On Tue, Dec 29, 2009, Alan wrote:
> First, the molecule is the ligand GNP in 1E0A from PDB:
> http://www.pdb.org/pdb/explore/explore.do?structureId=1E0A
The ligand there has a net charge of -4: is that what you used?
This is not a crazy charge assignment, although I would expect the terminal
phosphate group to have an extra proton attached at pH 5.5....
> The problem is that in the complex 1E0A (done in pH 5.5), GNP has only 13 H
> and not 17 H as shown in formula or in the ligand links above.
The "ligand links" show a (hypothetical) neutral ligand, that would only have
such a protonation state at extremely low pH.
> So, trying to summarise. GNP_1E0AS0 failed, even after 10 hours. From
> sqm.out log file, some excerpts:
> QMMM: WARNING!
> QMMM: Unable to achieve self consistency to the tolerances specified
> QMMM: No convergence in SCF after 1000 steps.
Ross (and other developers): we need to issue a bug fix to make this a fatal
error, certainly for stand-alone sqm. Do you agree?
Further exploration: if I use the ligand from the 1E0A file with mopac (from
Amber10) it works OK (with a charge of -4).
If I use sqm, however, I get the convergence failure noted above. However,
if I take the optimized structure from the mopac run, and feed it to sqm,
things work fine.
For developers (Ross, Andreas): I'm including an sqm_bad.in (which is the
original geometry), and an sqm_good.in (which has the mopac-optimized
geometry). The geometries are not that much different -- some changes in
torsional angles, but nothing drastic. So: is there some parameter that could
be tweaked to improve SCF convergence for the "bad" case? Something else
that is inhibiting sqm for this molecule?
...good luck....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Dec 29 2009 - 12:00:03 PST
