AMBER Archive (2004)

Subject: Re: AMBER: ions in mm_pbsa

From: Thomas E. Cheatham, III (cheatham_at_chpc.utah.edu)
Date: Thu May 06 2004 - 14:07:42 CDT

> I am using mm_pbsa to calculate protein-ligand affinity. I used
> 16 Na+ to neutralize the system. My question is: should I keep ions
> when I generate complex snapshots? Is the implicit solvent model smart
> enough to neutralize the charge?

This is a debatable/research question. Both approaches have been utilized
in the past. Jayaram, Beveridge and co-workers have at least two papers
where the MM_PBSA type analysis includes the explicit ions (JACS 1998,
120: 10629-10633; JPC B 1998, 102: 9571-9576). A number of other groups
including myself, Dave Case, and others, have done MM_PBSA excluding the
bound ions. One issue relates to the fluctuations; you can imagine that
including the explict ions may lead to significantly larger fluctuations
in the electrostatic energies.

While it is true that you can add an effective salt concentration into the
Poisson-Boltzmann or generalized Born analysis, this ionic strength does
not implicitly neutralize the system. If you vary the "implicit" salt
concentration, the shift in the numbers is small (which verifies this).
If, on the other hand, you compared the energies with 1 explicit ion vs. 2
explicit ions and so on, the results would be massively different (and
non-comparable).

Unlike the periodic unit cell / crystal, there is no worry about
divergence in the total charge to -infinity since in the MM_PBSA the
system is no longer periodic. The overall charge will shift the solvation
free energy (implicit solvent) component which will be partially offset by
the explicit electrostatic energy. The KEY is that you are consistent and
have a free energy cycle of comparison, i.e. if you are trying to compare
a -16 DNA to a -15 DNA this will not work.

Another issue is the role of the neutralizing ion(s). If they represent a
diffuse counterion distribution and are not involved in specifically
stabilizing the molecule (say by bridging a drug or a bend across the
major groove in the some but not all of the structures you are comparing),
its inclusion could be important.

We have investigated this in the context of simulations of a G-DNA
quadruplex and various shifted and perturbed models. There we
demonstrated that it was necessary to include (a consistent set of) some
of the bound ions. For a quadruplex binding three ions in the central
channel, we included these three bound ions. To compare to a simulation
where only two ions were bound, we used the two bound ions plus the
closest free Na+ ion (so that we were comparing apples to apples). This
is described in Biophys J. (2003) 85, 1787-1804. We also suggest that the
same may be true for explicitly bound water (such as waters mediating drug
binding); see JACS (2003) 125, 1759-1769. For both those papers, there is
a lot of info in the supplementary material.

So, short answer: Try some things out and see what you get. There are
multiple ways to do things as long as the energetic comparisons are
consistent. Keeping a subset of explicit ions/water can be done using the
"closest" command of ptraj; to do the MM_PBSA you will have to build
"in-vacuo" prmtops that include matching numbers of the ions/water you
have present.

Good luck,

\ Thomas E. Cheatham, III (Assistant Professor) College of Pharmacy, Depts of
| Medicinal Chemistry and of Pharmaceutics and Pharmaceutical Chemistry
| Adjunct Asst Prof of Bioengineering; Center for High Performance Computing
| University of Utah, 30 South 2000 East, Skaggs 201, Salt Lake City, UT 84112
|
| tec3_at_utah.edu (801) 587-9652; FAX: (801) 585-9119
\ BPRP295A / INSCC 418 http://www.chpc.utah.edu/~cheatham

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu