[gmx-users] ligand force field parameters
mic0404 at yahoo.com
Tue Dec 16 13:22:00 CET 2003
> m b wrote:
> >>3.suppose if such parameters are not available to one such ligands of
> >>interest can some suggest me means of generating FF parameters.
> > The way I normally go is:
> > * use the OPLSAA force field
> > * calculate partial charges as ESPD charges with, e.g., ChelpG
> > in Gaussian (or anything compareable) for the ligand
> > (to be consistent this should also be done for the lipid/protein,
> > however, if these are too big then have to use the built-in OPLS charges)
> I'm not sure if I would call that 'consistent'. If you consider that
> the charges and the other parameters in the forcefield are interdependent,
> IMHO it doesn't make sense to replace one set of charges with another...
I find that here we are talking about THE most important issue when
trying to estimate the reliability of classical molecular simulations,
and I have the feeling that people often give this issue not
the amount of consideratuion that it deserves.
I appreciate that playing around with the parameters of
a well established force-field is always problematic.
But what alternative do you see when introducing new
molecular types (in many cases drug-like molecules which
can theoretically have all imaginable kinds of topologies
and functional groups) ??
You have a number of choices to determine the partial charges
of your new molecule/drug-candidate:
1) use partial charges found in the Gromacs force field
on similar groups and residues
2) use an independent set of charges as for example
obtained from ab-initio calculations (ESP, AIM, Gasteiger, etc)
or iteratively via "charge-increments" according to the relative
electronegativities of the involved atoms. (in my experience ESP
charges are the best choice in most cases)
3) optimize the charges in your molecule in a similar fashion
as was done for the remainder of the (Gromacs) FF by fitting
everything to experimental thermodynamic and structural data
(this IS the way the Gromos/Gromacs FFs were optimised
isn't it ?)
Option 3 would be the only really consistent approach, but
we clearly have to dismiss it because this approach
is simply not feasible.
Option 1 I consider to be rather problematic, the ramifications of this
approach are very hard to estimate, especially if (parts of) the new
molecule do not have any resemblance to any molecules that are
already an integral part of the FF (mostly AAs I take it).
In this case the consequences of this approach are un-predictable.
Option 2 would clearly be the best way to go if the VdW, electrostatic
and the intra-molecular parameters of the force-field had been
optimised independently of each other, which AFAIK, is not the case
for the Gromos/Gromacs FFs (and most other common FFs for peptides
for that matter)
Does not sound too good, does it !?
In spite of its limitations I personally favor approach 3,
ideally for ALL atoms in the system.
And here I come to the your concerns about the "interdependent"
parameters: Normally one is interested in the INTERactions between
a molecule and some poly-peptide. When it comes to SPECIFIC
interactions which is the most interesting part in ligand-receptor
simulations than in most cases electrostatic interactions will
play a (if not the) major role here.
Thus, I do believe it is much less a severe approximation to use
consistent charges throughout for the price of sacrificing the
consistency between VdW/intra-molecular parameters in the protein
as compared to using inconsistent charges in ligand and receptor.
Do you Yahoo!?
Protect your identity with Yahoo! Mail AddressGuard
More information about the gromacs.org_gmx-users