MCMCtree is a Bayesian MCMC program for estimation of species divergence times using molecular and trait data. MCMCtree was written by Ziheng Yang from University College London. This is a fork of MCMCtree version 4.9j with small modifications suitable for large-scale analyses. The latest version can be found in PAML's repository.
If you use this version of MCMCtree in your publications, please indicate so as MCMCtree version 4.9j-forked (available at https://github.com/dosreislab/mcmctree). General citations for MCMCtree are:
- Yang, Z., and B. Rannala. 2006. Bayesian estimation of species divergence times under a molecular clock using multiple fossil calibrations with soft bounds. Molecular Biology and Evolution, 23: 212–226.
- Rannala B, Yang Z. 2007. Inferring speciation times under an episodic molecular clock. Systematic Biology, 56: 453-466.
The first paper describes the MCMC algorith and the birth-death prior conditioned on soft fossil calibrations. The second paper describes the relaxed clock implementation. If you use the approximate likelihood method please cite:
- dos Reis M, Yang Z. 2011. Approximate likelihood calculation for Bayesian estimation of divergence times. Molecular Biology and Evolution, 28:2161-2172.
To compile the program go into the src/ directory and type make (this assumes you are in a Unix-type system and with build tools available). This will compile the mcmctree executable. If you're planning on running MCMCtree in a high-performance cluster, it may pay off to check if dedicated compilers are available for your cluster's architecture. For example, for Intel clusters, the Intel compiler may be available. In that case, you may want to set CC = icc and CFLAGS = -fast in the Makefile, then compile with make in the cluster. Our tests indicate the Intel-compiled version of MCMCtree is up to 2.5x faster than the GCC-compiled version.
MCMCtree is run in the command line, with program settings spcified in a control file, usually called mcmctree.ctl:
$ mcmctree mcmctree.ctl
A simple tutorial is available in the tutorial/ folder. A book chapter on how to run MCMCtree using phylogenomic datasets is available at github.com/mariodosreis/divtime.
This sectiond describes settings in the control file for this version of MCMCtree which may behave slightly different than the main production version. Please do not assume compatibility with other versions.
seed = -1
seqfile = ape4s.phy
treefile = ape4s.tree
outfile = out
checkpoint = 1
ndata = 1 * number of partitions
usedata = 1 * 0: no data; 1:exact likelihood; 2:approximate likelihood; 3: prepare approx. lkhd files
clock = 1 * 1: global clock; 2: independent rates; 3: correlated rates
model = 4 * 0:JC69, 1:K80, 2:F81, 3:F84, 4:HKY85
alpha = .5 * alpha for gamma rates at sites
ncatG = 5 * No. categories in discrete gamma
cleandata = 0 * remove sites with ambiguity data (1:yes, 0:no)?
BDparas = 1 1 0 * birth, death, sampling
kappa_gamma = 2 .2 * gamma prior for kappa
alpha_gamma = 2 4 * gamma prior for alpha
rgene_gamma = 2 2 * gamma prior for mean rates for genes
sigma2_gamma = 1 10 * gamma prior for sigma^2 (for clock=2 or 3)
print = 1
burnin = 4000
sampfreq = 6
nsample = 5000
checkpoint activates checkpointing. If checkpoint = 1 the state of the MCMC chain will be saved to file mcmctree.ckpt every 10% percentile of the total MCMC run length. Burn-in is used as set in the control file and states during burn-in are not saved. If checkpoint = 2, the program will resume the MCMC chain from the mcmctree.ckpt file. In this case the burnin variable is ignored (i.e., is set to zero internally). The state of the chain is also saved every 10% percentile. Note in all cases mcmctree.ckpt is overwritten everytime the chain is saved, and thus only the last state saved is available. This version of MCMCtree allows unlimited checkpoint runs, simply rename the mcmc.txt file once the run has finished and restart the MCMC chain. The varios MCMC files can then be merged into a larger file and process with print = -1 in the control file. In short, use checkpoint = 1 for the first run in which you need to use burn-in, then use checkpoint = 2 for subsequent runs.