[gmx-users] PBC and large hydrogen-bonded molecules

[Mark Abraham]

mernst at tricity.wsu.edu wrote:
> Hello fellow Gromacs-users,
> 
> Today I was extracting PDB snapshots from a solvated DNA 12mer simulation trajectory.
> When I went to view the snapshots, I saw that one of the DNA strands suddenly jumped
> some nanometers away from its partner during certain frames. I remembered that there was
> a note about fixing this artifact in the trjconv documentation, and soon found that the
> -whole or -nojump options produced an easier-to-view snapshot series.

Well done.

> However, viewing a few original snapshots with water left in place, I was struck more
> forcefully by how an entire strand jumps when even part of it reaches the boundaries of
> the current simulation box. A colleague who saw the original snapshots was concerned,
> and seemed to think that both DNA strands should jump together if they were going to
> jump at all. I didn't think it was a problem apart from needing an extra command line
> switch when converting trajectories.

The colleague only thinks it is a problem because he is applying 
chemical knowledge that these things "should" be together. It's no more 
a problem than two H-bonded water molecules being represented on 
opposite sides of the PBC cell. If you're using VMD, there is an option 
to show the periodic images which should set peoples' minds at ease.

> You can see it happen here in the original snapshot extraction, with both strands in the
> same box with solvent:
> 
> http://205.234.232.134/DNA-inbox.jpg
> 
> and then one strand jumping:
> 
> http://205.234.232.134/DNA-pbc-jump.jpg
> 
> Is there anything wrong with just one strand of the DNA jumping? I didn't think so,
> since Gromacs makes whole molecules jump but the 2 strands of DNA are not held together
> by explicit bonds. It looks strange to have one strand almost entirely outside the
> central box, while no water molecules lurk out there, but I have formerly presumed that
> this has no special physical significance and looks strange only because the DNA strand
> is a much larger molecule.

The simplest hypothesis for these images is that the visualisation 
program is placing the first atom it finds in each molecule inside the 
box and the rest of it consistent with that placement. More visually 
appealing might be an algorithm that located the center-of-mass 
inside... shrug. Playing with the trjconv options before the 
visualisation program gets into the mix may help here.

Mark