• Base frequency adaptation: Can switch between base frequencies specified by substitution models and frequencies estimated directly from the alignment.
• Rate heterogeneity and invariable sites: Supports inclusion or exclusion of gamma rate heterogeneity and a proportion of invariable sites and estimates associated parameters.
• Bayesian framework: Operates within a Bayesian inference framework that leverages priors for joint estimation of models, model parameters, and trees.

Scientific Applications:

• Reducing model-selection bias: Mitigates bias in phylogenetic estimation caused by selecting a single substitution or site model.
• Accurate phylogenetic reconstruction: Improves accuracy of tree inference from amino-acid alignments by integrating over substitution and site-model uncertainty.
• Parameter estimation: Enables estimation of parameters such as the proportion of invariable sites and the gamma shape parameter across multiple models.

Methodology:

Performs Bayesian inference with model averaging using trans-dimensional MCMC proposals to switch among empirical amino-acid substitution models (e.g., Dayhoff, WAG, JTT), allows switching between model-derived and alignment-estimated base frequencies, and includes options for gamma rate heterogeneity and a proportion of invariable sites while leveraging priors for joint tree and parameter estimation.