[gmx-developers] Adding an artificial energy term for structure refinement

[Alexander Björling]

My aim is to reproduce x-ray scattering data of a molecule in
solution. The data not only describes the ground state, but also
intermediate conformations of various reactions.

To get the time-averaging right and to make sure the fitted structures
make some mechanical sense, the idea is to use GROMACS with an
artificial potential energy, proportional to some deviation of the
predicted scattering of each MD frame to experimental data. An
expensive thing to calculate, sure, but let's assume that we can
handle this somehow.

Similarly to the problem posed here,
http://gromacs.5086.n6.nabble.com/Editing-the-GROMACS-energy-function-tp4669473.html,
the artificial energy term is a real number resulting from a
complicated operation involving all coordinates. The difference is
that this energy is really all I have - I've no way of calculating
forces.

While the hacking seems doable, my question is whether this approach
is meaningful with GROMACS, or if I need to provide the force on each
atom?

Thanks,
Alexander Björling,