AMBER Archive (2003)

Subject: RE: Why the NMA discrepancies?

From: Jones de Andrade (johannes_at_titus.phy.qub.ac.uk)
Date: Mon Mar 17 2003 - 08:35:05 CST

> Can you be more specific?
> Could the difference you observed be the consequence of different
> solvation environment? The reason for using RESP charges is that they
> mimic condensed phase electrostatics, rather than typical gas phase. I

        I don't feel like that. The original references os the resp
method mentions that the resp charges were developed to overcame problems
on ESP statistics for strongly buried atoms and different configurations.
To overcame the gas phase charges the idea is to use calculations on the
level of HF/6-31G(d), which will always overestimate the dipole moment of
the molecule in about 10%, then really "mimicing" the condensed phase
electrostatics.

> would anticipate difference if your QM NMA is done in gas phase.

        Also, what's is strange is that I've compared perfectly my NMA
results with the ones from a paper by P. Kollman on JPCB (I think it was
1998... I can look forward for the reference later) and the published
charges. The aim of that paper was to develop parameters for new solvents,
and the NMA of the CCl4 (or CHCl3, I'm not quite sure now) and DMSO
reproduced the QM scalled results. And I could reproduce those results
with this program, what would mean that there should not be any problem on
it. ;-)

> Nevertheless, I would be interested in knowing the detail of your
> comparison.

        Step-by-step: QM minimization of the benzene molecule. QM
calculation of the ESP grid. RESP charges calculation from the ESP grid.
QM NMA calculation. minimization by MM of the benzene molecule atomic
coordinates with the amber force field parameters. MM NMA calculation of
the normal modes from the MM minimized coordinates and amber force field
(also combined with the calculated amber charges). compare the MM NMA
results with the scalled (by a magic factor of 0.8953 developed by Pople
et al and also used in the amber parametrization process) QM NMA results.

        Many molecules work perfectly, some don't. BUT what called my
attention is that the aromatic parameters doesn't work perfectly for the
most simple aromatic molecule: benzene. Were those parameters not
developed using the benzene molecule as the model molecule?

        Thanks for everything,

                                                        Jones