I'm have some questions about doing real space refinements using phenix.real_space_refine. I am refining a model for an intermediate-resolution (3.79 Å) cryo-EM map. As the starting point I use a model from a higher-resolution map (3.25 Å) of the same system in another conformation. My idea is to first do rigid body fitting of domains and then gradually relax constrains and restraints. - Is it possible to restrain the position of certain parts of the model (i.e. "freeze" them so that they move at all) to do focused refinement of subdomains or loops? - In the instructions it says that you can do both NCS constraints and restraints. How you do the latter? I can only do the former by default or by defining ncs_constraints=True and give a NCS definition file. - When I do a run with run=rigid_body, some rigid body groups move in to each other and overlap. Is there no collision detection using this algorithm? If I do a run=minimization_global, things work a bit better in this regard, but is not really what I want (at this stage). Am I doing something wrong? Here is the run command and the rigid body definition file: phenix.real_space_refine \ model.pdb \ map.ccp4 \ rigid.eff \ resolution=8.0 \ ncs_constraints=False \ rotamer_restraints=True \ ramachandran_restraints=True \ c_beta_restraints=True \ run=rigid_body \ macro_cycles=10 # rigid.eff: refinement { rigid_body { group = chain A and resseq 101:266 group = chain A and resseq 273:388 group = chain B and resseq 101:266 group = chain B and resseq 273:388 group = chain C and resseq 104:266 group = chain C and resseq 273:388 } } Regards, Daniel Larsson