AMBER Archive (2004)

Subject: Re: AMBER: question about delphi and UHBD

From: Thomas E. Cheatham, III (
Date: Tue Sep 28 2004 - 14:34:59 CDT

Dr. Luo wrote:

> Aside from different numerical implementations of molecular surface and
> finite difference solvers, Delphi and UHBD also use very different
> methods to compute reaction field energies. So in the end, you can't get
> agreement at all for a complex molecule, such as a large protein.

>From the original e-mail of Shawn:

> The final solvation energies for ligand, receptor and complex are
> -1466, -1529 and -2730 kcal/mol in UHBD and -1035, -1056, -1224
> kcal/mol in delphi.

Despite the different algorithms used and the size of the systems being
studied, I find it rather difficult to believe that these differences in
energetics are reasonable between the two programs. One program suggests
a free energy difference of 265 kcal/mol and the other 867 kcal/mol for
the complex - (receptor + ligand) favoring the separated proteins. Both
of these are completely unreasonable (if they are the final solvation
energies). The disagreement between the two methods makes me wonder if
the numbers are indeed comparable. Do the continuum solvation numbers
include anything else that you do not realize? Is the conversion from
reaction field energy to kcal/mol correct? Are both running with the same
internal dielectric? Even generalized Born and continuum methods, in
general, show better agreement than is seen between UHBD and delphi shown
above leading me to speculate that something is amiss either in the
running or interpretation of the results. I do not think this relates too
heavily to the number of iterations.

To test out your knowledge of delphi and UHBD, try running a really simple
test case, like a continuum run on an isolated Na+; both programs should
give a number around 97 or so for the Na+ parameters (Aqvist adapted)
present in AMBER.

Good luck.

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to