From: Mayne, Christopher G (
Date: Wed Apr 17 2013 - 12:51:05 CDT

The MM values will never match the QM values. The idea is to compromise so that the charges are not too far from QM while performing as well as possible in an MM simulation. It is not always necessary to use all three files, based on the 3D topology of your molecule and other symmetry arguments. You should try to use the target data from at least one of the 120 positions, in addition to the 180 position.

Christopher Mayne

On Apr 17, 2013, at 12:46 PM, Wenhao Liu wrote:

I read the CGenFF paper and I think it didn't discuss much about carbonyl group. So you mean the reason why ffTK generate three water-molecule complexes for oxygen in carbonyl group is to let MM values approach QM values, right? So when I optimize the charges in ffTK for the moelcule contain carbonyl group, I should use all these three(120a, 120b, 180) gaussian log files for carbonyl oxygen in my QM target data, right?


On Wed, Apr 17, 2013 at 1:26 PM, Mayne, Christopher G <<>> wrote:
Carbonyls are a special case in ffTK. The optimal water positioning in QM is 120 degrees (where the electron density is located. In MM, the force field uses a point charge model that doesn't account for electronic structure, hence, the optimal water position is 180 degrees to reduce steric interaction with adjacent bound atoms. In this special case, ffTK generates input files for BOTH 120 and 180 degree positioning of waters, as per the suggestion made either on the CHARMM forums or in the original CGenFF paper (I don't remember where off the top of my head). This essentially hedges against errors due to discrepancies between QM and MM theories.

Christopher Mayne

PS Wenhao, we should keep this discussion on-list

On Apr 17, 2013, at 12:18 PM, Wenhao Liu wrote:

I see your point about charge optimizer. Thanks. So the plugin can generate more than one acceptor or donor bonding with water molecules for the atoms in my new molecule(like oxygen) if it can form more than one bond, right? I think this should be right since I've just tried and find that the plugin generated three different water bonding structures for the oxygen in Carboxyl group and that make sense.


On Wed, Apr 17, 2013 at 10:52 AM, Mayne, Christopher G <<>> wrote:
I don't think so. The optimizer judges the ability of a charge set to reproduce the interaction, but the positioning of the water is predetermined based on geometry, i.e., not a stochastic calculation. The best time to modify your weights is when you want to bias the charge sets towards getting a particular interaction correct (relative to qm).

On Apr 17, 2013, at 9:18 AM, "Wenhao Liu" <<>> wrote:

So you mean I can set the weight like to 2 when I deal with oxygen-water bonding in the condition that oxygen in my new molecule can form two hydrogen bonds with water? Thank you!


在 2013年4月17日星期三,Mayne, Christopher G 写道:
It is not required due to the symmetry of water--there is no extra information to be gained by flipping the identities of the hydrogens. ffTK handles the positioning of the water automatically.

Christopher Mayne

On Apr 17, 2013, at 1:53 AM, Wenhao Liu wrote:

> Hello vmd people,
> I was just wondering is it required to set up more than 1 hydrogen bonding for Acceptor (like the oxygen that can bonding with 2 hydrogen of the water) when I need to do the water interaction Gaussian calculation in charge optimization step? But I can only have 1 possible conformation to put water molecule bonding with oxygen in Acceptor indices, the other possible hydrogen boding between oxygen and hydrogen of water should be in different orientation, right? Or I can just set the weight of this calculation to 2 when I do the charge optimization? Thanks in advance.
> Wenhao