Summary: vdW parameters



 Hello,
 About week ago I asked:
 > Recently I analyzed vdW parameters in different
 > force fields
 > (sybyl, few flavors from MSI, Charmm, MM3 etc) and I
 > noticed
 > that the parameters differ  substantially!
 > In crystallografer community the most popular are
 > Bondi vdW radii
 > and they are in most cases completely different from
 > those used in forcefields.
 > The only exception is sybyl forcefield, but here
 > also parameter for
 > H atom is 1.5 A (instead of 1.2 A) which gives 0.6A
 > difference for H-H
 > interaction.
 > My question is: what is origin of these (huge)
 > differences ? Physical
 > origin of the vdW is known and should not change
 > from forcefield to
 > forcefield ...
 > In a case of overlapping H-H pair a repulsive force
 > calculated using
 > sybyl vdW radii is substantial, (e.g. if distance is
 > 2.5A) but using
 > Bondi's radii the force  is small and even
 > attracting
 Here are responses I've got. Thanks again to all for valuable
 discussion.
 G.Bakalarski
 +++++++++++++++++++
 *From: "Michael K. Gilson" <gilson (- at -) umbi.umd.edu>
 The variation in H-bond radii may come in part from the fact
 that most current force-fields model hydrogen bonds in
 terms of electrostatics (Coulomb's law) + vdw interactions.
 In order to get a good strong H-bond, the radius of a polar
 H must be artificially reduced so the H's positive charge
 can get close to the H-bond acceptor.  If you need a reference
 for this, let me know and I'll dig one up.
 Regards,
 Mike Gilson
 ++++++++++++++++++++
 *From: Marvin Waldman <marvin (- at -) msi.com>
 With regard to your CCL question below about differences in vdW parameters,
 it has been found that the parameters can be quite sensitive to various
 factors such as the form of the potential function (e.g. exp-6, LJ 12-6, LJ
 9-6, ...), the form of the combination rules used for mixed parameters, the
 charge model used, and most importantly, the type of data used to fit the
 parameters (e.g. crystal structures, liquid simulations, ...).  Also, the
 parameters will be correlated with the torsional parameters of a force
 field as both are generally involved in determining vicinal (1-4)
 interactions.  As such, the parameters indeed do tend to vary quite
 substantially.  The vdW parameters are _not_ experimental_
 observables.  Rather, they are part of an overall model (the force field)
 which is used to predict/fit various experimental results.  Depending on
 the particular types of experiment you are trying to predict/fit and the
 rest of the force field model, the vdW parameters themselves can be quite
 different.  Discussion of some of these points can be found in the
 following references:
 Z.Peng, C.S. Ewig, M.-J. Hwang, M. Waldman, A.T. Hagler, J.Phys.Chem. A,
 vol. 101, pp. 7243-7252 (1997)
 C.S. Ewig, T.S. Thacher, A.T. Hagler, J.Phys.Chem. B, vol. 103, pp.
 6998-7014 (1999)
 Regards,
 Marvin Waldman, Ph.D.
 Senior Director, Rational and Combinatorial Drug Design
 Molecular Simulations Inc.
 ++++++++++++++++++++++
 *From: Alan.Shusterman (- at -) directory.reed.edu (Alan Shusterman)
 --- You wrote:
 what is origin of these (huge) differences ?
 --- end of quote ---
 Your question is a good one. The answer turns out to be fairly simple, but it
 has several parts.
 First, the vdW parameters used by force fields (FF) are designed to reproduce
 geometries of
 +particular molecules selected by the FF developer. The identities of these
 molecules vary from one
 +developer to the next, so the target data is not the same for all FF.
 Also, one can think of a FF as having two parts, the vdW part and everything
 else. The "everything
 +else" part varies from one FF to the next, so even if two FF tried to
 reproduce the same molecular
 +geometries they would have to use different "vdW" parameters to get
 there. (What is more, different
 +FF often use different equations to model vdW interactions, so different kinds
 of parameters are
 +required for this reason.)
 It also needs to be recognized that most FF are not designed to reproduce the
 data that so-called
 +experimental vdW radii are based on. Most (but not all) FF try to reproduce the
 geometries of
 +isolated molecules. Experimental vdW radii come from a variety of sources, but
 they generally
 +reflect some type of intermolecular interaction, and these interactions may
 contain more than vdW
 +interactions. As a result, comparing experimental and FF vdW parameters (and
 energies) may be
 +nonsensical.
 I hope this helps. You might consult Allinger's book, "Molecular
 Mechanics", for a more detailed
 +explanation.
 -Alan
 +++++++++++++++
 *From: Rick Venable <rvenable (- at -) gandalf.cber.nih.gov>
 One issue may be the exact meaning of the "radii"; CHARMM radii
 represent the minimum of a Leonard-Jones 6-12 potential, while other
 force fields (like Merck) may use a somewhat different interpretion, as
 well as additional terms.  You really have to look at the underlying
 equations used to explictly compute the VDW energy, and not just the
 radii.
 Although the physical basis for VDW may be understood as a diffusive
 interaction of the electron clouds which may have some long range
 character, most force fields use a simple Morse or LJ curve to model
 this interaction, which is mostly a short range approximation of the
 real VDW forces.
 --
 Rick Venable
 +++++++++++++++
 *From: Irilenia Nobeli <nobeli (- at -) biochemistry.ucl.ac.uk>
 Dear Grzegorz,
 Force fields are commonly empirically fitted which means that the van der
 Waals parameters are allowed to vary until  the value of some
 property(ies) that depends on them agrees with either experimental or ab
 initio results.
 Hence although these parameters are normally restricted to some
 "acceptable" values, the optimum fitted values are unlikely to be
 identical to what one would expect from Bondi radii, the most common
 reason for this being that they have to absorb other inaccuracies in the
 potential, e.g. approximations to the dispersion energy by R6 terms and
 neglect of other terms such as polarisation energy.
 Hope this helps.
 Irilenia
 ------------------------------------------------
 Irene (Irilenia) Nobeli
 Biomolecular Structure and Modelling Unit
 Department of Biochemistry and Molecular Biology
 University College London
 &&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&(end
 forever)