On Wed, Feb 3, 2016 at 12:27 PM, Krantzman, Kristin D <KrantzmanK.cofc.edu>
wrote:
> Dear Amber Mailing List:
> I am using AMBER to run md simulations of a 20 residue polypeptide in a
> TIP3BOX explicit solvent.
> I have performed simulations using an initial solvent box of 10.0
> Angstroms and also with 15.0 Angstroms.
> The non-bonded cutoff I have been using is 9.0 Angstroms.
> What affects should I be observing if the solvent box is too small?
>
There's the obvious one that your simulation simply won't work. The
minimum image convention is only valid if your cutoff is at least half the
distance between the closest two planes of the triclinic unit cell being
used for periodic boundary conditions. If this is not satisfied, then
multiple periodic images of some atoms *may* be inside your cutoff. But
most codes will catch this situation and quit with an error rather than run
incorrectly (Amber included).
The less obvious one is to check solvent radial distribution functions.
The RDF densities should approach bulk solvent density asymptotically, and
the initial peaks and valleys result from the ordered arrangement of
solvent in the various solvation shells. Your box is too small if the
solvent still feels the effect of the solute at the boundary of the unit
cell, because that indicates a direct, solvent-mediated interaction between
periodic images of the biomolecule.
Basically what you are looking for is evidence of periodicity artifacts or
evidence that periodic images of the solute are influencing each other.
The RDF check is clear evidence that such an interaction exists, but is not
(to my knowledge) sufficient that such interactions don't exist (if the
RDFs *are* asymptotically "correct"). Other people may have other tidbits,
though.
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Feb 03 2016 - 11:00:03 PST
