Welcome, Guest. Please Login
YaBB - Yet another Bulletin Board
 
  HomeHelpSearchLogin  
 
a different freezing problem (Read 1379 times)
jssdenton
YaBB Newbies
*
Offline


Feed your monkey!

Posts: 3
a different freezing problem
Jan 17th, 2011 at 11:17am
 
Hi folks.

I'm running HyPhy 2.0 on OSX 10.6.6, and trying to do a partitionwise model selection using nucmodelcompare.bf and a user-input tree topology (I removed the branch lengths).

The dataset is 196 taxa and 4833 bp total, or 4833/3 for each partition.

I try to run nucmodelcompare, and HyPhy freezes. Freezing occurs generally after the initial likelihood calculation of the (012345) GTR model.

Is there a best way to specify a partitionwise model test in HyPhy? Could that be the problem?

I have a SETS block in my NEXUS file. The file is attached.

Thanks!

~John
Back to top
 
 
IP Logged
 
Sergei
YaBB Administrator
*****
Offline


Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male
Re: a different freezing problem
Reply #1 - Jan 19th, 2011 at 12:01pm
 
Hi John,

The default version of 'nucmodelcompare.bf' does not perform partition-based testing, as far as I recall. Did you modify the file -- if som please attach it for me to try to recreate the issue.

Otherwise, could you please list the options you were feeding in to the analysis?

Sergei
Back to top
 

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego
WWW WWW  
IP Logged
 
jssdenton
YaBB Newbies
*
Offline


Feed your monkey!

Posts: 3
Re: a different freezing problem
Reply #2 - Jan 20th, 2011 at 9:19am
 
Hi Sergei,

Thanks. I did not modify the nucmodelcompare.bf script. What I was going to try to do was to use a SETS block to define a single codon position and make three files--one for 1st position, one for 2nd position, and one for 3rd position and run them separately. The entire dataset would be there each time, but only an individual partition would be assigned. I was assuming that HyPhy would then only fit a model to the defined partition.

The guide tree I was using is a NJ tree built with GY94 distances (output from HyPhy). I was going to use this tree for fitting all partitions.

The options I used for nucmodelcompare did not alter the freezing. What I tried most frequently were

nucmodelcompare->global w/variation->every time->gamma->4->input NEXUS+tree->alpha=0.05

I have attached a zip of the file + tree.

Thanks!
Back to top
 
Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login Login (179 KB | )
 
IP Logged
 
Sergei
YaBB Administrator
*****
Offline


Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male
Re: a different freezing problem
Reply #3 - Jan 21st, 2011 at 11:34am
 
Hi John,

I am able to run your data on v2.0 of HyPhy without 'freezing'. Which version, OS combination is giving you problems?

In terms of partitioned model selection:

1). If this is coding data, I would suggest running a CODON-based model selection, because nucleotide positional models are simply approximations for those. Take a look at Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login Login
If you are interested in the nucleotide bias component of the model, you can try CodonModelCompare.bf

2). HyPhy will ignore CHARSETS in nucmodelcompare.bf -- you should split the file into 3 input files (one for each position) and then feed them through nucmodelcompare.bf.

Sergei
Back to top
 

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego
WWW WWW  
IP Logged
 
jssdenton
YaBB Newbies
*
Offline


Feed your monkey!

Posts: 3
Re: a different freezing problem
Reply #4 - Jan 21st, 2011 at 12:42pm
 
Hi Sergei,

I am using 32-bit OSX 10.6.6 and the version of HyPhy 2.0 based on a previous post in the forum from Dec 23rd, I think.

Thanks for the paper. I am using the codon position approximation given the size of the dataset.

Do you have any recommendations on how to generate the datasets of individual positions? I had used an 'exclude' command for regular modeltest-like implementations.
Back to top
 
 
IP Logged
 
Sergei
YaBB Administrator
*****
Offline


Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male
Re: a different freezing problem
Reply #5 - Jan 21st, 2011 at 1:08pm
 
Hi John,

The GUI version for OS X has an obscure bug (which I thought I had squashed) that 'freezes' for multithreaded analysis (multiple CPUs). I suspect this is what is happening in your case -- I'll try to replicate the issue and attempt to debug it. Thanks for letting me that it's still there.

For your model selection problem I suggest the following:

1). Run the analysis in the command line version of HYPHY and see if the problem goes away.
2). Datamonkey.org can do model selection on up to 500 sequences (using an MPI cluster -- should be much faster than a desktop).
3). I would use the comb filter (Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login Login) to split your alignment into individual positions, save each one to a separate file and then run model selection.

Based on your use case, I am considering adding default support for including/excluding positions if the input file is in NEXUS format and the user provides partitions.

Cheers,
Sergei
Back to top
 

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