Subject: Re: AMBER: About RESP Calculation
Date: Sun, 21 Mar 2004 08:30:41 +0100 (CET)
From: FyD

Dear Qing Zhang,

> The published charges can be reproduced well, although
> not exactly, if one strictly follows the paper:
> Piotr Cieplak, Wendy D. Cornell, Christopher Bayly, &
> Peter A. Kollman, "Application of the multimolecule
> and multiconformational RESP methodology to
> biopolymers: charge derivation for DNA, RNA, and
> proteins", J. Comp. Chem., 16: 1357-1377, 1995.

- Using program Gaussian, the charges "can be" reproduced "although not exactly"
as you said. ;-)
- Using program GAMESS the differences in charges "might be" slightly more
important...
Now, what is the meaning of "can be", "although not exactly" and "might be" ?
Reproducibility after 2, 3 or 4 digits after the decimal point ?

As I said before, in a previous email, from our tests, we found a charge
difference up to 0.043 e. Such differences strongly depend on the model studied
(It is why RESP charge derivation might be tricky in 'some' cases). Good
examples are models EtOH or DMSO which present 2 different orientations (after
minimization using the re-orientation scheme available in Gaussian) and a
maximum charge difference of 0.043 e for one heavy atom. On the contrary, for
some other models we were only able to generate a unique orientation using
Gaussian. In this case, the charge differences were obviously smaller and the
charge reproducibility really better (directly related in this case to the
minimization threshold criteria used in Gaussian).

> One can also
> easily apply it to other unusual deoxynucleotides. The
> protocol is at:
> http://monod.biomath.nyu.edu/~qzhang/BP-adenine/protocol.txt

Let's follow your link: dimethylphosphate (conformation gauche gauche) is
another good example showing that RESP charge derivation can be tricky... Why
deciding that Methyl II is 'the' group II and not the group IV and why deciding
that O3' is O3' and not O5'? Indeed, the charge differences between the group II
and group IV is exactly the ones induced by the orientation of the 2 methyls in
the conformation gauche gauche. This means that there is an incertainty
concerning the charge of atom O3' and O5'. What is the difference in charge
obtained in this case for the 2 methyl carbons ? 0.016 This is smaller than
0.043 but however means that molecular orientation represents an incertainty of
20 % of the charge value. Such difference may not alter MM simulation but the
orientation is anyway responsible of an charge incertainty of 20 %

Thus our goals in developping RED were:
- To solve "inaccuracy problems" related to the RESP ESP charge derivation
protocole as those described above.
- To derive highly reproducible RESP and ESP charges using Gaussian BUT ALSO
GAMESS, i.e. WHATEVER the ab intio software is.
- To provide an automatic way to derive RESP (and ESP charges).
- To define a set of charges for a particular orientation allowing to reproduce
published values and checking errors.

Best regards, Francois

F.-Y. Dupradeau