Re: AMBER: Atoms overlapping

From: David A. Case <>
Date: Mon, 11 Dec 2006 12:08:58 -0800

On Mon, Dec 11, 2006, Nitin Bhardwaj wrote:

> I am having a strange problem during my simultion. For one of my lipids, I
> have a hydrogen (say H1) connected to an Oxygen (say O1). O1 is in turn
> connected to Phosphorus (P). Also connected to P is another oxygen (say O2)
> that is negatively charged. Picture is sthg like this:
> O2-----P----O1---H1
> |
> |
> C
> Now if I use amber atomtype (instead of gaff), in the first step of
> minimization itself, H1 almost overlaps with O2 (distance being 0.001 Ang)
> while still being bonded to O1. I checked the parameters:
> H1 is atomtype 'ho' (charge 0.4233), O1 is 'oh' (charge -0.7908) and O2 is
> 'o' (charge -0.8586).

This is a common problem with phosphates and similar types of structure (such
as yours). Basically, the ho atom type (which is gaff, not amber) has zero
van der Waals parameters, which allows it potentially to overlap with
negatively charged atoms (such as O2). For water and hydroxides, this can
be prevented by shaking the O1-H1 bond, (because then the repulsion with the
O1 vdW shell prevents fusion). But this does not seem to be sufficient for
the sort of geometry you have above.

In my view, this is a deficiency in the gaff (and Amber) parameterizations.
Junmei may have some comments here. One workaround we have used is
illustrated in the "modified tRNA" parameters you can find at the contributed
parameters data base (off the Amber web site): we added an artificial P-H1
bond, effectively keeping the P-O1-H1 unit fixed (via shake, somewhat like a
water molecule). This "works" in the sense that it prevents the bad overlaps
(or did for us), but we never extensively tested the realism of the resulting
configurations. But maybe something like this would work for you.

> Also, the charges written in the prmtop file for all these atoms are
> different (they are, respectively,
> 7.71386404E+00 -1.44094659E+01 -1.56451201E+01).

Charges in the prmtop file are the "real" charges multiplied by 18.2223.


The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Wed Dec 13 2006 - 05:21:45 PST
Custom Search