[gmx-users] Re: Questions

[Justin A. Lemkul]

lloyd riggs wrote:
> -------- Original-Nachricht --------
>> Datum: Tue, 17 Apr 2012 07:49:06 -0400
>> Von: "Justin A. Lemkul" <jalemkul at vt.edu>
>> An: lloyd riggs <lloyd.riggs at gmx.ch>, Discussion list for GROMACS users <gmx-users at gromacs.org>
>> Betreff: Re: Questions
> 
>>
>> lloyd riggs wrote:
>>> Dear Justin,
>>>
>>> So I get no answere on the mail list so decided to ask you (and most
>> likely
>>> Dr. Abrahams if you dont answere) some questions as follows.
>>>
>> Please keep all discussion on the mailing list.
>>
>> http://www.gromacs.org/Support/Mailing_Lists#Mailing_List_Etiquette
>>
>> Your problems are still rather inexplicable to me (hence why I didn't
>> reply), so 
>> I'll ask a few questions and perhaps someone else will have an idea or
>> I'll 
>> stumble upon it based on your answers.  I assumed as of this morning that
>> this 
>> issue was solved:
>>
>> http://lists.gromacs.org/pipermail/gmx-users/2012-April/070528.html
>>
>> Is this not the case?
> 
> No it worked for one, and I happily wrote that, then it didnt work for a single other one?
> 
>>> So my experience with Umbrella and WHAM up untill now,
>>>
>>> I followed the tutorial and a paper using similar proteins which worked
>> well.
>>> I did make one slight change in the pull code for the ki 2000 (set by a
>> paper
>>> with the similar proteins) changing to 2500 kCal/mol over 4 ns. 
>> Basically
>>> the internal code does something where you have to have two seperate
>> values
>>> or the written dhdl always equals zero due to the equation in the code.
>>>
>>> In an case, I did this twice and generated models for an NPT
>> equilabrated
>>> system, with and without a small molecular complex as well generating
>> evenly
>>> spaced models across a 4 ns run each (before the pull code).  I then
>> cliped
>>> the first and last for two sets, and chose 10 randomly spaced models in
>> there
>>> for each with and without (20 total)
>>>
>>> I then production ran these with a rather complicated index file as to
>> later
>>> pull out the contributions energy wise to the overall effects of things.
>>  I
>>> had done this already by hand for one run not included here and was able
>> to
>>> get relativly decent figures by summing energies and pluging these into
>>> standard thermo equations (neglecting obviously some contibutions so
>> relative
>>> but fitting wet laboratory eqperimental error so still good).
>>>
>>> So, I then now have 10 and 10 pullx, pullf, and tpr (along with
>> energies,
>>> etc...).  I started to look at each one individually using WHAM to
>> compare
>>> the profiles and found the following.
>>>
>> How can you use individual files as input into WHAM?  By necessity, the 
>> algorithm works by assembling multiple files.
> 
> same way, using the .dat files with one single file and no error (it just prints 0's for error if you try) and then make error  or Standard deviation analysis in excell or qti plot with all your histograms.
> 

Well, I know you can do it, it just doesn't make much sense and I was trying to 
rationalize what you were doing as a potential source of problem.

>>> As I used a N Y N (my origional axis were on the Y rather than the Z as
>> they
>>> were arbitrarily generated) and found the pullx.xvg files write out the
>>> coordinates for the wrong position, but these can be found using other
>> How so?  I have never seen this happen.
> 
> I dont know.  All I know is if I have it print one of the other deminsions, I get the other curves mentioned.
> 

When accusing the code of doing something wrong, "I don't know" isn't a very 
good justification ;)  The contents of pullx.xvg are the COM coordinates of the 
reference group on the axis or axes along which the restraint was applied, 
followed by the distance between the reference and pulled group, again along 
each axis.  In your case you should have the y-coordinate of the COM of the 
reference, and dY, which represents the distance between the two groups along 
the y-axis only.  These values should be easy to confirm with g_traj and g_dist.

>>> gromacs tools.  Still when I use these the end results (energy
>> calculated at
>>> time 0 and end) are correct, but there is just a jump half way through
>> the
>>> run from a low to high point, but no curve)
>>>
>> I'm not clear on what you mean here.
>>
>>> For the pullf.xvg files (individually), I went through and found only
>> the
>>> runs where there was no interuption (where I didnt have to start from a
>> .cpt
>>> file) the WHAM takes the files and generates nice curves which vary only
>>> slightly as expected.  However any run where the pullf.xvg was
>> interupted and
>>> then continued as pullf.part0001.xvg etc...) and I cut and pasted the
>> ends
>>> into another full pullf.xvg file it cant find the proper columns. 
>> Example.
>> That seems very weird.  These are simple text files; if pasted together
>> properly 
>> I see no reason why their interpretation should be flawed simply by
>> appending 
>> them with a text editor.
> 
> So, A couple of them have almost all (90% of the 2000 points).  If I do not append the final data, they generate a curve that fits in the guasian.  As soon as I cut and paste the ends, it no longer works with the sam error mentioned.  So yes, its wierd to me.  I even looked at the .txt file in a script editor and there is nothing, no carriage no other things, same format as the rest, nothing on the ends...so can not figure out how the software finds an error?
> 
> I even just looked at the code (just the reading and passing for lines portions) and it still should not give an error even with some returns, or other?
> 

The only idea that comes quickly to mind is some problem with end-of-line 
characters.

> 
> The couple I have that crashed early generate a curve but it looks complete crap, pretty much looks like a bell really early?
> 
> 

The following couple of questions I asked are the most important ones, but you 
haven't answered them.  I won't be able to contribute anything else useful 
without an answer to these.

-Justin

> 
>>> ______________________
>>>
>>> Only 2 columns on line 2012 in file pullf_14.xvg Only 2 columns on line
>> 2013
>>> in file pullf_14.xvg
>>>
>> Can you post what these lines are?
>>
>>> ------------------------------------------------------- Program
>> g_wham_d,
>>> VERSION 4.5.5 Source code file:
>>>
>> /tmp/var-tmp/portage/sci-chemistry/gromacs-4.5.5-r1/work/gromacs-4.5.5/src/tools/gmx_wham.c,
>>> line: 1805
>>>
>>> Fatal error: Found 1 pull groups in traj_14.tpr, but 2 data columns in
>>> pullf_14.xvg (expected 1)
>>>
>>> Maybe you confused options -ix and -if ?
>>>
>> Is this a possibility?  Are your files named correctly with respect to
>> their 
>> contents?
>>
>> Further, a complete .mdp file (or at least the complete pull code section)
>> would 
>> be useful.
>>
>> -Justin
>>
>>> For more information and tips for troubleshooting, please check the
>> GROMACS 
>>> website at http://www.gromacs.org/Documentation/Errors 
>>> -------------------------------------------------------
>>>
>>> When I feed it just the column for the energies, it then say it expects
>>> another column?
>>>
>>> I looked at things just to see if there were hidden marks or something,
>> and
>>> con not figure out why it wont take the files.  Theres no hidden marks,
>>> spaces or anything?
>>>
>>> In any case, I now have all these files and need to figure out why I
>> have
>>> such a wierd error.  I should note, the way our software was compiled on
>> the
>>> Linux cluster doesnt allow any append options or I would have just done
>> that
>>> in the first place.
>>>
>>> Sincerely
>>>
>>> Stephan P. Free Watkins stephan.lloyd at students.unibe.ch University of
>> Bern 
>>> Insel Spital RIA Sahli Haüs II Freiburstrasse 21 Bern 3011
>>>
>>>
>> -- 
>> ========================================
>>
>> Justin A. Lemkul
>> Ph.D. Candidate
>> ICTAS Doctoral Scholar
>> MILES-IGERT Trainee
>> Department of Biochemistry
>> Virginia Tech
>> Blacksburg, VA
>> jalemkul[at]vt.edu | (540) 231-9080
>> http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin
>>
>> ========================================
> 
> Thank you for answereing,
> 
> I do sometimes post pretty quickly with problems, but something like this is inigmatic and waste alot of time, and I know its something crazily simple.  Its just anoying as I was trying to keep to a standard for analysis.
> 
> SIncerely,
> 
> Stephan Watkins

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================