MrBayes MPI version and weird tree

Question

[cryomics]

Hi all,

Lately I've been using MrBayes 3.2.2 on two different supercomputer platforms.

I ran a first analysis using a GUI version MrBayes and the Mixed jumping model. I didn't reached convergence with 2 000 000 generations but the resulting tree looked pretty normal (featuring sequence clusters).

I couldn't resume the analysis with this GUI so I started to use another platform and ran the analysis from the scratch with a MPI version of MrBayes. Today I finally hit convergence with 3 000 000 generations. However, the concatenated tree looks like an urchin and makes me think that I missed a flag or something.

Does anyone has a suggestion of what is going on here, please?

Happy holidays to you all and thanks in advance!

Anne-Lise

[Michael]

Hard to say if something is missing without knowing the parameters or input data you used! Did you use the same nexus file as input?

[cryomics]

Hi Michael,

Yes I used the same nexus file. Here is the script I used for the batch analysis:

>begin mrbayes;
>execute 06122015Mixed.nexus;
>prset aamodelpr=mixed;
>mcmc nruns=2 ngen=3000000 samplefreq=250 printfreq=1000 nchains=4 temp=0.0025
>checkpoint=yes append=yes;
>sump;
>sumt;
>end;

I think it's pretty basic.I should add that I got the same type of aberrant tree with an analysis that I run on the MPI version as well, but using the GTR model this time and a very similar script.

Does it help ?

Thank you, Michael, for answering. :-)

Answer 1

[cryomics]

I just realized that the first analysis that looked good even if convergence wasn't reached was run with 3.2.0 version while my recent analysis was run on 3.2.2. It has to be a difference in the default parameters of the 2 versions...

And it looks like that is coming from a bug on the 3.2.2: https://sourceforge.net/p/mrbayes/bugs/1599/

[Michael]

Ok, good you found it out. The latest version is 3.2.5, maybe time to update.