HyPhy message board - problem running LocalGARD.bf

rubi

YaBB Newbies

Offline

Feed your monkey!

Posts: 9

problem running LocalGARD.bf
Apr 19^th, 2010 at 8:42am

Hi,

I've downloaded the v2.0 developmental build of HyPhy.
Then I ran the buildFromSVN.sh shell script in ./trunk/Scripts.
I then tried to run build.sh MPI but that failed. So I followed on of the previous discussions and changes the COMPILER to mpic++ and deleted lmpich from the LINKER_FLAGS. These changes seemed to work as it compiled and linked and produced a HYPHYMPI executable.

I then downloaded a test MSA file from ran LocalGARD.bf using the command: mpirun -np xx <PATH>/HYPHYMPI <PATH>/LocalGARD.bf with the parameters:
<test MSA file>,010010,2,4,<output>

Unfortunately, the run crashes with the following std output:
Master node received an error:
Invalid Matrix Index [0][0] in a 0 by 2 matrix.
Call stack
1 : sortedScores[0][0]=2*(res[1][0]-myDF*(baseSites/(baseSites-myDF-1)))

[lecs.tau.ac.il:01508] MPI_ABORT invoked on rank 0 in communicator MPI_COMM_WORLD with errorcode 1
[lecs:01508] *** Process received signal ***
[lecs:01508] Signal: Segmentation fault (11)
[lecs:01508] Signal code: Address not mapped (1)
[lecs:01508] Failing at address: 0x31
mpirun noticed that job rank 0 with PID 1508 on node lecs.tau.ac.il exited on signal 11 (Segmentation fault).

what's wrong?

thanks a lot,
rubi

Back to top

IP Logged

Sergei YaBB Administrator Offline Datamonkeys are forever... Posts: 1658 UCSD Gender:	Re: problem running LocalGARD.bf Reply #1 - Apr 19^th, 2010 at 10:05am Hi Rubi, Are you allocating more than two processors when you start a HyPhy MPI instance, i.e. using mpirun? Sergei
Back to top	Associate Professor Division of Infectious Diseases Division of Biomedical Informatics School of Medicine University of California San Diego WWW IP Logged

rubi YaBB Newbies Offline Feed your monkey! Posts: 9	Re: problem running LocalGARD.bf Reply #2 - Apr 19^th, 2010 at 10:40am now that i've allocated 16 it seems to have solved the problem. thanks a lot.
Back to top	IP Logged

rubi

YaBB Newbies

Offline

Feed your monkey!

Posts: 9

Re: problem running LocalGARD.bf
Reply #3 - Apr 21^st, 2010 at 1:19am

Hi again,

As you probably noticed, I'm new in using HyPhy (GARD).
I've tried going through the posts looking for an answer on how exactly to interpret GARD's results on whether there are statistically significant recombination breakpoints or not.
From the output I get after running the HYPHYMPI LocalGARD.bf the only indication for this is in the html result file which either writes: "GARD found no evidence of recombination" or "GARD found evidence of k breakpoints" (k being the number of breakpoints). The html output also suggests running RecombinationProcessor.bf for post process, but when I try it it freezes in the "Fitting a single-tree, multiple partition model" stage.

So my questions are:
1. What is the standard/customary way to determine recombination according to the output.
2. If the answer to 1 indeed involves running RecombinationProcessor.bf, any idea what's wrong?

Thanks a lot

Back to top

IP Logged

Sergei

YaBB Administrator

Offline

Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male

Re: problem running LocalGARD.bf
Reply #4 - Apr 21^st, 2010 at 2:12am

Hi Rubi,

Try

Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to

Login
Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to

Login

You do need to run GARDProcessor.bf

Cheers,
Sergei

Back to top

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego

WWW

IP Logged

rubi

YaBB Newbies

Offline

Feed your monkey!

Posts: 9

Re: problem running LocalGARD.bf
Reply #5 - Apr 21^st, 2010 at 4:01am

Hi Sergei,

I tried running the GARDProcessor_001.bf I downloaded from the posts you referred me to (I also replaced the KHTest.bf) with the command:
<PATH>/HYPHYMPI <PATH>/GARDProcessor_001.bf but still it freezes in the "Fitting a single-tree, multiple partition model" stage.

the nucleotide data file and the GA partition analysis result file are attached in the zipped folder.

Thanks a lot.

Back to top

Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to

Login (28 KB | )

IP Logged

Sergei YaBB Administrator Offline Datamonkeys are forever... Posts: 1658 UCSD Gender:	Re: problem running LocalGARD.bf Reply #6 - Apr 21^st, 2010 at 4:06am Hi Rubi, Do NOT supply the .details file as input to GARD_Processor. It expects the input alignment and the "_finalout" file as input. Best, Sergei
Back to top	Associate Professor Division of Infectious Diseases Division of Biomedical Informatics School of Medicine University of California San Diego WWW IP Logged

rubi

YaBB Newbies

Offline

Feed your monkey!

Posts: 9

Re: problem running LocalGARD.bf
Reply #7 - Apr 21^st, 2010 at 6:09am

Hi again,

thanks for the clarification.

Unfortunately, after the "Fitting a mutilple tree, multiple partition model" notification I get the following error:

Master node received an error:
Unbalanced Parentheses in Expression:]?
Call stack
2 : Build Tree tree_1 from string ]
1 : ExecuteCommands in string "Tree tree_"+pccounter+" = "+partLines[pccounter*2+3]+";" using basepath ./hyphy/HYPHY/./.

[lecs.tau.ac.il:21196] MPI_ABORT invoked on rank 0 in communicator MPI_COMM_WORLD with errorcode 1
[lecs:21196] *** Process received signal ***
[lecs:21196] Signal: Segmentation fault (11)
[lecs:21196] Signal code: Address not mapped (1)
[lecs:21196] Failing at address: 0x31
Segmentation fault

I would be grateful for your help again.

btw,
any use of path (absolute or relative) doesn't help

Back to top

IP Logged

Sergei YaBB Administrator Offline Datamonkeys are forever... Posts: 1658 UCSD Gender:	Re: problem running LocalGARD.bf Reply #8 - Apr 21^st, 2010 at 7:09am Hi Rubi, Can you please attach the xx._finalout file? Sergei
Back to top	Associate Professor Division of Infectious Diseases Division of Biomedical Informatics School of Medicine University of California San Diego WWW IP Logged

rubi YaBB Newbies Offline Feed your monkey! Posts: 9	Re: problem running LocalGARD.bf Reply #9 - Apr 21^st, 2010 at 7:16am Hi Sergei, In the attached zipped folder are the sequence data and all the output files generated by running LocalGARD.bf through HYPHYMPI. Best, Rubi
Back to top	Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to Login (41 KB \| ) IP Logged

Sergei

YaBB Administrator

Offline

Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male

Re: problem running LocalGARD.bf
Reply #10 - Apr 21^st, 2010 at 1:05pm

Hi rubi,

Apologies -- I gave you incorrect advice. The second input file for GARDProcessor.bf should the "splits" file (not _finalout).

On your data, GARDProcessor.bf returns the following test results:

Code:

Breakpoint | LHS Raw p | LHS adjusted p | RHS Raw p | RHS adjusted p
	 428 |   0.00120 |	  0.00720 |   0.01750 |	  0.10500
	1106 |   0.00010 |	  0.00060 |   0.00010 |	  0.00060
	1884 |   0.00010 |	  0.00060 |   0.00010 |	  0.00060

At p = 0.01 there are 2 significant breakpoints
At p = 0.05 there are 2 significant breakpoints
At p = 0.1 there are 2 significant breakpoints

Mean splits identify: 0.116667

Sergei

Back to top

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego

WWW

IP Logged

rubi YaBB Newbies Offline Feed your monkey! Posts: 9	Re: problem running LocalGARD.bf Reply #11 - Apr 21^st, 2010 at 11:29pm Hi Sergei, Thanks a lot for the clarification. Just to make sure I follow, a breakpoint is determined to be significant at a given significance level if its computed p-values according to all tests (LHS raw, LHS adjusted, RHS raw, and RHS adjusted) are below that significance level? thanks, rubi
Back to top	IP Logged

rubi YaBB Newbies Offline Feed your monkey! Posts: 9	Re: problem running LocalGARD.bf Reply #12 - Apr 22^nd, 2010 at 2:17am another question btw, it seems that GARD often determines a breakpoint in the first position in the alignment (significant according to all statistical tests performed by GARDProcessor.bf). is this reliable or a know artifact? thanks, rubi
Back to top	IP Logged

Sergei

YaBB Administrator

Offline

Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male

Re: problem running LocalGARD.bf
Reply #13 - Apr 22^nd, 2010 at 5:05am

Hi Rubi,

Both of the adjusted KH tests (Left hand side and Right hand side) must be below a certain alpha-level (p-value) for a breakpoint to support significant topological incongruence.

Sergei

rubi wrote on Apr 21^st, 2010 at 11:29pm:

Hi Sergei,
Thanks a lot for the clarification.

Just to make sure I follow,
a breakpoint is determined to be significant at a given significance level if its computed p-values according to all tests (LHS raw, LHS adjusted, RHS raw, and RHS adjusted) are below that significance level?

thanks,
rubi

Back to top

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego

WWW

IP Logged

Sergei YaBB Administrator Offline Datamonkeys are forever... Posts: 1658 UCSD Gender:	Re: problem running LocalGARD.bf Reply #14 - Apr 22^nd, 2010 at 5:06am Hi Rubi, This behavior is not normal -- unless the original GARD analysis returned a result of NO RECOMBINATION (if I recall correctly). Would you be able to attach an example? Sergei
Back to top	Associate Professor Division of Infectious Diseases Division of Biomedical Informatics School of Medicine University of California San Diego WWW IP Logged

	Welcome, Guest. Please Login