Phycas

Phycas is a software package for Bayesian inference of phylogenies from DNA sequences. It is written by Paul Lewis,Mark Holder and Dave Swofford.

Phycas can read data and trees from NEXUS formatted files. The command language for Phycas is the Python programming language. Python 2.5 is recommended (2.3 is the minimum version that might work). Phycas is open source software available through the GNU Public License. It can be run on Mac, Linux, or Windows, but binaries are only distributed for Mac 10.5 on Intel, and Windows.

Phycas features include:

• Markov chain Monte Carlo simulation to approximate the posterior probabilities of trees and clades (as well as credible intervals for parameters of the model of molecular evolution) under the GTR+Gamma+I model. The most notable features of the MCMC sampler that are not common in other software packages are:
  • Non-zero prior probabilities can be assigned to trees with polytomies.
  • The use of slice sampling to update the parameters of the model
• Summarization of a collection of trees (usually those produced during an MCMC simulation). This includes the calculation of majority-rule consensus tree, as well as some graphical representations of the mixing of the simulation over splits in the tree that can help identify cases in which the MCMC has not been run long enough.
• Simulation of trees and DNA sequences on arbitrary tree topologies (under GTR+Gamma+I model)
• Calculation of the likelihood given a fully specified model (maximization of the likelihood is not implemented).
• A very high level of support for scripting (because the front end to Phycas is the Python interpreter).

The 2008 Workshop on Molecular Evolution marked the official release of the software as version 1.0.0-alpha. The lab exercise at the workshop focused on using Phycas to conduct a Bayesian phylogenetic inference in which polytomies are allowed.

Experimental features that are still in testing include:

• more flexible methods of accommodating rate heterogeneity, and
• a new tool for conducting model selection.

Features that are planned for release in the new version include:

• partitioned and mixture models,
• heated chains, and
• a wider range of MCMC proposals for updating the tree.