Hi Jason,
Thanks for your continued help. We have made progress with your
suggestions, but are still confronting problems with SHAKE. Namely,
minimization/equilibration appear to run fine with pmemd (mpi), but the
same equilibration procedure fails with pmemd.cuda (Amber 14):
785: 31632 - 31633 and 787: 31633 - 31634
Hydrogen atom 31633 appears to have multiple bonds to atoms 31632 and 31634
which is illegal for SHAKEH.
Exiting due to the presence of inconsistent SHAKEH hydrogen clusters.
I've verified that the guilty hydrogen should indeed be bonded to both the
oxygen and opposing hydrogen (vmd output below) and that the geometry
appears normal. Oddly, this error is thrown for the TIP3 water with resid
10001, so there are 10K TIP3 waters before that entry that appear to be ok.
The .in file we ran was the following:
&cntrl
imin = 0,
irest = 0,
ntx = 1,
ntb = 1,
cut = 10,
ntr = 1,
ntc = 2,
ntf = 2,
ig=-1,
tempi = 0.0,
temp0 = 300.0,
ntt = 3,
gamma_ln = 1.0,
nstlim = 1000, dt = 0.002,
watnam='TIP3',OWTNM='OH2',
hwtnm1='H1',hwtnm2='H2',
ntpr = 100, ntwx = 100, ntwr = 1000
/
Hold the protein fixed with weak restraints
10.0
RES 1 90
END
END
/////
I've uploaded all the files we've used to
http://users.mccammon.ucsd.edu/~huskeypm/share/filez.tgz. To summarize, we
created the input files with CHARMM-GUI (in order to be consistent with
previous runs) and the charmm-36 FF. We used parmed (latest version from
the git repo) to obtain prmtop/inpcrd files:
chamber -top toppar/top_all36_prot.rtf -param toppar/par_all36_prot.prm
-str toppar/toppar_water_ions.str -psf step2_solvator.xplor.psf -crd
step2_solvator.crd -box 80.,80.,80.
outparm apo.prmtop apo.inpcrd
Minimizations were done using pmemd. All input files should be in the link
I provided above.
If you're able to look into this and get this running via pmemd.cuda, we'd
be very grateful. I'm sure were missing something subtle, so a third pair
of eyes would be helpful.
Thanks again for your time,
Pete
Is there something else obvious that we are missing?
Info) ------------
Info) molecule id: 0
Info) trajectory frame: 9
Info) name: H2
Info) type: HT
Info) index: 31633
Info) residue: 10108
Info) resname: TIP3
Info) resid: 10001
Info) chain: S
Info) segname: SOLV
Info) x: 16.492001
Info) y: 34.381001
Info) z: -0.762000
Info) picked atom:
Info) ------------
Info) molecule id: 0
Info) trajectory frame: 9
Info) name: H1
Info) type: HT
Info) index: 31632
Info) residue: 10108
Info) resname: TIP3
Info) resid: 10001
Info) chain: S
Info) segname: SOLV
Info) x: 17.830999
Info) y: 34.262001
Info) z: -1.458000
Info) picked atom:
Info) ------------
Info) molecule id: 0
Info) trajectory frame: 9
Info) name: OH2
Info) type: OT
Info) index: 31631
Info) residue: 10108
Info) resname: TIP3
Info) resid: 10001
Info) chain: S
Info) segname: SOLV
Info) x: 17.347000
Info) y: 33.957001
Info) z: -0.690000
On Fri, Oct 31, 2014 at 11:51 AM, Jason Swails <jason.swails.gmail.com>
wrote:
> On Fri, Oct 31, 2014 at 10:20 AM, Caitlin Scott <cescott.caltech.edu>
> wrote:
>
> > Dear AMBER users,
> >
> > I've been using CHARMM-GUI to generate psf and pdb files of my solvated
> > and neutralized protein. Previously, psf files were generated in the
> > CHARMM format, but recently, CHARMM-GUI changed its psf file format to
> > XPLOR. I can still generate the inpcrd and prmtop files, but the job
> > crashes during MD.
> >
> > When I submit it with AMBER 12 pmemd.cuda, I get the error message:
> > Hydrogen atom 2882 appears to have multiple bonds to atoms 2881 and 2883
> > which is illegal for SHAKEH.
> >
> > According to the psf file, hydrogen atom 2882 is bonded to an oxygen and
> > the other hydrogen in same the water molecule. Is this unusual for the
> > TIP3 water model?
>
>
> No, this is normal. That is how the geometry of the TIP3P water model is
> defined.
>
>
>
> > When I check the md output file, it says:
> > Number of triangulated 3-point waters found: 0
> >
>
> Amber identifies triangulated 3-point water molecules by looking at atom
> and residue names. tleap always names water residues WAT and the atom
> names within WAT residues as O, H1, and H2. If the names are not what
> sander/pmemd expect them to be, triangulated waters will not be identified
> and this error will result.
>
>
> >
> > When I was using CHARMM psf files to generate the inpcrd and prmtop
> files,
> >
> > I would get this:
> >
> > Number of triangulated 3-point waters found: 32385
> >
> >
> >
> > Can someone please explain what is a triangulated water, and why are they
> >
> > missing when the psf is in the XPLOR format?
> >
> >
>
>
> I don't think this has anything to do with XPLOR vs. CHARMM format, but
> rather that CHARMM-GUI gives water residues (and water atoms) different
> names than CHARMM itself does?
>
> Look at the "watnam", "owtnm", "hwtnm1", and "hwtnm2" variables (p. 293 of
> the Amber 14 manual) -- you can change those variables in the &cntrl
> section of your mdin file to get pmemd.cuda to recognize the water naming
> scheme for your system.
>
>
> >
> > When I run this job with AMBER 12 sander, there are still no triangulated
> >
> > waters, but the job doesn't crash.
> >
>
> My guess is that means that the regular SHAKE algorithm is used (rather
> than the fast, analytical SETTLE) algorithm to fix the water geometry (you
> can measure some water geometries to make sure that the constraints are
> being correctly applied).
>
> HTH,
> Jason
>
> --
> Jason M. Swails
> BioMaPS,
> Rutgers University
> Postdoctoral Researcher
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
======================
Peter Kekenes-Huskey, Ph.D.
Assistant Professor
Department of Chemistry
University of Kentucky
http://pkh.as.uky.edu/
======================
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Oct 31 2014 - 14:00:02 PDT
