How To Show Mrbayes Pp Values?

Question

[michal.oskiera]

Hello, I have work to do with MrBayes, but I can`t see the posteriour probabilites in final file .nex.con.tree - I opened it with TreeView and there was nothing to show.

I used commands (set 39seq 700bp):

begin mrbayes; lset nst=6 rates=invgamma ploidy=haploid; mcmcp ngen=5000000 nchains=4 temp=0.1 filename=chi18consPROMALSGblockMrB_out starttree=random savebrlens=yes samplefreq=100 printfreq=500 diagnfreq=1000; prset tratiopr=fixed(2.37) brlenspr=unconstrained:exp(10) topologypr=uniform popvarpr=variable statefreqpr=dirichlet(0.23,0.33,0.22,0.22) revmatpr=dirichlet(1.15,2.42,0.59,0.59,5.03,1.00) shapepr=fixed(1.169) pinvarpr=fixed(0.428); end;

Now, I dont know whats wrong - commands I used, or program to open the file (TreeView)?

Does any one know where I do mistake?

Another question:

Could anyone help to configure 8 CPU intel Xeon on Windows 7 to work parallel with MrBayes? (I know than I should configure MPI on Windows 7, but how to do this?)

Best Regards

Answer 1

[Steve Moss]

Its always a good idea to read the paper and the manual first (although, saying that, the manual for 3.2 isn't yet complete).

I assume you are using MrBayes 3.2.x? The new version of MrBayes outputs a single .con.tre file with the consensus tree. The posterior probabilities are located within the nexus comments. Most programs, not being aware of the new MrBayes format, will not display the posterior probabilities (TreeView is quite an old program). I tend to use FigTree for most of my tree visualisation, and according to the manual (page 34) it supports this format. One way to work around this, if you wish to use different tree viewing software, is to enter the command conformat=simple in order to revert to the more simple form of tree output (see page 35 of the manual). This will output two trees; one with branch lengths and support values and one with only branch lengths.

One of the problems with MrBayes, is it's lack of support for multithreading, however it does support MPI. From what I can tell the MPI versions are focused towards UNIX based systems, such as OS X and Linux, however, Windows does have MPI software available. I've used OpenMPI in the past and you would want the 64bit version, OpenMPI_v1.5.5-1_win64.exe based on your architecture. Unfortunately, the compilation section is one of those missing from the 3.2 manual. Looking at the source code, the Visual Studio project file is missing, which was included in previous versions. I'll need to do a bit more tampering, but at the moment, it looks like you'll have to make do with serial, or install Linux for easier MPI support.