Hi,

When using the MEME model on my data, I get the following error message:

Error:
perBranchTree is undefined in call to ReplicateConstraint.

Function call stack
1 : Replicate the following constraint "this1.?.alpha:=_sFactor*this2.?.synRate__" using these variables:perBranchTree;codonTree;
-------
2 : result_String["BRANCHES"]=obtainBranchWiseEBEstimates(sFactor,nsFactor1,nsFactor2,mixingP,filterString)
-------
3 : Branch under condition 'SendJobMEME()'
     to
           <END EXECUTION>
     else
           <END EXECUTION>

Am I doing something wrong?

Regards,

Hi there,

This looks like a bug in the MEME script: you are not using MPI, are you (e.g. running HYPHYMPI on a computer cluster)?

Sergei

Dear Sergei,

Thanks for your reply.

I'm not using MPI. Just MEME_mf.bf with one data partition (I realised aftwerwards
that QuickSelectionDetection.bf, which works well, was more appropriate for the
analysis I am interested in).

On a slightly different topic, I understand that the p-values returned by MEME
correspond to the null hypothesis of no positive selection at each codon site. How can
I use this information to test for the null hypothesis of no positive selection anywhere
in the alignment? Do you recommend applying a Bonferonni correction to the site-specific
p-values and reject the null hypothesis of no positive selection anywhere in the alignment
if at least one of them is below a nominal level? Or, shall I reject the null hypothesis if at least
one non-corrected (i.e., an 'original' site-specific) p-value is below the nominal level?

Regards,

Hi there,

Please take a look at Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login
In our group we generally consider the question of alignment wise testing uninteresting (e.g. you can almost certainly expect there to be some type of positive selection somewhere in a biological alignment). If you are coming from the PAML background, the need to do a gene-wide test before doing site-wise inference was technical : the empirical Bayes estimates of positive selection do not reliably include error estimates in the omega values, etc. Even there, very few people cared about global selection tests, but rather where this selection took place.

In the MEME setting you are directly testing for selection at sites, and if you can say with confidence that AT LEAST one reported site is selected, you should be able to declare that selection operates on the level of a gene. Bonferroni (and other FWER) correction is way too strict -- it checks that ALL of the reported sites are selected.

Sergei

Dear Sergei,

Thanks again for your explanations. Just to make sure I understand properly: the type-I error frequencies reported in your Murrell et al. (2012) paper (SI page 2, Null simulations) correspond to the proportion of neutrally evolving sites (not alignments) at which positive selection was detected. Is this correct?

Regards,

-Stephane-

Hi,

I am using HYPHYMP to run MEME and I get a similar error message, see below.

I used as input a file generated by Gard that works well on the web site. I have compiled HYPHY from the current tar ball, and was able to run Gard and FEL without problems.

Any help will be appreciated.

Regards,

Jose Ribeiro

Follows MEME output till crash:

[MEME PHASE 0] Retuning branch lengths and nucleotide rates under the codon model...
Improved Log(L) BY 14.81274319508702 points
[MEME PHASE 1]. Fitting a codon model site-by-site
| Codon |     Beta- |   Weight- |    Beta+  |   Weight+ |     alpha |      LRT  |   p-value |Full Log(L)|   Est. remaining time (secs)
Error:
perBranchTree is undefined in call to ReplicateConstraint.

Function call stack
1 : Replicate the following constraint "this1.?.alpha:=_sFactor*this2.?.synRate__" using these variables:perBranchTree;codonTree;
-------
2 : result_String["BRANCHES"]=obtainBranchWiseEBEstimates(sFactor,nsFactor1,nsFactor2,mixingP,filterString)
-------
3 : Branch under condition 'SendJobMEME()'
    to
        <END EXECUTION>
    else
        <END EXECUTION>

Hi Jose,

Just to make sure it's not an already fixed issue, can you confirm the bug by pulling down and compiling the latest version from github?

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

Also, are you running MEME in an MPI environment, or on a desktop?

Sergei

Hi Sergei,

I actually made it to work now! Just for the record, this is an example of the command line I am using, it may be of help to some other reader.


(echo 1;echo 1;echo 1;echo 1;echo /home/hp/gard/AGAP028193-PA.gard_finalout;echo /home/hp/meme/AGAP028193-PA.gard_finalout-meme;echo 0.05;echo /home/hp/meme/AGAP028193-PA.gard_finalout-meme.csv;) | /usr/local/bin/mpirun -np 16 HYPHYMPI /usr/local/lib/hyphy/TemplateBatchFiles/MEME_mf.bf > /home/hp/meme/AGAP028193-PA.gard_finalout-MEME-out


Many thanks for making all this available to the community,

Jose Ribeiro