Hi Yuri,
1- I figured out the occupancy problems. I had to define the ligands as group occupancy rather than individual atoms...(duh!!...) It looks better now.
yes.
2- Is there any case/resolution when one should try anisotropic refinement for a ligand?
High resolution, more or less fully occupied. If resolution better than ~1.3-1.2A then probably refine it rather with aniso ADP than iso. There are no general rules, it varies case by case.
3- About the waters occupancies/B-factors I definetely see some density peaks that correspond to waters (good distance and angles for H-bond interactions with protein backbone on the surface) Obviously they are weaker than peaks for waters that are ordered near the active site. I modelled those in and I got some negative peaks in the mFo-DFc map after isoptropic B refinement. That is what led me to think I should try occupancy refinement.
4-Still on the same topic B-factors, I am also encountering the same problem with surface side chains especially (D, E, Q). I am confident they should be where they are. backbone density looks great. But no matter where I put their side chains, I get negative peaks in the diff. map. Their B-factors usually refine to ~55 vs ~25 for ordered side chains.
These side chains are likely to have several conformations and your map shows you only one or a few (2-3?). If this is the case, you should try refining group occupancy for these residues (one occupancy per side chain), so it refines to something less than 1. If you have (you model) more than one conformer, then the sum of group occupancies must be <=1. Pavel.