From: Kenno Vanommeslaeghe (kvanomme_at_rx.umaryland.edu)
Date: Mon Jul 01 2013 - 14:37:51 CDT
Hi Massimiliano,
On 07/01/2013 07:41 AM, Massimiliano Porrini wrote:
> In fact, OPLS FF uses a *geometric* combination rule both
> for LJ interaction strength (\epsilon) and for size of atoms (\sigma),
> whereas CHARMM FFs use Lorentz-Berthelot combining rules, that is
> geometric for \epsilon and arithmetic for \sigma.
That's an excellent question, thanks for bringing it up. The short answer
is that transferring parameters between force fields with different
combining rules is generally wrong, and may cause significant deviations
in properties. Let's just say that the fact that we *often* get away with
it *in the specific case* of OPLS --> CHARMM is just an empirical
observation. There's probably a reason for this and I have a good hunch of
what it could be, but I don't want to start speculating on this
mailinglist. If you want to know for sure, you have to ask my mentor Alex
MacKerell.
> Is there any workaround about this different way of implementing LJ
> parameters?
Not that I know of.
> For instance: can one somehow specify a \sigma geometric combination rule
> only for those atoms for which OPLS FF parameters are implemented?
That would leave all the vdW interactions between an "arithmetic" and a
"geometric" atom type undefined.
> Or it has already been demonstrated that using OPLS LJ parameters
> within CHARMM, implementing Lorentz-Berthelot combination rules, is not
> harmful for the results of the simulations?
Generally spoken, no, but there are a large number of case studies where
it gave pretty good results. Assuming the OPLS nitrate parameters Amin
spoke of are well-optimized, porting them into CHARMM is almost certainly
better than the other options that came up in this discussion thread
(short of a big ion parametrization project).
> On 1 July 2013 09:52, Kenno Vanommeslaeghe wrote:
>> because
>> the CHARMM nonbonded parameters are historically based on OPLS ones.
I'd like to correct this - I meant or write: "*some* CHARMM nonbonded
parameters".
Cheers,
Kenno.
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:21:22 CST