On Tue, Mar 19, 2013 at 11:00 AM, Josep Maria Campanera Alsina <
campaxic.gmail.com> wrote:
> Dear all,
> I was checking some interactions in a pairwise calculation (MMPBSA
> perl, AMBER11) and I've notice some unrealistic or at least high
> values in the electrostatic term.
>
> For instance the electrostatic interaction between residue 13 (GLU)
> and 123 (LYS) is -5.330 kcal/mol altough they are 32 angstroms far
> away. This seems related to the cut parameter. If instead of the
> default value of 999.0 (in perl PBSA) one uses 9.0 then the pairwise
> interaction falls to 0 which seems more reasonable.
>
Setting a cutoff of 9 Angstroms will ignore all electrostatic interactions
beyond that distance, so the Glu13 -- Lys 123 interaction will be 0 simply
because those interactions are not calculated. Electrostatic interactions
are very strong and quite long-range -- they decay as 1/r (and the sum of
1/i for all integers i is actually infinity -- that's how slowly 1/r
decays).
As an example, consider two monovalent ions of opposite charges (at long
distances, GLU and LYS interact this way). At 32 Angstroms, the
electrostatic energy is 332*1*-1/32 = -10.38 kcal/mol. For comparison, at
1 Angstrom the interaction energy is -332 kcal/mol. Since reported
pairwise interactions are actually divided by 2 to avoid double-counting,
this number seems correct (-5.33 is roughly 1/2 of -10.38).
> TDC 13-> 123 0.000 -0.000 -5.330 5.263 0.000
>
> Therefore, my question is why particle-mesh Ewald (PME) procedure
> which handles long-range electrostatic interactions calculate such a
> value at 32 angstoms of distance? I can understand that with a cutoff
> of 999.0 the procedure includes long interactions but at 32 angstroms
> .... I would think that there is no electrostatic interaction. What
> am I missing? What I am doing wrong?
>
You never use PME in your example. PME is for periodic systems only, and
actually calculates the full electrostatic interaction between all
particles in all periodic boxes (out to infinity). The electrostatic
interactions calculated via Ewald is _not_ pairwise decomposable due to its
use of Fourier transforms in the long-range contribution (and the part that
is sum-over-pairs is artificially damped, so it is not a helpful quantity
on its own).
All the best,
Jason
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Mar 20 2013 - 07:30:02 PDT
