AMBER: Re: About RESP Calculation

From: FyD <>
Date: Mon, 22 Mar 2004 18:09:00 +0100 (CET)

Dear Qing Zhang,

> Thanks a lot for your detailed explanation on the
> inaccuracy caused by "orientation". I also visited
> your website. If I understand correctly, the
> "orientation" refers to both molecule transformation
> (whole molecule rotation and/or translation) and
> internal conformation change (such as the two methyl
> groups in dimethylphosphate). So I have three
> questions as following:

Orientation and conformation are different... If you take the same conformation,
you can align (by translatations and rotations) this conformation on the X axis
or on the Y axis (for instance) and this conformation will have two different
orientations (i.e 2 different set of Cartesian coordinates).

About conformation and ESP charges:
See Stouch et al JCC 1992 13 622-632.

About orientation and ESP charges:
See ref. 14, 15 in RED manual and previous emails.

> 1.
> > Thus our goals in developping RED were:
> > - To solve "inaccuracy problems" related to the RESP
> > ESP charge derivation
> > protocole as those described above.
> What is your method to solve the inaccuracy caused by
> orientation? I see a "new RESP fitting procedure"
> mentioned on RED website. Since the RESP fitting
> procedure is after Gaussian or GAMESS calculation on a
> specific orientation, how does your new procedure
> eliminate the orientation effects from Gaussian or
> GAMESS results?

A rigid-body re-reorientation algo. applied on the minimized structure (from
Gaussian or GAMESS). See RED manual pages 13-14.

> Do you average RESP results on many
> orientations?

Yes we do average RESP results on 'many' orientations: we called it
multi-orientation-RESP fit by analogy to the multi-conformational RESP fit, see

> 2. Clearly, AMBER 1995 partial charges were produced
> on a specific orientation. Do you think which choice
> is better for MD simulations: a "new RESP fitting
> procedure" only on a part of a molecule OR consistency
> with current AMBER 1995 partial charges for the whole
> molecule? We know that AMBER and CHARMM have quite
> different partial charges, but both perform well.

As I already said, the question is: Do you want that people that will read your
works can REPRODUCE your published charges ? If "yes", use RED with the rigid
body algo. if "not" you can still use RED however without using the rigid-body
algo. (keeping the orientation selected by the ab initio software)

On the contrary, we do NOT question the VALIDITY of such charges in AMBER and
CHARMM. As you said, it is known they perform well.

Reproducibility and validity is not the same thing...

> 3. Does the orientation "inaccuracy" really matter in
> MD simulations? How much is the "inaccuracy" caused by
> software updates and machine differences?

See RED manual pages 3-4.

As example, if you take the dimethylphosphate example the charge incertainty
(because of the orientation of the 2 methyl groups) is 0.016. It is unlikely
that such charge incertainty will affect MD simulation. Now, if you build a box
of solvent using EtOH or DMSO and you have a charge incertainty of .046, this
starts to be more important. I do not have the answer for this particular case.
Thus, using RED you can apply a multi-re-orientation RESP fit that allows to
average the differences in charge observed for one orientation over several

All this, is explained in the RED manual...

As I said before, with RED version 2.0, the user will be able to apply
multi-orientation and/or multiconformation RESP fit. This will allow to
associate the 'validity' of the RESP charges (using multi-conformational fit)
and the 'reproducibility' of the RESP charges (using multi-orientational fit).

Best regards, Francois

F.-Y. Dupradeau
The Scripps Research Institute, San Diego, CA
Faculte de Pharmacie, UPJV, Amiens, France
The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Mon Mar 22 2004 - 17:53:00 PST
Custom Search