Re: [AMBER] Should vdW cutoff distance be system-dependent?

From: David A Case <>
Date: Fri, 6 Dec 2013 14:22:27 -0500

On Fri, Dec 06, 2013, yunshi11 . wrote:
> I understand that AMBER implements (by default) a continuum model to
> correct energy and pressure for vdW interactions beyond the cutoff (e.g. 8
> or 10 angstroms). Does this model simply correct for the r^-6 term that is
> responsible for LJ attractions?


> It also seems to me that such a model assumes an isotropic environment
> beyond the cutoff distance, which is not necessarily true for all
> bio-molecular systems. If one has a large and strangely shaped protein or a
> lipid bi-layer to model, should a much larger cutoff distance be used?

This is certainly true. There have been many publications on the sensitivity
(sometimes rather extreme) of bilayer simulations to the LJ cutoff.

> I wonder if this would also, in theory, affect results for steered MD
> simulations that pulls a ligand away from a protein. When the ligand is,
> say, 8 angstroms from its binding site, it is not in an isotropic
> environment, with one side being amino acid residues and the other being
> waters. Should a larger cut (e.g. 15 angstroms) be applied for this kind of
> simulation?

I don't recall seeing any tests of this issue; perhaps others can chime in.
If you are already (artifically) steering your trajectory, it may be hard to
predict the effects of a bias in how long-range LJ interactions are handled.

> Generally speaking, if one has enough computational resources, would it
> always be safer to use a large cutoff distance for vdW interactions?

Well, a large cutoff implies a large box and a slower-running simulation.
In most cases, this in practice leads to less sampling (e.g. through a shorter
simulation). These sorts of trade-offs depend a lot on what you are trying to


AMBER mailing list
Received on Fri Dec 06 2013 - 11:30:02 PST
Custom Search