[gmx-users] Re: gmx-users Digest, Vol 68, Issue 86

[Alan Chen]

Hi All:

I just wanted to add my 2 cents to the already excellent comments posted
about the conversion of different types of ion Lennard-Jones parameters,
since I get lots of emails about this very subject.

It is misleading to think of an ion parameter as a SINGLE value of sigma and
epsilon.  The combination rules are a sort of shorthand, what was actually
parametrized were PAIRS of lennard-jones interactions between ions and
various other particles.  So in any attempt to "transplant" parameters from
one convention to another, at a minimum, one must refer to the original
parametrization literature to see precisely which interactions must be
"preserved" (via specific overrides in a topology or FF .itp file). For all
ion parameters, this definitely includes the ion-water interaction, since
this determines the free energy of solvation. In the case of the Cheatham &
Young ions,  the motivation for these parameters was to avoid
crystallization artifacts caused  by anomalously short cation-anion
distances, which was alleviated by including the crystal lattice as part of
the parametrization, therefore the cation-anion interaction parameter must
also be "overridden".

The sigma primes that are mentioned are derivable using very simple
algebra,  just by setting the sigma for a particular interaction PAIR to be
equal under both mixing rules, and solving for sigma prime.  For example,
here is a description of how Aqvist's ions were transplanted into AMBER
(note that this description is slightly more complicated because of a
conversion from R* to sigma, and that TIP3P O sigmas are identical in
either  combination rule).
http://ambermd.org/Questions/vdw.html (Note that this only preserves one
particular interaction PAIR, you would have to repeat this process for any
additional interactions in your simulation unless you are simulating only
ions and water.)

But just because you CAN do this doesn't mean it's a good idea. In
principle, one would have to override every possible pair interaction with
the  transplanted ion using the mixing rules appropriate to the ion. Even
then, you have no guarantees that any interaction that was not explicitly
part of the parametrization process are sane or correct, both because  the
"new" force field likely uses different charges and L-J parameters for
equivalent biomolecular atoms, and because you have no guarantee that these
interactions were at all validated even in the "original" force field unless
it was part of the parametrization process.

Practically speaking, however, it is only the water and highly charged
moieties that one will need to pay particular attention to, but my point is
that this  should always be done with great care, since small inaccuracies
in transplanting ion parameters can have dramatically bad results due to the
large energies involved (see Cheatham & Young's original report or my own in
http://pubs.acs.org/doi/abs/10.1021/jp0765392).

Cheers,

Alan Chen

postdoctoral research associate - Garcia Group
Rensselaer Polytechnic Institute, Troy NY
chena7 at rpi.edu




> > > On Tue, Dec 15, 2009 at 4:06 PM, Reza Salari <resal81 at yahoo.com
> > <mailto:resal81 at yahoo.com>> wrote:
> > >
> > > Hi All,
> > >>
> > >> Recently there has been a new set of ion parameters published by
> > Joung and Chetham and I am interested in running some test runs using
> > these parameters. These set of parameters are based on using LB rule
> > (arithmetic mean) for sigmas.
> > >>
> > >> However I am using OPLS-AA ff so I am using the combination rule 3
> > (geometric mean of corresponding A and B values). My question is that
> > can I use the exact sigma values from Cheatham for my simulations? I'm
> > almost positive that I have to change these sigma values to be
> > consistent with the combination rule that I am using. In fact there is
> > a paper by Horinek et al that has a nice table of different ionic
> > sigma and epsilon values from different parameter sets (Aqvist,
> > Jensen, Cheatham,..). The article is
> > >> here:
> > >>
> >
> http://scitation.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=JCPSA6000130000012124507000001&idtype=cvips&gifs=Yes
> > <
> http://scitation.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=JCPSA6000130000012124507000001&idtype=cvips&gifs=Yes
> >
> > >>
> > >> In that table, they have mentioned two sigmas; a usual sigma (which
> > is used with rule 2) and a sigma prime (which can be used with rule
> > 3). However it seems sort of unclear to me how they got these value
> > since in some references that they've mentioned I could find either
> > sigma or sigma prime, not both. So I am guessing there must be some
> > way to convert these two sigmas to each other.
> > >>
> > >> So does anyone know if there is such way? Does GROMACS internally
> > treats sigmas as "sigma prime" for OPLS-AA? I looked at the manual and
> > also searched the mailing list to find an explanation but without
> > luck. I really appreciate any help on
> > >> clarifying this.
> > >>
> > >> Regards,
> > >> Reza Salari
> > >>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20091217/59bf6c12/attachment.html>