Software
MrBayes
Homepage
MrBayes was
one of the first and remains one of the most popular programs
for Bayesian inference of phylogeny. It covers a large model
space, supports parallel tempering (Metropolis-coupled MCMC) to
accelerate convergence, and provides on-the-fly convergence
diagnostics based on comparisons among two or more independent
analyses run in parallel. The MrBayes code base is maintained in
collaboration with the (National Bioinformatics Infrastructure
Sweden (NBIS) .
RevBayes
Homepage
RevBayes implements Bayesian
inference for phylogenetic models that are described using a
dedicated model specification language, Rev, inspired by
concepts from the field of probabilistic graphical models. This
allows users to interactively specify their own variants of many
phylogenetic model components, while combining these components
with a wide range of more complex model components available
through flexible function calls in the environment. In this way,
RevBayes covers a much broader range of models than MrBayes. The
models can be used in simulation, computationally efficient MCMC
inference, or path sampling after specifying relevant samplers
of parameters in the model. RevBayes lead developer is former
student Sebastian Höhna, who is now
at Ludwig Maximilian
University in Munich. (Link-out).
TreePPL
Homepage
Our most recent software project, TreePPL is based on universal probabilistic programming. The user defines the phylogenetic model using a simple but Turing-complete programming language, and then the system automatically applies appropriate inference algorithms to learn the parameters in the model. This approach gives the user complete freedom in specifying the model. Developing inference algorithms for programmatic model descriptions is challenging but also rewarding, as the same machinery can cover an impressively wide range of different models. Programmatic model descriptions also facilitate the adoption of new inference strategies that have seen little or no use in phylogenetic previously.