[gmx-users] Ligand Parameters

David Mobley dmobley at gmail.com
Tue May 15 00:31:43 CEST 2007


Justin,

> I am interested in simulating some small molecules that involve oxygen in
> aromatic rings (which thus bears a +1 charge).  I tried to submit one of these
> molecules to PRODRG, but got an error indicating that there were too many bonds
> to O.  I thought this issue might arise, due to the unusual condition of the
> oxygen (+1 charge, hybridization, etc.)  I have looked into the list to see if
> anyone had come across a similar problem, but I can't find an explicit example.
>  I realize this odd case is probably limited by the developed force field
> parameters, which don't appear to include O+.  I have checked the *.atp entries
> to confirm this.  I read on the archive that it might be possible to develop
> parameters using the ffamber ports (which I have) and using ANTECHAMBER (which
> I have used) to get partial charges.  I would then have to create some entries
> in the applicable .rtp file, etc, and otherwise do a lot of reading in Chapter
> 5 of the manual :-)

Are you committed to a particular force field? For example, with
PRODRG, you are stuck with GROMOS or whatever else it can output
(hence, you would need to use the same force field for other elements
in the system).. Also, the charges from there would have, (ahem),
possible issues even if it would generate parameters for your
molecule. For example, their FAQ page says,

"Q: The atomic charges don't always come out right? A: This is a well
known limitation. PRODRG works with the concept of charge groups,
which are defined as a group of bonded atoms with an integer charge.
E.g. -COO would have the C (with a slight positive charge) and the two
O's (with an equal negative charge) in one "charge group" with the
total charge being -1. Thus in order for PRODRG to assign atomic
charges it needs to be able to recognise the charge groups first. This
is particularly difficult in ring systems with nitrogens, and several
charge group definitions (e.g. C-Cl) or still missing. I would
definitely recommend that you check these charges before starting your
2 microsecond simulation."

In other words, you can't count on the charges to be right (for
example, this charge groups concept is rather questionable).

As far as having worked with O+, I haven't. You could try running it
through antechamber (specifying the -nc flag for your molecule) and
see if it actually can generate parameters for it... If you can
parameterize it with Antechamber, you can use our conversion script
(http://www.alchemistry.org/FreeEnergyTools) to convert those to
GROMACS format (for use with the AMBER ports to GROMACS). But I can't
vouch for how Antechamber will work for something with O+.

> I realize this problem is not trivial, and I don't expect an easy solution, but
> I was wondering if anyone out there had worked with O+ in their molecules, and
> if they had any advice or topology building blocks that I might use as a
> template, seeing as none have been contributed online.

Have you searched the literature? If there aren't O+ parameters in
current force fields, I suppose you'll have to derive them yourself in
a manner consistent with the force field you plan on using for the
rest of your system (i.e. possibly from QM).

> I have also considered submitting an analog of my compound to PRODRG (normal
> aromatic ring instead of one containing O+) to get the dihedrals, etc. and then
> re-calculating charges under ANTECHAMBER.  Any thoughts on that approach?

Presumably dihedrals, etc are affected by the atom types,
hybridization, etc. Sounds like a bad idea. If I were a reviewer on a
paper that did this, I'd probably think it was an even worse way to
parameterize something than the "make the parameters up based on
chemical intuition" approach.

The way to do this: Figure out what force field you want to use; look
up how existing parameters for that force field were derived, and then
do something consistent with that to get your new parameters. The
procedure will be different depending  on the force field you want to
use.

Best wishes,
David



More information about the gromacs.org_gmx-users mailing list