Dear AMBER users and Junmei,
I've been doing a bit of poking around in the atom types assigned by
ANTECHAMBER (currently I'm using v 1.2.4, although I assume what I'm
describing also applies to the (older) version distributed with AMBER
8), and I think I've found a bug. In particular, nitrogens in some
rings like pyrrole and indole seem to be misassigned as "na" rather
than "nc", which seems more appropriate since these are aromatic. I
can provide input pdb's and output mol2 files if that would be
helpful. Is this a bug?
This is a bit concerning, since a number of us in the Dill and Pande
labs are working to use GAFF parameters to automatically compute
binding and hydration free energies for a relatively large number of
molecules. It's possible, of course, that problems with atom typing
that don't change vdW parameters (as in this case) won't change such
things very much, but it still seems a bit concerning.
Has there been any sort of systematic test of how the atom types
assigned by ANTECHAMBER compare to the atom types assigned by other
automatic atom typing programs (for example, prodrg, or similar tools
from OpenEye)? I've skimmed the recent ANTECHAMBER paper and all I can
find is a "success rate" which describes how many atoms actually
caused failure (in terms of crashes and so on). There is a similar
statement on the ANTECHAMBER web page. But this seems rather different
from the less-obvious "failure" we're seeing here (assigning improper
atom types).
Is there some reason this shouldn't be considered a failure? And is
there any work already being done to diagnose whether problems like
this are more widespread?
Thanks,
David Mobley
Dill Group
UCSF
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Tue Apr 18 2006 - 03:02:43 PDT
