[gmx-users] box size and pbc

[Mark Abraham]

Daniela S. Mueller wrote:
> Mark Abraham wrote:
> 
>> Wang Zhun wrote:
>>
>>> Editconf command generates a box for further solvation. When 
>>
>>
>>  > no definite box size and type are set, it will constructs a cubic
>>  > box which might be the smallest to contain the solute, as far as I 
>> know.
>>  > I set -d with 0.6 and full PBC for this box to simulate peptides
>>  > folding. With VMD I found in one case a peptide moved out of the box
>>  > while in another MD the two termini of the peptide both stretched out
>>  > of the box.
>>
>> This latter observation could be a problem. Now your termini might be 
>> closer to the periodic copies of other termini than you would actually 
>> want. If your termini are zwitterionic, you could be observing them 
>> attracting each other through space. If so, neutralise each with a 
>> counter-ion using genion.
> 
> 
> if the termini interact through the periodic boundary, the box is 
> clearly too small. neutralising the termini is just a cosmetic solution, 
> and maybe you don't want to change your system like that.

Indeed so, it is unwise to have a box so small that such an interaction 
exists, however if increasing the size of the simulation box is 
impractical, neutralizing the termini (either with counter-ions, or 
explicitly) might be an acceptable compromise to make. After all, 
isolated charges require charge separation, which requires work on the 
system... that doesn't happen in biological systems without serious 
machinery. Unless you know otherwise, the real system will have a 
counter-ion nearby somewhere.

>>> For the 1st one I think the PBC is ok, but what about the 
>>
>>
>>  > 2nd case if the two termini point to the opposite box sides? It seems
>>  > there is no post on this editconf option. I know I'd better choose a
>>  > large dodecahedron box but I just wonder if this box should be ok
>>  > to perform MD with PBC.
>>
>> Larger is probably better, but neither cubic nor dodecahedron of 
>> comparable size should be a problem.
> 
> 
> if your protein expands during the run, you either have to adjust the 
> system by increasing the box size; or if you can anticipate this, start 
> with a larger box to begin with.

A normal simulation system in equilibrium won't do this... Equilibration 
at constant pressure will allow the volume to adapt gradually before you 
start a subsequent simulation (at constant P or V) that you care about. 
This would normally be a more effective approach than guessing a size 
larger than the actual system size and hoping the ensuing mess will 
always sort itself out quickly into a meaningful model of reality.

Mark