Re: [AMBER] Concerning nonbonded exclusions and MMPBSA

From: Jason Swails <>
Date: Thu, 18 Oct 2012 11:57:58 -0400

On Thu, Oct 18, 2012 at 11:45 AM, Aron Broom <> wrote:

> Hello AMBER users,
> I'm wondering about how nonbonded exclusions are treated in MMPBSA. Are
> they included just as in a normal MD run in AMBER?
> The reason I ask is that I'm somewhat confused about these exclusions. I
> had written my own little code to calculate just the electrostatic
> potential energy portion of something in order to do part of an MMPBSA
> style calculation (for reasons that are too uninteresting to go into, it is
> infeasible for me to use MMPBSA or AMBER for this beyond testing
> purposes). I was confused that my numbers were coming out differently from
> those reported in a normal AMBER run, and tracked the problem to the fact
> that I was calculating all the pairwise interactions, and while I had the
> AMBER cutoff at 999 angstroms, there are still the 1-2, 1-3 complete
> exclusions and the 1-4 scaled exclusions. I had thought these exclusions
> were just there to save computational power because those interactions were
> weak compared to the bond, angle, and dihedral components that would also
> affect those atoms.

No, actually these interactions would be incredibly large since the
distances are so short. They are neglected because the bond and angle
terms are parametrized to account for the full bonded interaction. The
added cost of calculating these exclusions would be next to nothing (since
the exclusion list is so small compared to the full pair-list), so it is
excluded because that is the *actual* definition of the force field.

The way Amber handles this is to set up an "exclusion" list in which each
atom contains a list of atoms in which nonbonded interactions should *not*
be calculated. This is listed in the topology file (NUMBER_EXCLUDED_ATOMS
and EXCLUDED_ATOMS_LIST). The list includes all bonded, angled, and
dihedral-ed atoms; all of which are excluded from the core nonbonded
calculation. The 1-4 interactions are added back in as a sum over unique,
proper dihedrals, after which they are appropriately scaled.

The GB routines take this exclusion list directly from the topology file
(which means you can edit that to add your own exclusion rules via ParmEd
or something), but the PME routines rebuild the exclusion list based on the
system topology.

So anyway, my question basically is: if one wants to calculate the actual
> electrostatic potential energy of a protein for instance, is ignoring those
> exclusions accurate, or should they really be included in order to get
> accurate results? And then, does MMPBSA include them?

They must be included to get accurate results. And yes, MMPBSA includes
them. MMPBSA does not actually compute energies directly, it calls
external programs to do so (either a NAB program or sander). All external
programs that 'correctly' implement the Amber force field implements these


Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
AMBER mailing list
Received on Thu Oct 18 2012 - 09:00:03 PDT
Custom Search