Not a problem for CMAPs per se--just that if you want to debug the forces
that come out of *sander*, you cannot do it easily for a CMAP potential
calculator that you have written. *Sander*'s &debugf module will not
produce CMAP forces (or the energy they entail) in a transparent way. I
can create a system with no vdW or charge parameters (as I think I modified
the alanine dipeptide model included in this thread), such that it has
bond, angle, dihedral, and ff19SB CMAP properties. I can then compute the
bond, angle, and dihedral energies with some alternative program, and
compare them to the &debugf forcedump.dat output. The energies I compute
match the results printed in *sander* mdout, and all of the forces match
the results in forcedump.dat EXCEPT FOR the forces on atoms involved in the
CMAP term. The discrepancy is maintained throughout all of the various
force terms, and even the electrostatic and vdW forces (which should not
exist) show the same discrepancy in the atoms affected by the CMAP term for
the central alanine residue. When I compute the CMAPs with a program like
*mdgx* and reach in to have them printed from the actual force loop, they
match the discrepancy in the *sander* &debugf forcedump.dat output, and my
own program's forces now match that signal as well, so I am pretty
confident I am properly computing the CMAP derivatives and forces.
But, *sander*'s &debugf module should get an update to be consistent with
all of the valence (and probably 12-6-4 non-bonded) terms that the program
really knows.
See you around,
Dave
On Thu, Dec 23, 2021 at 10:41 AM Carlos Simmerling <
carlos.simmerling.gmail.com> wrote:
> Hi Dave,
> I'm not sure I follow what this means for the CMAPs - is there a problem?
> carlos
>
> On Tue, Dec 21, 2021 at 11:41 PM David Cerutti <dscerutti.gmail.com>
> wrote:
>
> > Hello Amberites,
> >
> > After some wrestling with it, I've managed to get *sander*'s &debugf
> > namelist
> > in gear to print out the forces due to various force field terms that I
> > should expect on each atom. However, I am trying to prepare / debug a
> CMAP
> > force calculation, and when I create a system with these terms in it I
> find
> > that the &debugf output is messed up in precisely the atoms affected by
> the
> > CMAP term. I can examine my other forces to verify that they are
> correct.
> > The &debugf module has no apparent capability to handle CMAPs (and,
> > presumably, other CHARMM terms). Instead, the forces seem to shine
> through
> > the output like speckling in a corrupted photograph, adding their image
> to
> > every existing quantity (including the total energy). The various
> > components do not, it seems, add up to the total energy, either, so there
> > is definitely more than one accounting issue here.
> >
> > The &debugf namelist is actually a great tool, if used properly. It's
> why
> > I tried to include an elaborated version of it in *mdgx*, although I
> > probably have to update that to give proper treatment to CMAPs as well.
> >
> > Dave
> > _______________________________________________
> > AMBER-Developers mailing list
> > AMBER-Developers.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber-developers
> >
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu Dec 23 2021 - 09:00:02 PST
