[gmx-users] Problems with Jacobi diagonalization

[Tsjerk Wassenaar]

Hi Dayle,

Quite a number of posts here :)

Okay, so there may indeed (still) be a mismatch. Maybe good noting
that g_covar will also accept a .pdb file as reference. One way to be
sure that things go right is to convert the .tpr  to .pdb using the
index with the group 'TDR':

editconf -f topol.tpr -o TDR.pdb -n index.ndx

Then use that file to convert the trajectory to .pdb:

trjconv -s TDR.pdb -f traj.xtc -o traj.pdb

You should be fine taking the system as the atoms should match now. To
check if they do, visualize both topol.pdb and traj.pdb. You'll notice
a mismatch by garbled atom types/names. You have to explicitly check
for that. Such mismatches are a bit hard to spot, as the .xtc file
does not contain atom metadata.

Aside of that there is another possible cause for this error to occur,
and that is when the system is split over the periodic boundaries.
That will also break the routine, but will also immediately show up
when visualizing the system as mentioned above.

Then, by means of 'by the way', are you sure you want to use the -ref
flag with g_covar? This means that the deviations are taken with
respect to the reference coordinates rather than with respect to the
average coordinates. This comes down to calculating a non-central
second moment. To illustrate that, imagine tossing a dice to
investigate the distribution of numbers. Fine, you come to an average
and get the standard deviation the usual way. But what you do here is
calculating the deviation from, say 5, rather than the average of 3.5,
or worse even, from 8. Nice exercise to calculate the second moment
you get in that case and to try and think of what it means. I don't
say it has no meaning; in some cases it makes sense, but you'll have
to have good reasons for doing it and justify the approach when it
comes to publication. More so, it usually requires more thinking of
what such results signify.

Cheers,

Tsjerk


2009/4/2 Dayle Smith <daylemariesmith at gmail.com>:
> Thanks, Mark-
> I re-ran a short MD simulation with TDR as the only xtc group, and used
> tpbconv to create a .tpr file with only TDR (tpxout.tpr) and ran g_covar
> with an index.ndx that contains only TPR (just to be safe!),
>
> g_covar -f traj.xtc -s tpxout.tpr -ref -n index.ndx
>
> and I still get the Jacobi error:
>
> Choose a group for the least squares fit
> Group     0 (         TDR) has    29 elements
> There is one group in the index
>
> Choose a group for the covariance analysis
> Group     0 (         TDR) has    29 elements
> There is one group in the index
> Calculating the average structure ...
> Reading frame       0 time    0.000
> -------------------------------------------------------
> Program g_covar, VERSION 3.3.3
> Source code file: nrjac.c, line: 129
>
> Fatal error:
> Error: Too many iterations in routine JACOBI
>
> ~Dayle
>
>
> On Wed, Apr 1, 2009 at 6:22 PM, Mark Abraham <Mark.Abraham at anu.edu.au>
> wrote:
>>
>> Justin A. Lemkul wrote:
>>>
>>>
>>> Dayle Smith wrote:
>>>>
>>>> Thanks for your help, Justin. I ran gmxcheck -c topol.tpr ("12284 atoms
>>>> in file") and gmxcheck -f traj.xtc ("# Atoms  29"). The difference is
>>>
>>> Indeed, that's the problem, then!
>>>
>>>> the SOL atoms. I made another .tpr file with xtcgroups = TDR SOL, and
>>>> now gmxcheck shows that the number of atoms in topol.tpr is 12284, but I
>>>> still can't get g_covar to work. Maybe these are unrelated problems, I'm not
>>>> sure.
>>>
>>> You haven't really changed anything.  The xtc-grps parameter defines what
>>> was saved in the simulation.  Setting it after the fact does not affect the
>>> already-produced .xtc file.  What you need is a .tpr file that contains only
>>> TDR, so you would have to make modifications to your .top in order to
>>> generate this TDR-only .tpr file.
>>
>> Or use tpbconv - this is the "other" application for that utility.
>>
>> Mark
>> _______________________________________________
>> gmx-users mailing list    gmx-users at gromacs.org
>> http://www.gromacs.org/mailman/listinfo/gmx-users
>> Please search the archive at http://www.gromacs.org/search before posting!
>> Please don't post (un)subscribe requests to the list. Use the www
>> interface or send it to gmx-users-request at gromacs.org.
>> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>
>
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>



-- 
Tsjerk A. Wassenaar, Ph.D.
Junior UD (post-doc)
Biomolecular NMR, Bijvoet Center
Utrecht University
Padualaan 8
3584 CH Utrecht
The Netherlands
P: +31-30-2539931
F: +31-30-2537623