Hi there - I was pleased to read in the manual: if no selections are provided (refine.occupancies.individual=None and refine.occupancies.group=None) then phenix.refine by default will refine individual occupancies for atoms that have partial occupancy values in input PDB file (except zeros) and it will perform group constrained occupancy refinement for all atoms in alternative conformations. Occupancies of all other atoms will be fixed. If selections are provided (refine.occupancies.individual=atom_selection or refine.occupancies.group=atom_selection) then the selected atoms will be added to the list of atoms defined above (atoms with partial occupancies or those in alternative conformations). See this document for examples of atom_selection syntax. However, I want to refine my MET residues in double conformations, one as MET (AMET) and the other one as MSE (BMSE) with each 0.5 occupancy to start with. When I try that I get: == ----------Initial model statistics (before refinement)---------- ... |-Occupancies statistics------------------------------------------------------| Sorry: There are atoms with negative occupancies. Check input PDB file. === The PDB file has 0.5 and 1.0 occupancies only. I am also using: == refinement.refine.strategy=tls+individual_sites+individual_adp +group_anomalous refinement.refine.adp.tls="chain A" refinement.refine.adp.tls="chain B" refinement.main.number_of_macro_cycles=4 refinement.simulated_annealing.start_temperature=5000 refinement.refine.adp.individual.anisotropic =" chain C" refinement .input .xray_data .labels=I_Mystery(+),SIGI_Mystery(+),I_Mystery(-),SIGI_Mystery(-),merged refinement.refine.anomalous_scatterers { group { selection = "name ZN " f_prime = -0.3 f_double_prime = 2.5 refine = *f_prime *f_double_prime } group { selection = "name SE " f_prime = -5.2 f_double_prime = 3.8 refine = *f_prime *f_double_prime } } == Any hints on how to proceed ? Should I define groups for the occupancy refinement explicitely and not rely to automatic assignment? And if so how? How do people treat this partial Se-Met issue ? We solved the structure with SeSAD ... if I out MSE I get pretty large negative density peaks on the Se ... if I put MET I get positive peaks ;-) And there are good reasons to believe in partial Se- substitution in this case. Thanks in advance, Tassos