Is it possible to get the gradient (i.e. the derivative with respect to each independent degree of freedom) of the energy score evaluated on a given protein pose? It seems like the Minmover subroutine moves the pose in a small step in the direction of this gradient, meaning the gradient in implicit in the Rosetta code. I'd like to access the value of the gradient itself at any fixed pose, for example in terms of dihedral coordinates or Cartesian coordinates. Technically one could numerically approximate the gradient by perturbing each degree of freedom (say dihedral angle) one at a time and seeing how the energy function changes, but this is a very inefficient solution given the high dimensionality of the situation.
Since the formula for energy is very complicated and probably a nightmare to differentiate by hand, I'd be quite content to have the energy function in a complete explicit form so that I could enter it into an automatic differentiation software package (of which there are many good ones these days). So far most of the references I've seen in the Rosetta literature give a high-level explanation of each term but not quite an explicit form (or the various constants and so on needed to get that explicit form).