From: Jiri Sponer
Subject: Re: AMBER: positive binding free energy
Date: Thu, 11 Mar 2004 18:02:18 +0100 (MET)

[ Charset GB2312 unsupported, skipping... ]

Dear hj zou

It may be a quite intricate problem with
many possible sources of problems and imbalances.
We have done ca 2 years ago a well defined DNA-drug system
and also in this case we could not match the absolute binding
free energies. Specifically, too positive if sol. entr. included.
Relative free energies looked quite good.
We had well defined starting X-ray as well as docked geometries,
well calibrated drug force field.
May be some discussion in that study could help you.
Molecular dynamics simulations and thermodynamics analysis of DNA-drug complexes.
Minor groove binding between 4 ',6-diamidino-2-phenylindole and DNA duplexes in solution
Spackova N, Cheatham TE, Ryjacek F, Lankas F, van Meervelt L, Hobza P, Sponer J
J. Am. Chem. Soc.
125 (7): 1759-1769 2003

I am sure others have very similar experience.

Note that continuum solvent models are in general (regarding
hydration and binding energies) very sensitive to
parameters such as the atomic radii.
This being a major problem also in QM approaches such
as PCM or COSMO etc. (not all published numbers are
justified).
In such systems we need to rely on compenstation of errors.


Best wishes, Jiri


>This is possible. I am not sure that Docking was the step that lead to
>the problem. Nevertheless, we might be more helpful if you can provide
>a bit more detailed information, like how did you do the simulations,
>the main features of the ligand and protein, the typical values of the
>individual energy terms (and the uncertainty of them).
>
>I doubt methods like MM/PBSA is this sensitive to the specifics of a
>force field, assuming you did the right thing for ligands. By the way,
>how did you obtain the ligand model? We are typically very shy on
>questions like "My results are wrong. Please help me to get it right."