MrBayes

MrBayes is a program for Bayesian inference on phylogenies.

MrBayes

License

Free to use and open source under GNU GPLv3.

Available

Roihu-CPU: 3.2.7a, via the bio-apps module.
Roihu-GPU: 3.2.8 (GPU-accelerated through BEAGLE), via the bio-apps module.

Usage

MrBayes is part of the bio-apps collection on Roihu. On Roihu-CPU, load the bio-apps module tree and then the MrBayes module:

module load bio-apps/v202603
module load mrbayes/3.2.7a

On Roihu-GPU, load mrbayes/3.2.8 instead; see Running on GPUs.

MrBayes is started with the mb command. The same MPI-enabled binary runs serially when started directly:

mb

and in parallel when launched with srun in a batch job. When using the parallel version, you should note that MrBayes assigns one chain to one core, so for optimal performance you should use as many cores as the total number of chains in your job. If, for example, you have specified nchains=4, nruns=2, you should use 4 * 2 = 8 cores.

Batch jobs

Running a MrBayes analysis might take a considerable amount of CPU time and memory. It is, therefore, recommended to run it through the batch job system on Roihu. Shorter test runs can be run in interactive mode using sinteractive. The serial version is recommended for interactive use.

To run a batch job you need to:

Write a MrBayes command file (here mb_com.nex) or include a MrBayes command block in your .nex file. For details, see Chapter 5.5.1 of the MrBayes manual.
Write a batch job script (here mb_batch)
Make sure you have all your input files (here primates.nex)
Submit your job into the queue

The MrBayes command file should include the commands you would type in MrBayes in interactive mode. This example runs the analysis mentioned in Chapter 2 of the MrBayes 3.2 manual.

begin mrbayes;
    set autoclose=yes nowarn=yes;
    execute primates.nex;
    lset nst=6 rates=invgamma;
    mcmc nchains=4 nruns=2 ngen=20000 samplefreq=100 printfreq=100 diagnfreq=1000;
    sump;
    sumt;
end;

Below is an example batch job script for Roihu using 8 cores. We are using 8 cores since our example uses nchains=4, nruns=2, so 4 * 2 = 8.

#!/bin/bash
#SBATCH --account=<project>
#SBATCH --job-name=my_mrbjob
#SBATCH --error=my_mrbjob_err%j
#SBATCH --output=my_mrbjob_out%j
#SBATCH --partition=small
#SBATCH --time=01:00:00
#SBATCH --ntasks=8
#SBATCH --cpus-per-task=1
#SBATCH --mem-per-cpu=4000

module load bio-apps/v202603
module load mrbayes/3.2.7a

srun mb mb_com.nex >log.txt

To submit the job:

sbatch mb_batch

See creating a batch job script for Roihu for more information about running batch jobs.

Running on GPUs

On the Roihu GPU (GH200) nodes, MrBayes 3.2.8 is built with the BEAGLE library with CUDA support, which computes the likelihoods on the GPU. Log in to roihu-gpu.csc.fi and load the modules there to get the GPU build:

module load bio-apps/v202603
module load mrbayes/3.2.8

To use the GPU, add this line to the start of your MrBayes block, before execute:

set usebeagle=yes beagledevice=gpu;

BEAGLE can compute in single or double precision. To use double precision, add beagleprecision=double to the same line:

set usebeagle=yes beagledevice=gpu beagleprecision=double;

The showbeagle command lists the BEAGLE resources that MrBayes can use, so you can check that the GPU is found.

The GPU speeds up the analysis mainly for large alignments with many unique site patterns. For small datasets, the CPU version is usually as fast or faster.

Below is an example batch job script that runs the same 8 MPI processes as the CPU example on one GPU. Here mb_com_gpu.nex is the command file from above with the set usebeagle=yes beagledevice=gpu; line added. All MPI processes share the reserved GPU.

#!/bin/bash
#SBATCH --account=<project>
#SBATCH --job-name=my_mrbjob_gpu
#SBATCH --error=my_mrbjob_err%j
#SBATCH --output=my_mrbjob_out%j
#SBATCH --partition=gpumedium
#SBATCH --time=01:00:00
#SBATCH --ntasks=8
#SBATCH --cpus-per-task=1
#SBATCH --gres=gpu:gh200:1

module load bio-apps/v202603
module load mrbayes/3.2.8

srun mb mb_com_gpu.nex >log.txt

See the GPU partitions for the available GPU resources.

Support

CSC Service Desk