From:	Howard Alper <halper %! at !% uno.edu>
Date:	Mon, 12 Oct 1998 15:06:13 -0500
Subject:	Parameterization - summary

Hello,

  Several weeks back I asked this list for help with the process of
parameterizing a model of Nitromethane for molecular dynamics
simulation.  Many people on the list responded, with much helpful
information as to how I should proceed (being a total novice in
parameter development).  Thanks to their help, and some off-line
conversations, I am now happy immersed in the parameterization process,
and will hopefully converge to reasonable parameters for Nitromethane
in the near future.  Many thanks to all who responded.  The summary is
below.

  Howard

____________________________________________________________________

  My original post was the following:

Hello,

  I'm sorry if this is a repeat, but as far as I could tell this message
did not post when I originally sent it - twice.

  Several weeks back I wrote about parameterization and units in
simulation (still waiting for some replies...)  Now I have some
questions about parameterization to create forcefields for MD
simulations.

  Specifically, I am interested in parameterizing Nitromethane.
While any information regarding a set of parameters (bond, angle,
torsion, out-of-plane,charges,van der Waals, etc) would be welcome,
we are interested in generating the neccesary parameters, and in a form
suitable for use in CHARMM.  To this end I have several questions:

1) I need to derive, somehow, structural parameters and force constants
for bond-stretching and angle-binding.  The force constants should
be in units of energy/(angstroms**2) and energy/(radian**2) (I have
seen some force constants in old papers that give force constants
in units of millidynes/angstroms, but I do not understand how that
relates to the case of angle-bending constants...)  I can perform
an optimization and frequency calculation in Gaussian, for example, but
from what I can tell the resulting frequencies and force constants
are for the normal modes of the molecule, not bonds and angles/etc.
Does anyone know if Gaussian has any options to derive the type of
force constants I require?  Or, does anyone know of a post-Gaussian
analysis program that would do the above?

2) I also need to derive van der Waals parameters.  One person I asked
said something on the order of "you can derive that from the interaction
of two nitromethane molecules".  To a parameterization novice, that is
not very specific advice.  Could anyone clarify what in detail needs to
be done, provide references, describe how programs like Guassian might
be used to get VDW parameters, etc?  (If one calculates the interaction
energy of two nitromethane molecules at verious distances and
orientations, is not the total energy for a configuration the sum
of coulombic and VDW terms?  Does that mean the charges must first
be derived?)

  Thanks in advance,

  Howard Alper
____________________________________________________________________


  And now your responses:

____________________________________________________________________


From: Craig Burkhart 


Hello Howard,

Some ideas, etc on your request:

Howard Alper wrote:

> 1) I need to derive, somehow, structural parameters and force constants
> for bond-stretching and angle-binding.  The force constants should
> be in units of energy/(angstroms**2) and energy/(radian**2) (I have
> seen some force constants in old papers that give force constants
> in units of millidynes/angstroms, but I do not understand how that
> relates to the case of angle-bending constants...)  I can perform
> an optimization and frequency calculation in Gaussian, for example, but
> from what I can tell the resulting frequencies and force constants
> are for the normal modes of the molecule, not bonds and angles/etc.
> Does anyone know if Gaussian has any options to derive the type of
> force constants I require?  Or, does anyone know of a post-Gaussian
> analysis program that would do the above?

The conversion from mdyn/Ang to kcal/mole/Ang^2 is a throwback to the
old days
when forcefields were used for spectroscopic purposes (see the Allinger
MMx
implementations for the conversions).

Ab initio-based parameterization is growing in importance, largely due
to
Allinger's and Hagler's groups work. Hagler, Hwang and Stockfish wrote a
nice
article, circa 1994, on using ab initio methods to parameterize their
newest 2nd
generation forcefield. If my memory is correct, they used
small random displacements in bonds, angles and torsions to parameterize
their
forcefield for the valence terms.

> 2) I also need to derive van der Waals parameters.  One person I asked
> said something on the order of "you can derive that from the interaction
> of two nitromethane molecules".  To a parameterization novice, that is
> not very specific advice.  Could anyone clarify what in detail needs to
> be done, provide references, describe how programs like Guassian might
> be used to get VDW parameters, etc?  (If one calculates the interaction
> energy of two nitromethane molecules at verious distances and
> orientations, is not the total energy for a configuration the sum
> of coulombic and VDW terms?  Does that mean the charges must first
> be derived?)

This part is a bit more tricky. How you do the nonbonded interactions is
going to
depend upon your usage. If you are interested in ideal gas-phase
simulations,
then maybe you only have to do a set of random configurations with a
pair of
molecules (perfect pairwise interactions). If you want to do more
concentrated
systems, then you have to worry about how to incorporate the many-body
interactions (higher order virial terms) into your forcefields. In
liquids, for
example, the parameterization is performed against thermodynamic data
such as
heats of vaporization, and so on. Jorgensen's group has used this
approach with a
great deal of success in their OPLS implementations.

The long and the short of this is: if you want to do ideal gas phase
simulations
of nitromethane, then a correlated ab initio method (Muller-Plesset, for
example)
is probably going to be necessary if you want really high quality
pairwise
parameters. You can use Hartree-Fock if you want lesser quality, as the
dispersion interaction will not be handled adequately.  And yes, you
will need to
obtain the partial charges separately, usually by electrostatic
potential-derived
charges. Even here, you must be careful.

If you want to do liquids, then using ELP in a continuum solvent field
would be
indicated on initial inspection, as the liquid state will polarize your
molecule--causing the partial charges to increase on each atom (CHARMM
is a
nonpolarizable forcefield). Most people ignore this complication, as you
would
have to have a table or "nomograph" of partial atomic charges for each
atom type
as a function of dielectric constant and molecular environment (Karasawa
and
Goddard did some nice work in this area about 10 years ago using a
variation of
charge equilibration via a classical grand canonical ensemble. They
could update
the charges on-the-fly). You will also need to use bulk thermodynamic
data to
create "effective" nonbonded van der Waals parameters.

I don't know if Bill Jorgensen would release his parameterization codes,
but I
believe that Jay Ponder's Tinker suite has that capability--and the last
time I
checked, his software is freely obtainable from his website (
http://dasher.wustl.edu).

Hope this helps...

Craig


____________________________________________________________________

  From: ulf;at;hugin.teokem.lu.se (Ulf Ryde)
     To:

Dear Dr. Alper,

I have implemented a simple version of the procedure suggested by
Seminario (Int. J. Quantum Chem.: Quantum. Chem. Symp. 30(96)59-65).
It converts the Hessian matrix produced by the Gaussian program to
internal force constants and it is said to be invariant to the choice
of internal coordinates. I find this procedure useful for a rapid
determination of force constants. Please note, however, that this is
only an approximate procedure since it ignors the interactions between
parameters of different types and it is not the method used in
standard parameterisations. For more accurate methods, you should look
in the original papers of the MMFF, Amber, CHARMM, etc. force fields
and references therein, e.g .
Cornell, et al, J. Am. Chem. Soc. 117(95)5179
Halgren, J. Comput. Chem.  17(96)490-641
Paverlites et al, J. Comput. Chem. 18(97)221-239.

Good Luck,
Ulf
____________________________________________________________________

  From: Jaromir MAREK 


> Hello,
>
>   Specifically, I am interested in parameterizing Nitromethane.
> While any information regarding a set of parameters (bond, angle,
> torsion, out-of-plane,charges,van der Waals, etc) would be welcome,
> we are interested in generating the neccesary parameters, and in a form
> suitable for use in CHARMM.  To this end I have several questions:

Hi  Howard !

Some general informations are at AMBER WWW pages

http://www.amber.ucsf.edu/amber/amber.html

and

http://www.amber.ucsf.edu/amber/newparams.html

Regards,

Jaromir
____________________________________________________________________

  From: "Donald E. Williams" 

Check my home page www.louisville.edu/~dewill01.  Program nbp will
accept dimer
data for optimization of the force field.

--
Dr. Donald E. Williams          email:dew01 { *at * } xray5.chem.louisville.edu
Department of Chemistry
University of Louisville        phone:502-852-5975
Louisville, KY 40292            fax:  502-852-8149

____________________________________________________________________

  From: ss <-at-> CLEMSON.EDU (Steven J. Stuart)

Howard,

Some direct & indirect answers to your CCL post questions:

1) I don't have any specific numbers for you.  However, if you get
normal
mode frequencies from Gaussian, it will be possible to turn those into
bond
stretches and angle bends.  I can't answer whether Gaussian has options
to
do this or whether there are postprocessing tools to do it for you.  But
as
a worst-case scenario, you could do the calculation by hand.  Tedious,
but
possible.  See Wilson, Decius, and Cross "Molecular Vibrations" for
details.  Nitromethane is small enough that it might even be worked out
for
you.

2) LJ parametrization is more my forte.  One approach, that you
partially
described, would be to:  (a) evaluate interaction energy vs. distance
for 2
nitromethane molecules at fairly distant separations, for a given
orientation, using Gaussian or ; (b) select LJ parameters which
reproduce
the PE vs r curve.  In general, this will not completely specify the LJ
parameters for your 4 different atoms, and you will have to do this for
several different configurations.  Choose these either based on symmetry
or
based on selecting configurations such as H-bonding arrangements that
are
expected to contribute heavily in your applications.  Another very
common
approach is to steal some of your LJ constants from similar compounds.
The
LJ constants for C and H are not likely to be very different in
nitromethane and other substituted methane species.

You are right that you will need to know the partial charges before
doing
the LJ fit.  There are lots of ways to select these.  You will need to
reproduce the molecule's dipole moment (either in gas phase or in
solution,
depending on your particular application and the potential you are
combining it with).  This will not uniquely specify the charges, though.
Beyond that, you can optimize the charges to reproduce the electrostatic
potential around the molecule, or to give the correct dimer energies
when
optimized in conjunction with the LJ parameters, or any of several other
methods.

I suspect I haven't been as explicit as you'd like.  If so, ask more
questions and I'll try to provide more details where I can.

-Steve Stuart
ss (- at -) clemson.edu
____________________________________________________________________

Guido Germano 
Organization:
               Physics Department University of Bristol

You asked a 1 million $ question.

I have, perhaps, an answer to this curiosity of yours:

> I have seen some force constants in old papers that give force
> constants in units of millidynes/angstroms, but I do not understand
> how that relates to the case of angle-bending constants

You can treat bending as a (1,3) stretching.

More in general, the equilibrium bonds and angles can be easily found
with Gaussian; the force constants not, they must be guessed, but for
quick and dirty calculations it's not difficult, as the order of
magnitude of strecthing constants is always the same and ten times
higher than the bending ones, which are again more or less the same and
higher than the torsional ones, etc.

The hardest parameters to guess are the VdW ones. In principle you need
to do test MD runs with lots of molecules and calibrate physical
observables like the density by trial and error. I don't think ab initio
calculations on couples are enough, if the objects are not spherical.

Bottomline: if you find any new papers or algorithms to extract e.g.
stretching and bending constants in an automatic fashion from a normal
mode analysis, let me know (I suspect it's an ill-determined problem:
once you have a set of constants, you can check with the normal modes;
but I have never gone this much into detail).

Regards

Guido Germano
____________________________________________________________________

From: Ya-Jun Zheng 


Hi  Howard,

Since you are trying to get parameters to use in Charmm, you may not
need
very accurate force field parameters.  One simple approach is as follow:

  For internal parameters, you could take the ab initio bond lengthes
and
angles.  For the torsion parameters, you could use either experimental
or
ab initio (e.g., HF/6-31G*) values.  For charges, use ESP fitted
charges.
van der Waals parameters could be obtained by analogy (for example using
the normal carbonyl O parameters for the nitro O).  The bond and angle
force constants can be obtained by trying to reproduce the IR spectra of
CH3NO2 (normal model analysis in Charmm).
Unless you are trying to simulate nitromethane liquid, otherwise the
above
approach should give you reasonable parameters.  You could also use the
ab
initio CH3NO2---H2O interaction energy to refine your van der Waals
parameters.  Allinger and coworkers published MM3 (or MM2) parameters
few
years ago in J. Mol. Structure.  You could use some of these parameters
as
initial guess.

As for the two questions you asked, the simple answer is as follow:

 1.  Basicaly, you need a transformation program to do that.  There is
an
old program called G-matrix that some people use to do the
transformation.
I had a hardcopy of that program, but never had the chance to make it
work,

2. You could check some of the papers by former Biosym people (A. Hagler
and others), which gave detailed description of how to do
parameterization
from ab initio data.

Yajun Zheng

********************************************************
Dr. Ya-Jun Zheng
DuPont
Stine-Haskell Research Center
P. O. Box 30
Newark, DE 19714
E-mail: Ya-Jun.Zheng #*at*# usa.dupont.com
********************************************************
____________________________________________________________________

From: "Alexander D. MacKerell, Jr." 


Howard,

   If you want to determine parameters that are compatible with CHARMM
you should follow the same procedures used for that force field.  The
appropriate references are on my web page, along with the recent release
of charmm parameters.  Those papers include the use of normal modes in
the optimzation procedure (note that any normal mode is a combination of
individual internal coordinates, you just change the parameters
associated
with those coordinates to reproduce the desired frequency).

   Concerning the vdw parameters, i would use default values from CHARMM
as a
first (and quite reasonable) guess.  Using these optimize the rest of
your force
field.  When that is complete, then reconsider the quality of the vdw
params.
Note that optimization of the vdw params is, in my opinion, the most
difficult
aspect of parameter optimazation.

good luck, alex

--
Alex MacKerell, Ph.D.
Associate Professor
School of Pharmacy
University of Maryland
20 North Pine Street
Baltimore, MD  21201
410-706-7442
410-706-0346 (fax)

alex %-% at %-% mmiris.ab.umd.edu or
alex %-% at %-% rx.umaryland.edu
http://www.pharmacy.ab.umd.edu/~alex/
____________________________________________________________________

From: Wayne Steinmetz 


The frequencies of the vibrational normal modes can be expressed in
terms
of the local force constants which you require.  The details are given
in
the classic E. B. Wilson, J. C. Decius, and P. C. Cross,
Molecular Vibrations, McGraw-Hill (1955).  With some luck, a
moelcular spectroscopist has performed an exhaustive study of the Raman
and infra-red spectra of normal nitromethane and isotopic modifications
and has completed a normal coordinate analysis.  A complete study should
include the parameters that you require.  An example of what you might
find
can be found in T. Shimanouchi, H. Matsuura, Y. Ogawa, and I. Harada,
J. Physical Chemical Reference Data, Vol. 7 ,  pp. 1323-1443 (1978).
Unfortunately nitromethane is not in this paper but the paper that I
cited
is Part 9 of the series entitled Tables of Molecular Vibrational
Frequencies.
Some of Takehiko Shimanouchi's other papers in this series might contain
what you require.  Since nitromethane is a small molecule with no
internal rotation, I would be surprised that no one has beaten it to
death.
A source of chem abstract with keywords such as nitromethane and normal
coordinate analysis might lead the path to your data.
Another possible source of references is JANAF Thermochemical Tables.
The spectroscopic data that form the basis for the thermodynamic data
are provided that stat mech is the basis for the thermo.  I checked
my copy but nitromethane is not present but I have an older edition.
____________________________________________________________________

From: ross (- at -) cgl.ucsf.EDU


See

        http://www.amber.ucsf.edu/amber/newparams.html

probably all but the "improper torsions" part would
be a decent intro for your purposes.

Bill Ross
____________________________________________________________________

From: Rick Venable  deimos.cber.nih.gov>


On Wed, 16 Sep 1998, Howard Alper wrote:
> Now I have some questions about parameterization to create forcefields
> for MD simulations.  Specifically, I am interested in parameterizing
> Nitromethane.  While any information regarding a set of parameters
> (bond, angle, torsion, out-of-plane,charges,van der Waals, etc) would
> be welcome, we are interested in generating the neccesary parameters,
> and in a form suitable for use in CHARMM.  To this end I have several
> questions:

I've omitted the questions, because I'm not answering them directly, but
supplying some references from Alex MacKerell's web site at

        http://www.pharmacy.umab.edu/~alex/research.html

The papers below describe the development of various topology and
parameter sets which are distributed with the academic version of
CHARMM; additional citations and the parameter sets themselves (in
CHARMM format) are also available on the web page.  Typically, force
constants for bond stretching and angle bending are derived from
published IR/Raman spectroscopic data, after conversion to the CHARMM
system of units.  Atomic point charges and VDW radii parameters require
extensive ab initio calculations, while torsions are usually
parameterized to match experimentally determined rotational barriers
and/or conformational distributions.


MacKerell, Jr., A. D.; Bashford, D.; Bellott, M.; Dunbrack Jr., R.L.;
Evanseck, J.D.; Field, M.J.;  Fischer, S.; Gao, J.; Guo, H.; Ha, S.;
Joseph-McCarthy, D.; Kuchnir, L.; Kuczera, K.;  Lau, F.T.K.; Mattos, C.;
Michnick, S.; Ngo, T.; Nguyen, D.T.; Prodhom, B.; Reiher, III, W.E.;
Roux, B.;  Schlenkrich, M.; Smith, J.C.;  Stote, R.; Straub, J.;
Watanabe, M.; Wiorkiewicz-Kuczera, J.; Yin, D.; Karplus, M.  All-atom
empirical potential for molecular modeling and dynamics Studies of
proteins.  Journal of Physical Chemistry B, 1998, 102, 3586-3616.

Yin, D. and MacKerell, Jr. A.D. Combined Ab initio/Empirical Approach
for the Optimization of Lennard-Jones Parameters. Journal of
Computational Chemistry, 1998, 19:  334-338.

Pavelites, J.J., Bash, P.A., Gao, J. and MacKerell, Jr., A.D. A
Molecular Mechanics Force Field for NAD+, NADH, and the Pyrophosphate
Groups of Nucleotides. Journal of Computational Chemistry, (1997)
18:221-239.

A.D. MacKerell Jr., J. Wi-rkiewicz-Kuczera and M. Karplus, An All-Atom
Empirical Energy Function for the Simulation of Nucleic Acids, Journal
of the American Chemical Society (1995)  117:11946-11975.

--
Rick Venable                  =====\     |=|    "Eschew Obfuscation"
FDA/CBER Biophysics Lab       |____/     |=|
Bethesda, MD  U.S.A.          |   \    / |=|  ( Not an official
statement or
rvenable - at - deimos.cber.nih.gov  |    \  /  |=|    position of the FDA; for
that,
http://www.erols.com/rvenable
\/   |=|    see
http://www.fda.gov  )
____________________________________________________________________

From: Per-Ola Norrby 


Howard Alper wrote:

 .....

>  Specifically, I am interested in parameterizing Nitromethane.
>While any information regarding a set of parameters (bond, angle,
>torsion, out-of-plane,charges,van der Waals, etc) would be welcome,
>we are interested in generating the neccesary parameters, and in a form
>suitable for use in CHARMM....

 .....

        Dear Howard,

        We just published a method for parameterization, specifically
implemented for MacroModel, but we put a lot of effort into making the
routines general enough so that you should be able to use it in a
similar
manner with any MM program running under Unix.  Anything force field
specific is handled by Unix scripts, you only need to be able to handle
Unix to modify the method for the force field you are interested in.
The
method is described in J. Comput. Chem., 1998, 19, 1146-1166.  I'll
share
the routines (note! they are NOT user-friendly, and you definitely need
the
article to make sense of them).  Send me an email if you're interested.
The stuff I'm distributing right now is a year old, I'm right in the
middle
of putting together an updated version, should be ready in a couple of
weeks.'

        Best regards,

        Per-Ola Norrby


^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 *  Per-Ola Norrby, Associate Professor
 *  The Royal Danish School of Pharmacy, Dept. of Med. Chem.
 *  Universitetsparken 2, DK 2100 Copenhagen, Denmark
 *  tel. +45-35376777-506, +45-35370850    fax +45-35372209
 *  Internet: peon "-at-" medchem.dfh.dk, http://compchem.dfh.dk/
____________________________________________________________________

From: Konrad Hinsen 


>   Specifically, I am interested in parameterizing Nitromethane.
> While any information regarding a set of parameters (bond, angle,
> torsion, out-of-plane,charges,van der Waals, etc) would be welcome,
> we are interested in generating the neccesary parameters, and in a form
> suitable for use in CHARMM.  To this end I have several questions:

Have a look at

  21) K. Hinsen, B. Roux
      An accurate potential for simulating proton transfer in
acetylacetone
      J. Comp. Chem. 18, 368 (1997)

It describes a complete parametrization procedure, although of
course for a different molecule and probably with different
priorities.
____________________________________________________________________

From:"Klaus-D. Warzecha" 


At 17:04 14.09.98 -0500, you wrote:

>[...] Now I have some questions about parameterization to create
>forcefields for MD simulations.
>Specifically, I am interested in parameterizing Nitromethane.
>While any information regarding a set of parameters (bond, angle,
>torsion, out-of-plane,charges,van der Waals, etc) would be welcome,
>we are interested in generating the neccesary parameters, and in a form
>suitable for use in CHARMM.
>[...]

Howard,

unfortunately I'm no expert in computational chemisty and therefore not
sure whether the following will be helpful at all, but you might
consider
to have a look at an earlier article on

Molecular Mechanics Force-Field Parametrization Procedures.

A. J. Hopfinger, R. A. Pearlstein, J. Computational Chemistry, 1984, 5,
486-499.

I have to admit that I never used it myself and thus won't be able to
give
you any help on that, sorry.

Greetings,
Klaus-D. Warzecha

Klaus-D. Warzecha
Max-Planck-Institut für Strahlenchemie
Mülheim an der Ruhr
Germany


____________________________________________________________________


--

Howard E. Alper, Ph.D.
Dept. of Chemistry and The Advanced Materials Research Institute
University of New Orleans
New Orleans, LA 70148
504-280-7216

 - Helping molecules find happiness for almost a 5th of a century.



Similar Messages
09/18/1998:  Re: Parameterization
09/16/1998:  Parameterization
11/16/1993:  MM Parameters/Macrocycles
09/22/1998:  Re: CCL:Re: Parameterization
08/01/1996:  Re: CCL:M:Heat of formation calculation using MOPAC.
03/02/1992:  Oh, boy, oh, boy!  Real scientific controversy!
05/08/1995:  AM1 vs. PM3
10/01/1993:  torsion of conjugated systems -- summary
01/04/1995:  Summary of Answer to Ln3+ and water
11/02/1995:  summary AM1 vs PM3


Raw Message Text