[gmx-users] An argument about long range electrostatics

[Justin A. Lemkul]

Hassan Shallal wrote:
> Dear Gromacs users,
>  
> While I was using Gromos force field in simulating a protein in explicit 
> solvent, I used the following parameters:
>  
> *rcoulomb = 1, rlist =1, rvdw = 1.4*
> *Electrostatitcs : PME*
> *VDW : Twin range cutoff*
>  
> The above situation will not allow the calculation of any *long-range 
> electrostatics (LR-Coul)* while calculating the energy of interaction 
> among the specified energy groups of the the studied system.
>  

Actually, you're calculating lots of long-range electrostatic interactions, but 
with PME it is the reciprocal space term, Coul-recip.

> I am facing the question of to what extent that could affect the 
> accuracy of the calculation of the energy of interaction among the 
> specified energy groups of the the studied system.
>  
> I argue that electrostatic interaction is mainly composed of H-bonds and 
> salt-bridges, both with distance cutoffs of 0.35 nm and 0.4 nm 
> respectively. So there should not be any need for calculating any 
> electrostatic interaction beyond 1 nm (rcoulomb).
>  

If you maintain that all Coulombic interactions occur within 0.4 nm, why 
arbitrarily set a 1-nm cutoff?  By that logic, anything beyond 0.4 nm is 
unimportant.  Therein lies the flaw.  Electrostatic interactions decay over a 
longer range than do other interactions like van der Waals forces.  If an 
interaction at 1.0 nm is worth calculating, why is one at 1.001 nm completely 
unimportant?  Simple truncation leads to notable artifacts, which have been 
documented in the literature.  The first paragraph of the 1995 PME paper 
describes several.

> On the other hand, I came across another argument that calculating the 
> long range electrostatics is *"complicated by the practical limitations 
> of dividing lattice sum energies into energy groups"*. I could not find 
> any explanation of this point in the Gromacs manual! But I assume this 
> kind of argument could be valid if ignoring the long range 
> electrostatics would drastically affect the accuracy of interaction 
> energy calculations mentioned above.

The manual is not a complete repository of all knowledge, but in conjunction 
with the 140 or so references therein, it comes pretty close ;)  The mesh term 
in the PME calculation cannot be decomposed pairwise, that much is true.  But 
the effects on the dynamics of the system when not calculating long-range 
electrostatics (i.e. plain cutoffs) is what is very troubling.  By all means, if 
you want to use the pinnacle of 1980's methodology, use a plain cutoff :)

In any case, the counterargument to yours played out in the literature long ago. 
  As I said, within the first few lines of the 1995 PME paper, you'll find lots 
of reasons to use a long-range electrostatics method.  There are several recent 
demonstrations that PME is the most accurate for a variety of systems.

-Justin

>  
> I would appreciate any feedback or comment concerning the above arguments.
>  
> Thanks a lot
> Hassan
>  
> 

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

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================