You can look in the pmemd code for the derivative calculations--the file is
going to be ${AMBERHOME}/src/pmemd/src/cuda/kBWU_cmap.h. You can thank /
blame me for making all of the little includables. There is no finite
difference calculation, it's a straight-up change-of-coordinates chain rule
application. You can also look in mdgx,
${AMBERHOME}/AmberTools/src/mdgx/BondContrib.c, to see a different
expression of the same thing for dihedrals. The CMAP is two dihedral
angles, computed one after the other. I think the only difference is that
the magnitude of the force comes not from computing [ K cos(N theta) ] but
rather evaluating the bicubic spline for magnitude and derivative in each
dimension. The chain rule for each set of four atoms A-B-C-D and B-C-D-E
should be the same as for the standard dihedrals after that. I'm working
in that code myself to try and port CMAP functionality to mdgx.
Dave
On Fri, Sep 6, 2019 at 8:21 PM Lod King <lodking407.gmail.com> wrote:
> It seems like, it does not transmute back to Cartesian coordinates, but
> uses the finite difference method to calculate the force (respect to phi
> and psi)
>
>
> On Fri, Sep 6, 2019 at 1:29 AM David Cerutti <dscerutti.gmail.com> wrote:
>
> > You've got two dihedral angles, A-B-C-D and B-C-D-E. The potential
> energy
> > is computed using bicubic spline interpolation on a grid of values,
> > typically 24 x 24 point in each direction (the potential is known at 15
> > degree intervals, and interpolated based on that). The forces on atoms
> are
> > computed form the derivatives of that bicubic spline interpolation, which
> > must then be transmuted via the chain rule back to Cartesian space on
> each
> > of the atoms.
> >
> > HTH,
> >
> > Dave
> >
> >
> > On Thu, Sep 5, 2019 at 6:35 PM Lod King <lodking407.gmail.com> wrote:
> >
> > > Hi Amber
> > >
> > > How to explicitly understand the force computed from U(CMAP) energy
> > term. I
> > > have been reading the papers "Mackerell, Feig, and Brooks Vol. 25, No.
> 11
> > > Journal of Computational Chemistry" and "Martin Karplus 1995 Journal of
> > > Computational Chemistry", and still having difficulty understanding the
> > > math applied here.
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Sep 09 2019 - 10:30:02 PDT
