Docking structures with ligands in cryo-EM maps

Greetings Rosetta gurus-

We are attempting to use Rosetta 3.3 to flexibly dock crystal structures into relatively high resolution (around 4.8 Angstroms) cryo-EM maps. Our protein of interest binds and hydrolyzes GTP, and oligomerizes in a nucleotide-dependent manner. We have cryo-EM structures of the protein in GDP and GTP states (bound to a non-hydrolyzable analog) in physiological oligomers, and crystal structures of the protein in both nucleotide states in non-physiological conformations/oligomerization states.

Since nucleotide is critical to the protein's function and resolving its role in oligomerization was the point of our experiments, we really really want it to be incorporated into our docking models. We are confident that we can resolve the nucleotides sufficiently to include them. A bonus would be if we could also include magnesium ions and coordinating water molecules, which we cannot resolve. The crystal structures include Mg and HOH but do not resolve the position of hydrogens, which would need to be modeled. The position and orientation of the nucleotides has changed in our structures relative to the crystal structures, and thus Mg and HOH would need to be repositioned based on Rosetta's forcefield.

We have successfully generated param files for our ligands and performed a few dozen runs of relax into one of our density maps, including both Mg and waters using the TP3 params, having approximately placed the hydrogens in positions which seemed chemically reasonable in the starting models. The protein main chains are consistently being remodelled to fit into the density and the nucleotides are being repositioned. The Mg ions are being moved around in a more variable manner; however, there is a cluster at approximately the same site. The water molecules, on the other hand, are all over the place.

Based on reading other posts in the forum, this is not entirely a surprise, but we are wondering if there is anything we can do to improve the modelling of the water molecules (i.e. using TP5 params, Rosetta 3.4, etc.) or if it is not reasonable to include them. Is there implicit solvation/coordination of the Mg ions in the calculations? Also, our expertise is in cryo-EM and not modelling, so if anyone has suggestions of a better approach we would be most grateful.

Cheers!