Repository files navigation

CoupledMCMC is aBEAST 2 package, which you can install through thepackage manager that comes with BEAST.ChooseCoupledMCMC from the list of packages.

After you installed theCoupledMCMC package (version 0.1.5 or better), theMCMC2CoupledMCMC app becomes available in the app launcher.

1. Create MCMC analysis in BEAUti with any of the available templates, save asmcmc.xml

Now there are 2 ways to proceed:

2a. from a terminal, run

/path/to/beast/bin/applauncher MCMC2CoupledMCMC -xml mcmc.xml -o mc3.xml

This creates a filemc3.xml containing a CoupledMCMC analysis with the same model/operators/loggers etc as themcmc.xml analysis.

2b. from BEAUti, use menuFile > Launch apps, selectMCMC to Coupled MCMC converter from the available apps, fill in form and click OK

In order to set up a pre-prepared xml to run with coupled MCMC, open the*.xml and change the MCMC line in the xml.

To do so, go to the line with:

<run spec="MCMC" chainLength="....." numInitializationAttempts="....">

To have a run with coupled MCMC, we have to replace that one line with:

<run spec="beast.coupledMCMC.CoupledMCMC" chainLength="100000000" storeEvery="1000000" deltaTemperature="0.025" chains="2" resampleEvery="10000">

• chainLength="100000000" defines for how many iterations the chains is run
• deltaTemperature="0.025" defines the temperature difference between the chainn and chainn-1. This value should be changed such that the acceptance probability of a swap is between 0.25 and 0.6
• chains="2" defines the number of parallel chains that are run. The first chain is the one that explores the posterior just like a normal MCMC chain. All other chains are what's calledheated. This means that MCMC moves of those chains have a higher probability of being accepted. While these heated chains don't explore the posterior properly, they can be used to propose new states to the one cold chain.

Parallel Metropolis coupled Markov chain Monte Carlo for Bayesian phylogenetic inferencehttps://academic.oup.com/bioinformatics/article/20/3/407/186341