On Feb 11, 2010, at 2:35 PM, M. L. Dodson wrote:
> Hello all,
>
> I am seeing very strange behavior equilibrating a QMMM system prior to steered MD runs. 3 hydrogens from the guanidinium group of an ARG spatially adjacent to, but not part of, the quantum region dissociate (i.e., the ARG is part of the classical region). There is a single water in the quantum region (QR), and the QR contains components from a modified DNA sugar and two residues of the enzyme protein, as well. The run is done with iwrap=1 and ntc=ntf=2, but qmshake=0, since the production SMD run to follow will involve proton transfers (dt=0.0005 to compensate). The equilibration uses ncsu_smd with a sum of two distances collective variable held constant at its starting value. As expected, the CV is observed to not significantly change over the course of 1ns of simulation (2000000 steps). The distances comprising the collective variable are between atoms in the QR. The system is being run with sander.MPI modified to incorporate the LCOD ncsu collective variable (I have extensive experience with this modification and have seen no other problems with it). 8 processors are used in the simulations. The input file and qmmm_region.pdb files are attached.
>
> Does any one have any suggestions what might cause this (amazingly strange) behavior?
>
> Thanks,
> Bud Dodson
>
Replying to my own email:
Just to let the list know, I think I have diagnosed the problem leading to hydrogen dissociation from ARG guanidinium groups spatially adjacent to the quantum region in QMMM simulations. It appears to be the interaction of two variables: qmcut and qmshake. In the simulations showing the H-dissoc, qmshake=0 AND qmcut > cut (in the &cntrl section, where ntc=ntf=2, also). If qmshake=1, no dissociation, independent of qmcut relationship to cut (not all combos tested, only qmcut > cut and qmcut=cut). If qmcut=cut=8.0, no dissociation, independent of qmshake setting.
For some reason the input files I used (and reused) set both cut and qmcut in their respective namelist sections. (I'm sure I had good reasons at the time I first wrote those input files, but can't remember without reviewing my notes from several years ago when I started doing QMMM with AMBER9.) I had been using cut=qmcut=9.0. Before running the recent simulations that have had problems, I changed to cut=8.0, but did not change qmcut from 9.0. I set qmshake=0 since I want to study proton transfer reactions.
To recapitulate for the archives (all with ntc=ntf=2, H dissoc. observed only from ARG guanidinium immediately spatially adjacent to the quantum region):
cut qmcut qmshake results
8.0 9.0 0 ARG guanidinium hydrogen dissoc.
8.0 9.0 1 no ARG guanidinium hydrogen dissoc.
8.0 8.0 0 no ARG guanidinium hydrogen dissoc.
8.0 8.0 1 no ARG guanidinium hydrogen dissoc.
cut > qmcut 1 not done
cut > qmcut 0 not done
QMMM simulations without the ncsu SMD restraint were not done.
Bud Dodson
--
M. L. Dodson
Business email: activesitedynamics-at-comcast-dot-net
Personal email: mldodson-at-comcast-dot-net
Phone: eight_three_two-five_63-386_one
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Feb 13 2010 - 12:30:02 PST
