Hello Justin and Mark.
>>> All angles in CHARMM27 use Urey-Bradley type angles, so the angle
>>> bending energies are the U-B term.
>>
>> Different forcefields/programs use the label "Urey-Bradley" to refer
to
>> either the sum of the harmonic angle potential and the harmonic 1-3
>> potential (e.g. CHARMM27 in GROMACS), or just the latter (e.g.
CHARMM27
>> in CHARMM). So mileage does vary.
Ah, good to know. Yeah, I was used to separate terms, but I probably
should have figured that out when there wasn't an explicit "BEND" term.
>> You didn't say how you were measuring the energies, but if it was MD
>> averages, then your averages are over different ensembles, even from
the
>> same starting point with notionally the same algorithm.
It's from a zero step MD run with GROMACS, and a single point energy
calculation with CP2K, so shouldn't be an issue.
>> Probably. Getting the output from grompp -pp, and then choosing
atom(s)
>> to have zero charge, and/or zero VDW parameters is a good way to
>> trouble-shoot, assuming your other software can be massaged to
replicate
>> the effect. Before that, check that the energy of a single
amino-acid
>> can be replicated.
Good ideas. I'll give that a shot and see what comes up (I wasn't
familiar with the -pp option, and didn't want to change the original ff
files). Thanks!
Cheers, Matt
> > -------- Forwarded Message --------
> > From: Mark Abraham <Mark.Abraham at anu.edu.au>
> > Reply-to: Discussion list for GROMACS users <gmx-users at gromacs.org>
> > To: Discussion list for GROMACS users <gmx-users at gromacs.org>
> > Subject: Re: [gmx-users] Force-field checking options
> > Date: Wed, 03 Aug 2011 22:47:39 +1000
> >
> > On 3/08/2011 9:20 PM, Justin A. Lemkul wrote:
> > >
> > >
> > > mcgrath wrote:
> > >> Hi Mark.
> > >>
> > >>> Uh, they're available wherever you were getting energy break-downs
> > >>> from (e.g. in the .log file, or via g_energy on the .edr file). You
> > >>> can't get a break-down for each interaction, however. Simplify your
> > >>> system to probe things here. Do zero-step MD (not EM) without
> > >>> constraints, to evaluate the energy of a single conformation, and
> > >>> compare that with your other software. Complex things are complex to
> > >>> compare. :-) Reduce the complexity.
> > >>
> > >> I must be missing something really obvious, then, because this is where
> > >> I'm getting the energies from in the log file.
> > >>
> > >> Energies (kJ/mol)
> > >> U-B Proper Dih. Improper Dih. LJ-14
> > >> Coulomb-14
> > >> 2.42228e+02 1.23963e+02 6.15742e-02 1.43130e+02 -5.01705e
> > >> +03
> > >> LJ (SR) Coulomb (SR) Potential Kinetic En. Total
> > >> Energy
> > >> 8.30424e+01 -4.63865e+02 -4.88849e+03 1.05098e+02 -4.78339e
> > >> +03
> > >> Temperature Pressure (bar) Constr. rmsd
> > >> 3.00959e+02 -7.08408e+02 6.28488e-06
> > >>
> > >> I don't see any bend energies there. No bonds, either, but that's fine,
> > >> because I was constraining them (constraints all-bonds). Removing that
> > >> constraint gives a bond energy in that section, but still no bend
> > >> energy.
> > >>
> > >
> > > All angles in CHARMM27 use Urey-Bradley type angles, so the angle
> > > bending energies are the U-B term.
> >
> > Different forcefields/programs use the label "Urey-Bradley" to refer to
> > either the sum of the harmonic angle potential and the harmonic 1-3
> > potential (e.g. CHARMM27 in GROMACS), or just the latter (e.g. CHARMM27
> > in CHARMM). So mileage does vary.
> >
> > > -Justin
> > >
> > >> Agree completely about simplification. The first system I quoted for
> > >> Justin was for two molecules (ATP + Mg), which is the smallest system
> > >> that I see the problem for (as I mentioned, using 29 or 25000 waters
> > >> gave a difference of only 1%, which is explainable by slight rounding of
> > >> the coordinates...at least, possibly).
> >
> > You didn't say how you were measuring the energies, but if it was MD
> > averages, then your averages are over different ensembles, even from the
> > same starting point with notionally the same algorithm.
> >
> > >> He then asked for the breakdown
> > >> for the big system, so I gave him that, too.
> > >>
> > >>> ...and the magnitude of its energy contribution is available.
> > >>
> > >> Also agreed; that's how I concluded that its contribution was only a
> > >> small part, and using only ATP there appears to be no CMAP (see the
> > >> above energies). But, even on this small system, I haven't been able to
> > >> hunt down the differences. The parameter files have the same values for
> > >> the nonbonded parameters (converting between the GROMACS and CHARMM
> > >> format) and charges, so I'm not seeing any simple solution. Unless
> > >> someone has an idea to save me, I guess I'm stuck with looking at
> > >> individual contributions for this system.
> > >>
> > >> Sigh. Some days this field isn't so much fun. :) Especially since the
> > >> bug will probably be something stupid I did in setting it all up.
> >
> > Probably. Getting the output from grompp -pp, and then choosing atom(s)
> > to have zero charge, and/or zero VDW parameters is a good way to
> > trouble-shoot, assuming your other software can be massaged to replicate
> > the effect. Before that, check that the energy of a single amino-acid
> > can be replicated.
> >
> > Mark
> >
> > >>