On Mon, Feb 4, 2013 at 11:22 AM, David van der Spoel <sp...@xray.bmc.uu.se>
wrote:
>
> On 2013-02-04 17:18, Geoffrey Hutchison wrote:
> >> * Are there cases where MMFF94s should be preferred to Ghemical?
> >
> > Everything. The heuristic should be something like "MMFF94 if it works,
UFF otherwise." Ghemical is included for backwards-compatibility -- it was
much easier to implement than MMFF94.
> >
> >> * Is GAFF particularly good at any systems that the others handle
poorly?
> >
> > It's the "Generalized Amber Force Field." It should be fairly good with
proteins, but I don't think it's been tuned for speed as much as MMFF94 and
UFF have been.
> >
> Speed? Should make no difference at all. Lennard Jones + Coulomb on
> point charges. It is rather not for proteins (that is regular Amber) but
> for small (organic) ligand molecules.
Are you talking about the analytic forms used in the forcefield, or the
OpenBabel implementation? Some of the forcefields in OpenBabel have been
tuned and tested more extensively than others, I think that this was the
point Geoff was making wrt speed.
We mainly want to use these for cleaning up drawn structures at this point,
so if there are no major drawbacks to using GAFF or MMFF94 to get
qualitatively correct geometries, we're most interested in speed and want
the faster/better-scaling one.
Another constraint we're working under is that we can only access openbabel
by interacting with an external process, such as obabel (This is due to
licensing -- the new Avogadro code is BSD licensed, so we can't link to the
GPLv2 OB libraries). Is there some analytic way to know if a molecule will
fail to setup under MMFF94 or GAFF without actually trying? It will be a
much easier process to say "this molecule has atoms of type X, so fall back
to UFF" than to launch a trial process, wait for it to complete, parse the
log for an error string, then repeat with another FF, etc etc etc. It will
be best if we can do most of the setup work ourselves and then just run
obabel once with a configuration that we know will work.
Dave
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_jan
_______________________________________________
OpenBabel-Devel mailing list
OpenBabel-Devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/openbabel-devel
