[gmx-users] pbc nojump with replica exchange behaves differently than with regular MD

[Ramon Crehuet]

Dear Justin, 
Thanks for your suggestion. It works, as long as I set a tpr file in the -s option. So this works: 

gmx trjconv -f md.trr -s md.tpr -pbc mol -center -o whole.xtc 

But the following does not work (where whole_center.gro is a system without water molecules with a whole centered protein and some ions): 

gmx trjconv -f md.xtc -s whole_center.gro -pbc mol -center -o whole.xtc 

I understand that this is what this warning explains: 

WARNING: If there are molecules in the input trajectory file 
that are broken across periodic boundaries, they 
cannot be made whole (or treated as whole) without 
you providing a run input file. 

This would seem anecdotal, but the problem is that I saved only the non-water system in the md.xtc. Therefore the first line works if I select to output only the protein atoms, but it gives the following error if I want the non-water atoms as output (protein + ions, all that is in the md.xtc). 

Fatal error: 
Index[413] 36378 is larger than the number of atoms in the 
trajectory file (433). There is a mismatch in the contents 
of your -f, -s and/or -n files. 

Again, this makes sense, as the ions are stored after the water molecules.But I am stuck with this two options. 
So my question is whether is there is a way to use the tpr for a xtc trajectory that contains fewer atoms than the ones defined in the tpr? Or another workaround? 

Thanks, 

Ramon 

On 11/28/19 9:44 AM, Ramon Crehuet wrote:
> Dear all,
> As a follow-up to my question, I have seen that in a regular MD, the coordinates of the original trajectory are always smaller than the unitcell vectors, whereas this is not true in the trajectory from the replica exchange (deviations up to 1.5%). Could this be confusing trajconv?
> Could it be that in an exchange attempt only the coordinates of the atoms are exchanged between temperatures but not the box size? (the dynamics use a Parrinello-Rahman barostat).

The entire state is exchanged (at least in the current version of the 
code that I'm looking at), so the box and everything else gets swapped.

For a simple system of solute in water, you should be able to just do 
trjconv -pbc mol -center without all the intermediate steps (-pbc whole, 
-pbc nojump, etc). I suspect -pbc nojump won't work because you do not 
have a continuous trajectory during REMD, so the coordinates may change 
quite dramatically between snapshots, leading to a breakdown in the 
algorithm.

-Justin

> Thanks again.
> All the best,
> Ramon
>
> ----- On 28 Nov, 2019, at 11:49, Ramon Crehuet < ramon.crehuet at iqac.csic.es > wrote:
>
>> Dear all,
>> I have run a temperature replica exchange and I would like to analyze the
>> resulting trajectory for a given temperature (i.e. not following a replica
>> across temperatures using demux.pl).
>> The simulations consist of a single protein chain. I can easily generate have a
>> whole molecule, named `whole.pdb`.
>> When I was running regular MD, the following worked fine to generate whole
>> structures that diffused out of the box:
>> gmx trjconv -f md.xtc -s whole.pdb -pbc nojump -o whole.xtc
>> But repeating that for one of the md.xtc of the replica exchange does not work.
>> Each atom diffuses or of the box (i.e. without jumps) but to different
>> directions, resulting in an apparently exploding molecule.
>> Should there be a different treatment of RE simulations? If so, what should I
>> do? (I have tried the -pbc nojump step followed or preceded by -pbc whole
>> without success).
>> Thanks for your attention.
>> Best,
>> Ramon

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

Justin A. Lemkul, Ph.D.
Assistant Professor
Office: 301 Fralin Hall
Lab: 303 Engel Hall

Virginia Tech Department of Biochemistry
340 West Campus Dr.
Blacksburg, VA 24061 jalemkul at vt.edu | (540) 231-3129 http://www.thelemkullab.com ==================================================