RE: AMBER: RESP vs. AM1-BCC charges for isoprene

From: <ajakalian.lav.Boehringer-Ingelheim.com>
Date: Mon, 26 Jul 2004 11:09:31 -0400

Hello Oliver,

The quick-and-dirty answer to your first question is: if AM1-BCC can
correctly charge-up a molecule (i.e. no BCCs are missing for that molecule)
then that molecule is suitable...

The slightly longer answer is: to verify the quality (suitability) of the
AM1-BCC charges for a specific molecule, you may want to calculate the
HF/6-31G* ESP and compare it to the ESP calculated from the AM1-BCC method.
This will give you a quantitative measure of the suitability of the AM1-BCC
method for that specific molecule.

The global parameterization of the 354 BCCs is published in the
"parameterization and validation" paper (JCC, 23(16), 1623, 2002). It is
already included in Amber7 and Amber8 distributions within the antechamber
module.

Take care,

Araz

-----Original Message-----
From: Oliver Hucke [mailto:ohucke.u.washington.edu]
Sent: Friday, July 23, 2004 1:41 PM
To: amber.scripps.edu
Subject: Re: AMBER: RESP vs. AM1-BCC charges for isoprene


Hi Araz,

thanks very much for your helpful reply.
Of course you are right saying "It does not make sense to design a
charge model based on another charge model."
What I actually meant was that the AM1-BCC method was designed to
reproduce the HF/6-31G* esp as the resp fit procedure was. Sorry for not
being precise.

I would like to add the following questions:

o Is there a way to quickly check if the training set used for the
AM1-BCC parameterization is suitable for a specific molecule?

o Is there a publicly accessible repository for the most recent
parameterization of the AM1-BCC method (In your paper you wrote that you
had begun the global reparamaterization of the charge model). Or is the
globally parameterized method already part of the amber8
distribution?

Best regards,
Oliver

ajakalian.lav.Boehringer-Ingelheim.com wrote:
> Hello Oliver,
>
> AM1-BCC was NOT designed to reproduce RESP charges.
>
> It does not make sense to design a charge model based on another
> charge
> model. We designed the AM1-BCC model to reproduce the QM electrostatic
> potential (ESP) at the HF/6-31G* level of theory. This is the same
> fundamental idea behind RESP, however unlike RESP, AM1-BCC does not use
> HF level of theory but rather the AM1 level to get a first approximation
> of the QM ESP with the Coulson charges. These first-pass charges are
> then corrected with bond-charge corrections (BCCs) in order to reproduce
> the QM ESP.
>
> That said, I would not worry about large differences between AM1-BCC
> and
> RESP for carbon atoms. It is well known that RESP charges suffer from
> numerical instabilities, especially for buried atoms (i.e. charges can
> vary widely while the RMS of the fit varies little). Table VI in the
> AM1-BCC method paper shows this perfectly. Look at the RESP charges on
> Csp3 atoms, they vary between -0.08 and +0.07, while for the same atoms
> the AM1-BCC charges vary between +0.09 to +0.1.
>
> Hope this helps,
>
> Araz
>
> ****************************
> Araz Jakalian, Ph.D.
> Research Scientist, Structural Research Group
> Boehringer Ingelheim (Canada) Ltd.
> Research and Development
> 2100 Rue Cunard
> Laval, Quebec, H7S 2G5, Canada
> Phone: (450) 682-4640 Fax: (450) 682-4189
> E-mail: ajakalian.lav.boehringer-ingelheim.com
> ****************************
>

-- 
_______________________________________________________________
Oliver Hucke, Dr.
                               Health Sciences Building - K418C
University of Washington      1959 NE Pacific St.
Dept. of Biochemistry         phone: (206) 685 7046
Box 357742                    fax  : (206) 685 7002
Seattle, WA 98195-7742        email: ohucke.u.washington.edu
_______________________________________________________________
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Mon Jul 26 2004 - 16:53:00 PDT
Custom Search