Date: Tue, 28 Jan 2003 10:35:14 -0500
From: "Bayly, Christopher"
Subject: RE: Intramolecular H-bonds in charge fitting


For charge-fitting using RESP, intramolecular H-bonds are a much smaller
issue than with AM1-BCC charges. With RESP, charges are fit to the actual
electrostatic potential "seen" by the grid of points just outside the vdW
surface, so charges on atoms involved in an intramolecular H-bond will only
reflect peculiarities "seen" at that distance outside the molecule. You
may, however, have some numerical stability problems because H-bonds bury
the atoms involved away from the vdW surface.

recommendation for RESP: for numerical stability reasons, use the RESP
"recipe" of fitting charges simultaneously to at least two conformations,
and let the other conformation not have that intramolecular H-bond.


In contrast, AM1-BCC charges "correct" AM1 atomic charges to ones that
should perform similarly to RESP charges, but it assumes the AM1 atomic
charges exhibit "typical" behaviour. These AM1 charges, coming as they do
directly from the AM1 wavefunction, exhibit very atypical behaviour in an
intramolecular H-bond... they are much more polarized than usual. The
additive BCC corrections then result in significantly overpolarized charges
compared to what RESP would give. Even then, the performance might be OK as
long as the intramolecular H-bond doesn't break in your simulation, but if
it does, the over-polarized charges will be interacting too strongly with
water (or whatever else).

Recommendation for AM1-BCC: fit charges to a conformation that does not have
the intramolecular H-bond. It's still a lot less CPU time and human time
than RESP.


I hope this clarifies the issue somewhat... good luck.

Christopher



-----Original Message-----
From: Lepsa
Sent: Tuesday, January 28, 2003 4:06 AM
To: amber@heimdal.compchem.ucsf.edu
Subject: Intramolecular H-bonds in charge fitting


Dear AMBER community,
I have some questions on charge derivation. Does it matter for RESP charge
fitting of QM ESP that intramolecular H-bonds are present in the molecule? I
know that for AM1 derived charges these H-bonds must be avoided.

What influence does it make whether these bonds are present/absent from the
crystal structure/MD? What
change is there for this case in using B3LYP derivation scheme for
polarizable FF?

Who could say anything relavant to this is warmly welcome.

Thanks for your replies.

Regards,

Martin Lepsik