Replying to my own posting for the archive: I think I have
identified the problem I described in my original posting. See
below.
On Wednesday 04 August 2004 07:08 pm, M. L. Dodson wrote:
> On Wednesday 04 August 2004 03:48 pm, Ross Walker wrote:
> > Dear Bud,
> >
> > > A question about sander.QMMM (AMBER8) input.
> >
> > First question, have you applied bugfix.13? If not, do this first of
> > all before you do ANY QMMM calcs.
>
> Hi Ross,
>
> I _think_ I did (I applied bugfix.all, IIRC), but I will check
> explicitly tonight.
>
I had, in fact, applied the bugfix (it was contained in bugfix.all).
> > The charge you specify in divcon.in should be just the charge on the QM
> > system. Bare in mind however that any bond you break will have hydrogen
> > atoms added in the QM region. E.g:
> >
> >
> > H H H H QM region = C1 and C2 H H
> > HC1-C2-C3-C4H ------------------------> HC1-C2H Charge =0
> > H H H H H H
> >
> > Hence check if you are breaking bonds and where. Then for every bond
> > you break add a hydrogen on the QM section and at the end of it see
> > what you are left with. Then calculate charge of this QM piece and set
> > that in the divcon.in file.
>
> Hmmm... That was my interpretation (in analogy with the full
> DivCon version, MOPAC, etc). I'll email you my input files Fri
> (I'm out tomorrow, and it is 6:50PM here at home). Briefly,
> though, the QM section of the run that failed is a dinucleotide + a
> carboxylate + a protonated amine, total charge of -1 (including
> the internucleoside phosphate):
>
> 5'O-nside1-3'O-PO_2-5'O-nside2-3'O (flanked by the rest of the DNA
> treated via MM). Plus the carboxylate and amine (residues on the
> enzyme). Two link hydrogens cap the 5'- and 3-hydroxyl oxygens
> (plus several for residues on the enzyme). CHARGE=-1 => "odd
> number of electrons in closed shell calculation" (or some such).
>
> The system that ran to completion with no error message left out
> the central PO_2, so would have two more link hydrogens in place
> of the phosphorus and two oxygens. CHARGE=0 => runs OK, but
> 3'-sugar ring opens between C3' and C4'. Oops!
>
> So it could be the bugfix problem, unless I've missed something.
>
See above. It was not any of the problems addressed by the
bugfix.13.
The problem appears to be in the way I set up the enzyme residue
sidechains contained in the QM region. One boundary atom was a
branch point, which would need two link hydrogens to satisfy its
closed shell valence. Evidently, the program was not smart enough
to read my mind on this one. :-) When I changed the definitions of
the QM atoms such that none required more than one link hydrogen,
the error message went away. I still have a problem with the
exact definition of the boundary atoms (two carbons in the QM list
get too far apart during 100 steps of minimization), but the
dinucleotide sugar rings do not open. A little more
experimentation, and I think I will have it.
So, for the archive:
No boundary atom (the QM atom separating the QM and MM regions)
can have more than one link atom assigned to it. (Evidently.)
Bud Dodson
--
M. L. Dodson bdodson.scms.utmb.edu
409-772-2178 FAX: 409-772-1790
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Thu Aug 05 2004 - 23:53:01 PDT
