Re: [AMBER] Calculate charges using better method/basis set?

From: Karl N. Kirschner <kkirsch.scai.fraunhofer.de>
Date: Wed, 16 Nov 2011 16:53:13 +0100 (CET)

Hi Farid,

  Determining partial atomic charges are a tricky thing since they are not an experimental observable. Because they can not be directly compared to an experimental observables, one can not explicitly say that one method is more correct than another. What can be done is to show how well they reproduce a given QM electrostatic potential using different approaches (e.g. including aliphatic hydrogen in the fit, not including the hydrogens, different restraint function) - this is often done in AMBER Resp calculations.

Another consideration point is how reproducible are the partial atomic charges. This is addressed in the
paper by F.-Y. Dupradeau and coworkers (Phys. Chem. Chem. Phys., 12 (2011) 7821-7839) and their development work of the R.E.D. program.

Instead, I would use the language that one set of determined partial atomic charges are better (or worse) at modelling a given experimental observable. For charges that will be used in gas-phase calculations, one might compare how well the molecular mechanic's dipole moment reproduces the experimental gas-phase value. (But note, that the MM dipole's accuracy is coupled to how well the molecular geometry is also reproduced.)

Furthermore, experimental observables (e.g. diffusion rate) that arise from nonbonded interactions will also depend on the model's Lennard-Jones parameters. For condensed-phase simulations, the goal is to find a proper balance between the partial atomic charges and Lennard-Jones parameters. Because of this, any new nonbonded parameters that are developed for use with an established force field need to be done in the methodology that was used in that force field's original optimization - otherwise the resulting forces become unbalanced.

Because of the coupling between partial atomic charges and Lennard-Jones parameters, some force fields use atom-type-dependent partial atomic charges. Jorgensen's original OPLS parameters are an example of this. A more recent example is Ilja Siepman's TraPPE united-atom force field.

Hope this helps some.

Cheers,
Karl

------------------------------------
Karl N. Kirschner, Ph.D.
Fraunhofer-Institute for Algorithms
    and Scientific Computing - SCAI
Department of Simulation Engineering
Schloss Birlinghoven
53754 Sankt Augustin, Germany
Tel: +49 (0) 2241-14-2052
Fax: +49 (0) 2241-14-1328
------------------------------------

----- Original Message -----
From: "Mohd F. Ismail" <farid.ou.edu>
To: "AMBER Mailing List" <amber.ambermd.org>
Sent: Wednesday, November 16, 2011 4:10:34 PM
Subject: Re: [AMBER] Calculate charges using better method/basis set?

Dr. Walker,

Looking back at the original GAFF paper (http://onlinelibrary.wiley.com/doi/10.1002/jcc.20035/abstract) it seems that all the force constants were developed using higher level methods (MP2 and MP4) with appropriate basis set (polarization, diffuse function for appropriate molecule). For charges, only HF with 6-31G* was used for all calculation. The reasoning given (as you mentioned, the dipole value is overestimated) seems to come from the origin RESP paper (http://pubs.acs.org/doi/abs/10.1021/j100142a004) paper. What I get from this is that the force constants and the charges are not coupled.

Besides, overestimating the dipole is the reason to use HF with 6-31G* might be OK for protein since it is a solid. But for a bulk solvent, is not better to match the dielectric constant since the 'correct' dipole moment should produce the correct dielectric constant? Or am I totally missing your argument here?

I get that the RESP fitted charges would come out different with HF and MP2. But how would one know that the HF fitted charges is more correct one?

--Farid

*******************************
Mohd Farid Ismail
Graduate Assistant
Dept. of Chemistry/Biochemistry
University of Oklahoma
Norman 73019

________________________________________
From: Ross Walker [ross.rosswalker.co.uk]
Sent: Monday, November 14, 2011 12:36 PM
To: vvchaban.gmail.com; 'AMBER Mailing List'
Subject: Re: [AMBER] Calculate charges using better method/basis set?

> If we assume that both HF and MP2 provide an equally good [optimized]
> geometries of the system (molecule), than the trick looks painless...

The optimized structure, yes. But I would expect the actual electrostatic
potential and thus the RESP fitted charges to come out different with MP2 vs
HF. The original choice of HF was because it underestimates the dipole
moment in gas phase by approximately the same amount as the dipole moment
changes when you solvate the molecule. This, I assume, will not be the case
with MP2. Thus using MP2 is likely to have a fairly large effect on the
electrostatic interactions outside of what the force field was parameterized
for.

FF03 changed the charge fitting scheme to include an implicit solvent term
to generate solution phase charges from gas phase calculations instead of
relying on the cancellation of errors that the HF approach uses. Hence
improving the gas phase representation of the charges in the RESP fit does
not necessarily correlate with better results for solution phase
calculations.

All the best
Ross

/\
\/
|\oss Walker

---------------------------------------------------------
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
| http://www.rosswalker.co.uk | http://www.wmd-lab.org/ |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
---------------------------------------------------------

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.





_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Nov 16 2011 - 08:00:02 PST
Custom Search