AMBER Archive (2006)

Subject: RE: AMBER: Semi-empirical parameterization -- van der Waals and H-bonding?

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Mon Apr 03 2006 - 18:53:06 CDT

Dear Evan,

> I am beginning a QM/MM study on a drug-protein interaction.
> Hydrogen
> bonding and van der Waals forces are expected to be very
> important in
> this system. My understanding is that AMBER has a QM/MM function
> built into the SANDER module which treats the QM portion semi-
> empirically.

The first piece of advice I will give you is that if you are planning to do
QM/MM studies with Amber you should seriously consider upgrading to Amber v9
which was released on March 29th 2006. This version has vastly improved
QM/MM support over version 8. Including huge speed improvements (>10x), and
support for periodic boundary and implicit solvent GB simulations. You can
find details on the amber website (http://amber.scripps.edu).

I will be adding a tutorial on QM/MM with Amber 9 to my begginning amber
workshop pages (http://www.rosswalker.co.uk/tutorials/amber_workshop/)
shortly.

> My question is then: does anyone know about the
> parameterization of the semi-empirical methods?

The supported methods in Amber 9 are: PM3, AM1, MNDO, PM3CARB1, PM3/PDDG,
MNDO/PDDG, MNDO-D (gas phase/solvent cap only via divcon) and SCC-DFTB(gas
phase/solvent cap only)

If you check page 142 of the Amber 9 manual:
http://amber.scripps.edu/doc9/amber9.pdf you will find all of the
semi-empirical hamiltonians listed along with references that give details
of their parameterisation.

The QM to MM interface is the same as that used by the Dynamo code and
details can be found here:

Field, M.J., et al., J Comp Chem., 2000, 21, 1088

You should also refer to the papers referenced in Section 6.4 of the Amber 9
user manual.

A paper detailing the exact implementation of the various methods (QM-MM
interface, link atom treatment, GB, PME, derivative accuracy etc.) will be
published shortly. If you contact me privately I can possibly send you a
draft of this paper.

> Mainly, are they
> trained to handle H-bonding and van der Waals forces? Have there
> been any studies done on AMBER's ability to treat QM/MM
> H-bonding, etc?

The current Amber implementation has not been specifically tuned for
handling H bonding and VDW interactions. Within the QM region the hydrogen
bond interaction will be the same as that of the semi-empirical hamiltonian
you are using, similarly the VDW interaction is purely quantum within the QM
region and so will match that of the semi-empirical hamiltonian. Within the
MM region the interactions are all classical and so H bonding and VDW will
be the same as that of the classical force field you use. The remaining part
is the QM/MM interaction. VDW interactions between MM atoms and QM atoms are
calculated classically using the selected classical force field parameters.
Away from any covalent QM/MM boundary this should give VDW interactions that
are as good as the classical VDW interactions. Potentially one could improve
the results by sepcifically parameterising for QM/MM interactions but this
would involve a lot of extra work for questionable benefit. Around QM/MM
link atoms the interactions may be over estimated due to the inclusion of
bonds, angles and dihedrals between the QM and MM region as well as VDW
interaction between those atoms. If you have a suitably sized QM region,
however, and minimise the number of link atoms this effect is expected to be
minimal. Although you will have to test and calibrate it based on your
specific system.

H-bonding between the QM/MM region has not been extensively tested. This is
the subject of ongoing tests. The issue is that if you have part of the
hydrogen bond pair in the MM region it is not polarised (since it simply has
a resp point charge) when the H-bond is formed while the part that is in the
QM region can polarise... This mismatch in the electrostatics 'may' cause
problems... So, for MM-MM hydrogen bond interactions you can expect the same
behaviour as classical simulations. QM-QM hydrogen bonding should be fairly
good since both parts of the bond can be polarised. Try a literature search
concerning hydrogen bonding with the various semi-empirical hamiltonians.
QM-MM hydrogen bond interactions will require some testing, try looking at
some small systems that you can treat both purely quantum, purely classical
and QM/MM. Then you can compare the results.

Note, with Amber 9 you can do explicit solvent periodic boundary simulations
with PME which can have a huge effect on the system stability and the
quality of the results you obtain, including H-bonding effects. Similarly
you can do implicit solvent GB simulations as well. The benefits of these
two methods are, in my opinion, so pronounced that you should NOT consider
doing gas phase or solvent cap QM/MM simulations except only for
comparisson.

I hope this helps. If you want any more information or help please contact
me directly.

All the best
Ross

/\
\/
|\oss Walker

| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
| http://www.rosswalker.co.uk | PGP Key available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

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