AMBER Archive (2008)

Subject: AMBER: FW: AMBER vs Gaussian : zero VDW parameters?

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Fri Jan 04 2008 - 11:01:14 CST

 
Hi Yi,

I am forwarding your message to the AMBER mailing list so that others can
comment as well. You should subscribe (see http://amber.scripps.edu) so you
can see any replies.

My understanding is that only water hydrogens (OPLS and TIP3P) have zero VDW
radius and this is because they were only ever designed to be used as
triangulated water with Shake. I believe that Oxygen radius and the
repulsion between two oxygen atoms is sufficient to avoid the situation of
having a hydrogen of one water ending up on top of a oxygen of a second
water. Although I have heard of people having problems with phosphate
groups.

Note I believe that in Charmm they put a very small VDW radius on the water
hydrogens to prevent problems.

So my first assumption would be that it is related to not using shake in the
Gaussian Onion calculation so you'll need to work out if Gaussian supports
this and or add it if you are using waters in the calculation.

Perhaps others on the list can give a more definitive answer though.

All the best
Ross

-----Original Message-----
From: Yi Mao [mailto:maoyi0_at_gmail.com]
Sent: Friday, January 04, 2008 08:34
To: ross_at_rosswalker.co.uk
Subject: AMBER vs Gaussian : zero VDW parameters?

Recently I have been using amber and gaussian (oniom) to study a
complex of a protein (made of standard amino acid, no metal atoms) and
an organic molecule. There is no covalent bonding between the protein
and the organic part. I first performed MD simulations on the complex
by AMBER, and things went fine.

Since I am interested in the excited spectrum of the organic molecule,
QM/MM is the method of choice. And currently QM/MM in amber cannot
deal with excited state calculations, so I switched to ONIOM (part of
g03). The protein is chosen as the low layer and treated by AMBER and
the organic molecule is treated by HF. The interesting thing is that
even the initial coordinates and force parameters are correct for the
system, the optimization process always generates infinite forces on
the MM part sooner or later. It turned out that this is a quite
common problem that ONIOM users (who use AMBER as part of the layer)
encounter. I have consulted the ONIOM expert at gaussian inc, and his
reply says

" My best guess is that two atoms collapse onto each other. Usually this
is prevented from happening by the (repulsive) vanderwaals interaction
or bonded interactions. But there are some atom types in the Amber
forcefield (such as HO) that have zero VDW parameters. If such an atom
gets close to an atom with opposite charge, there is only attraction,
which will become infinite when the distance between them becomes zero.

This is a problem that we see quite often. There is not much we can do
about it, because we implemented the Amber forcefield exactly according
to the original publications. You can in your calculations, however,
give those centers a small VDW term in order to prevent the collapse."

His explanation agreed with my findings that the addition or omission
of some hydrogen atoms in the g03 input can have a big difference on
the outcome. However the natural question to ask is that if what's
stated above is true, how come AMBER never have similar problems? From
my understanding of MD, if the force field is the same, MD results
from different algorithms/programs have to agree to a significant
degree. I would like to hear the MD expert's take on this issue. (the
amber force at g03 is based on parm96.dat )

Thanks!

Yi

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu