Hello,

I'm trying to perform a GA branch analysis method on 32 sequences of 1869 sequences so we had to build a version on our own cluster. Our cluster now has support for HYPHYMPI albeit be it in a non-interactive mode so arguments have to be fed prior to sumbitting the job. Clusterjobs kept dying without any error log files so I therefore tried running this analysis with HYPHYMP on a high-memory machine to see if there were any discernable errors. The following is the output from such a run:

--------------------------------------------------------------
$ ./HYPHYMP ModelSelectorBranchLocal.bf < tempinput


                       +-------------------+
                       |Choose Genetic Code|
                       +-------------------+


       (1):[Universal] Universal code. (Genebank transl_table=1).
       (2):[Vertebrate mtDNA] Vertebrate mitochondrial DNA code. (Genebank transl_table=2).
       (3):[Yeast mtDNA] Yeast mitochondrial DNA code. (Genebank transl_table=3).
       (4):[Mold/Protozoan mtDNA] Mold, Protozoan and Coelenterate mitochondrial DNA and the Mycloplasma/Spiroplasma code. (Genebank transl_table=4).
       (5):[Invertebrate mtDNA] Invertebrate mitochondrial DNA code. (Genebank transl_table=5).
       (6):[Ciliate Nuclear] Ciliate, Dasycladacean and Hexamita Nuclear code. (Genebank transl_table=6).
       (7):[Echinoderm mtDNA] Echinoderm mitochondrial DNA code. (Genebank transl_table=9).
       (8):[Euplotid Nuclear] Euplotid Nuclear code. (Genebank transl_table=10).
       (9):[Alt. Yeast Nuclear] Alternative Yeast Nuclear code. (Genebank transl_table=12).
       (10):[Ascidian mtDNA] Ascidian mitochondrial DNA code. (Genebank transl_table=13).
       (11):[Flatworm mtDNA] Flatworm mitochondrial DNA code. (Genebank transl_table=14).
       (12):[Blepharisma Nuclear] Blepharisma Nuclear code. (Genebank transl_table=15).
       (13):[Chlorophycean mtDNA] Chlorophycean Mitochondrial Code (transl_table=16).
       (14):[Trematode mtDNA] Trematode Mitochondrial Code (transl_table=21).
       (15):[Scenedesmus obliquus mtDNA] Scenedesmus obliquus mitochondrial Code (transl_table=22).
       (16):[Thraustochytrium mtDNA] Thraustochytrium Mitochondrial Code (transl_table=23).

Please choose an option (or press q to cancel selection):
$PATH/Codon file to analyze::
$PATH/Please select a tree file for the data::
Nucleotide a 6-character model string specification (e.g. HKY85 = 010010):?
[Preliminary Step 1] A total of 61 branches. Fitting a nucleotide model to approximate branch lengths...
[Preliminary Step 2] Nucleotide LogL = -30752.5
[Preliminary Step 3] Base model has 76 parameters. Log-L = -27319.3. Mean dN/dS = 0.0816161
Error:
Invalid Matrix Index [0][0] in a -2 by 2 matrix.

Function call stack
1 : sortedScores[0][0]=-crapAIC
-----------------------------------------------------------------

The file tempinput contains the input arguments (could also be given in on the prompt but I chose this option to ensure the problem lies not with the structure of this file):

-------------------------------------
1
nuc_aligned_noemptyrows.fasta
tree.newick
012345
---------------------------------------

nuc_aligned_noemptyrows.fasta is a valid fasta file, aligned for codon positions
tree.newick contains a valid newick string of species in the fasta file with branch lenghts included

I suspect this error I get is the reason why my analysis won't start properly in MPI mode (assuming this error is not caused by the fact this example is run on a single machine of course). Can anybody help me with this?
Thanks

Kevin

Thank you for your reply!

We had to do some searching and turned out it indeed had something to do with the path structure as you suggested. Our infrastructure uses a module load system that immediately loads all the correct binaries and libraries (with the underlying path structure) but the new BranchGA files I downloaded were not put in the right library. After fixing this we got the script working in MPI mode.
I however fear I have another question/problem. After the analysis was finished I tried post-processing the script (on a single high-memory machine) and I get the following error:


$ HYPHYMP BranchGAResultProcessor.bf

$path_to_hyphy/Please locate the output of a GA branch analysis::nuc_aligned_noemptyrows.fasta_ga_branch.out

[PHASE 0] Processing nuc_aligned_noemptyrows.fasta_ga_branch.out

$path_to_hyphy/Save HTML results to:results
[PHASE 1] Reading raw data
       Read 27460 model fits

[PHASE 2] Tabulating Rates
       Best overall c-AIC      = 54587.7
       Best c-AIC with 1 rates = 54812.000. Rate estimates (branches):    0.0816 (  61)
       Best c-AIC with 2 rates = 54624.500. Rate estimates (branches):    0.1705 (  33)   0.0518 (  28)
       Best c-AIC with 3 rates = 54601.500. Rate estimates (branches):    0.1830 (  27)   0.0475 (  21)   0.0802 (  13)
       Best c-AIC with 4 rates = 54594.200. Rate estimates (branches):    0.1669 (  17)   0.0447 (  17)   0.2267 (  13)   0.0676 (  14)
       Best c-AIC with 5 rates = 54589.000. Rate estimates (branches):    0.1724 (  14)   0.0438 (  14)   0.0587 (  11)   0.2351 (  13)   0.0901 (   9)
       Best c-AIC with 6 rates = 54587.700. Rate estimates (branches):    0.1723 (  15)   0.0440 (  13)   0.0901 (  11)   0.2350 (  11)   0.0000 (   2)   0.0587 (   9)
       Best c-AIC with 7 rates = 54588.500. Rate estimates (branches):    0.1721 (  15)   0.0447 (  12)   0.0306 (   1)   0.0901 (  10)   0.2348 (  11)   0.0000 (   2)   0.0587 (  10)

[PHASE 3] Computing Akaike Weights and Model Averaged Rate Estimates


[PHASE 4] Computing Confidence Sets and Generating Result Files
       All models in the 95% confidence set (CS):    3923
       Highest c-AIC in the CS: 54594.60
       Models in the 95% CS with 1 rates =      0
       Models in the 95% CS with 2 rates =      0
       Models in the 95% CS with 3 rates =      0
       Models in the 95% CS with 4 rates =    118
       Models in the 95% CS with 5 rates =   2421
       Models in the 95% CS with 6 rates =   1043
       Models in the 95% CS with 7 rates =    341

[PHASE 5] Making a PostScript tree plot
Error:
Operation 'MAccess' is not implemented/defined for a Number

Function call stack
1 : psLegend*(""+colorMx[0]+" "+colorMx[1]+" "+colorMx[2]+" setrgbcolor\n"+currentPosX+" "+currentPosY+" moveto\n")

Seeing how all scripts on our side should be in place now, could this be a hyphy problem? If necessary, I can attach the nuc_aligned_noemptyrows.fasta_ga_branch.out file but it seems to be a valid output file upon inspection.

K

Thanks again for your swift reply.

The new script works and processing runs properly until the end. However could it be there is still an error in the generation of the .ps plot file? When I open this file I get half always half my tree with coloured branches and all values but to 0 percent. I don't think this is due to my viewer or anything as the lysozyme.ps example in the GAbranch packet opens up just fine. I can again include the .ps file if this should be necessary of course. Other output files seem OK

K

Here it is.

K

Thanks again.

As you can see my technical knowlodge as a biologist sometimes comes short but with that we have everything we need to finish this analysis. Our purpose indeed also lies within finding epsodic bursts of selection so we're definitely also are going to explore this option, especially now that the whole HYPHY package is fully functional on all our servers and shares 

K