Question: MrBayes Log likelihood stuck on a single value
gravatar for Whetting
19 months ago by
Bethesda, MD
Whetting1.5k wrote:

Hi All,
I think the below phenomenon may be (in part) a Beagle library issue, but there may be more going on?
I am trying to run a MrBayes analysis on protein alignment (3 partitions). I think there might be a bug.
When I run the analysis locally (no Beagle), everything looks ok, however, when I run the analysis on a cluster (either at the U or on cipres) I get a LnL of -inf. It doesn’t matter how long I run. In addition, the other parameters are stuck at the initial estimate. I think that this is, at least in part, due to use of Beagle. When I turn of Beagle, I now get some of the estimates to change, but the LnL still gets stuck at a ridiculous value (-1.797693e+308; see attached parameter file).

The first couple of lines of parameter file run at cipres (no Beagle) look like this (note that LnL is not changing)

Gen LnL LnPr    TL{all} alpha{1}    alpha{2}    alpha{3}    pinvar{1}   
0           -1.797693e+308  1.557647e+02    4.740000e+00    1.000000e+00    1.000000e+00    
1000    -1.797693e+308  1.561742e+02    4.740000e+00    5.905880e-01    1.000000e+00    
2000    -1.797693e+308  1.551478e+02    4.740000e+00    1.616992e+00    1.000000e+00    
3000    -1.797693e+308  1.561978e+02    4.740000e+00    5.669044e-01    1.000000e+00    
4000    -1.797693e+308  1.564849e+02    4.740000e+00    2.798501e-01    1.000000e+00    
5000    -1.797693e+308  1.567126e+02    4.740000e+00    5.213185e-02    1.000000e+00
6000    -1.797693e+308  1.562584e+02    4.740000e+00    5.063503e-01    1.000000e+00

When I run it locally (notice that at least the LnL is changing)

Gen LnL LnPr    TL{all} alpha{1}    alpha{2}    alpha{3}    pinvar{1}   
0         -1.797693e+308    1.557647e+02    4.740000e+00    1.000000e+00    1.000000e+00
1000    -1.652381e+05   1.565472e+02    4.740000e+00    2.175323e-01    1.000000e+00    
2000    -1.634193e+05   1.566048e+02    4.740000e+00    1.599154e-01    1.000000e+00    
3000    -1.623047e+05   1.566130e+02    4.740000e+00    1.517907e-01    1.000000e+00

Below is the MrBayes block from the Nexus file

begin mrbayes;  
    set autoclose=yes;

    charset Subset1 = 1-979;
    charset Subset2 = 980-2133;
    charset Subset3 = 2134-2838;

    partition PartitionFinder = 3:Subset1, Subset2, Subset3;
    set partition=PartitionFinder;

    unlink statefreq=(all) revmat=(all) shape=(all) pinvar=(all) tratio=(all);

    lset applyto=(1) rates=invgamma;
    prset applyto=(1) aamodelpr=mixed ratepr=variable;
    lset applyto=(2) rates=invgamma;
    prset applyto=(2) aamodelpr=mixed ratepr=variable;
    lset applyto=(3) rates=invgamma;
    prset applyto=(3) aamodelpr=mixed ratepr=variable;

    mcmcp ngen=50000000 savebrlens=yes samplefreq=1000 printfreq= 1000;
    propset ExtTBR$prob=0;

I would appreciate any insights! Thanks

beagle phylogenetics mrbayes • 640 views
ADD COMMENTlink modified 19 months ago by Biostar ♦♦ 20 • written 19 months ago by Whetting1.5k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1491 users visited in the last hour