Computation of likelihood values failed

[GoogleCodeExporter]

What steps will reproduce the problem?
1.Loading NEXUS alignment of 13 sequences (1521 sites)
2.Read perfectly the alignment
3.Launch Computation of likelihood values (same error with all parameters). It 
doesn't launch the error with BioNJ.

What is the expected output? What do you see instead?
---------------------------------------------------------------
*                                                             *
*        COMPUTATION OF LIKELIHOOD SCORES WITH PHYML          *
*                                                             *
---------------------------------------------------------------

::Settings::

 Phyml version = 3.0
 Phyml binary = PhyML_3.0_macOS_i386
 Candidate models = 40
   number of substitution schemes = 5
   including models with equal/unequal base frequencies (+F)
   including models with/without a proportion of invariable sites (+I)
   including models with/without rate variation among sites (+G) (nCat = 4)
 Optimized free parameters (K) = substitution parameters + 23 branch lengths + topology 
 Base tree for likelihood calculations = ML tree
 Tree topology search operation = SPR

::Progress::

Model        Exec. Time      Total Time      -lnL
-------------------------------------------------------------------------

Computation of likelihood scores interrupted. It took 00h:00:16:07.

ML tree for TrN+I+G is invalid.TrN+I+G      00h:00:16:07    00h:00:16:08    0.0000

What version of the product are you using? On what operating system?
jModelTest-2.1.4 on MacOSX 10.8.5

Please provide any additional information below.

Original issue reported on code.google.com by erir...@gmail.com on 28 Sep 2013 at 6:27

[GoogleCodeExporter]

That seems to be a problem with PhyML. If it runs with BIONJ probably the 
alignment is OK.

Could you please paste the PhyML tab/log? There is a PhyML tab in the GUI. If 
you are running jModelTest from console, you should have a 
jmodeltest-phymlxxxxxxxx.log file in the log directory.

Original comment by DiegoDL84 on 28 Sep 2013 at 9:53

• Changed state: Accepted

[GoogleCodeExporter]

I don't run jModelTest from console, I attached the log.

Original comment by erir...@gmail.com on 28 Sep 2013 at 5:18

Attachments:

• PhyMLlog.log

[Alqueno]

I've have a similar issue with PhyML. It turns out that if the sequence names are too long something goes wrong with PhyML. I think that it may be related with some limitation in the strings lengths when phyml build the tree string (Newick format).

As workaround i've renamed the sequences to something like $1, $2, $3, ... $N and then after the analysis renamed back to my original encoding.

I've prepared a bash script which uses AWK and SED to "encode" the alignment and create a .sed file that allows to "decode" the tree to the original sequence names.

Best regards

Andrés