Re: AMBER FF in Gaussian98

From: Eberhard von Kitzing <>
Date: Thu 8 Jul 1999 08:18:44 +0200

Dear James Duncan,

At 6:43 Uhr +0200 08.07.1999, James Duncan wrote:
>I wish to obtain a good starting structure for an 11-mer duplex DNA
>(sequence 5'-TATCGCCGAGA and its complement) to which to add lesions and
>then study the resulting DNA-adduct conformations using a QM/MM method.
>So far, to get my feet wet, I have tried to optimize this 11-mer duplex
>with the AMBER Force Field implemented in Gaussian98, using a Spartan
>(Wavefunction, Inc.) generated B-DNA structure for the input geometry
>and using the charges found in the Kollman article (JACS, 1995, 117,
>5179-5197). Unfortunately, the structure I get from this optimization is
>greatly distorted; only the three most central GC pairs appear

There are two mayor reasons why such a calculation can go wrong.

The first reason could be that the force field is still not
good enough to make the required calculations. E.g., it could have
been designed to work together withe explicit water and ions.
In this case one cannot do very much.

But even if the force field is a reasonable one, such calculation
can go into a wrong direction. Often the intial structure has much
energy in bonds and bond angles. This energy can push the optimization
into unfavorable local minima. If you have a reasonable overall
global structure you could simply start with an optimization
in which the atoms are bound by a weak harmonic force (e.g.,
1 kcal/mol/A^2). This allows considerable internal adjustments
but keeps the global structure. This step takes out high bond

For more details see:

E von Kitzing (1992) Modeling DNA Structures: Molecular Mechanics and
Molecular Dynamics. Methods in Enzymology, DMJ Lilley and JE Dahlberg,
editor(s). San Diego, Academic Press. 449--466.
Received on Wed Jul 07 1999 - 23:18:44 PDT
Custom Search