Hello, I recently started to refine a structure with phenix.refine for a first time. It is a 3A dataset with NCS of 8. For the refinement i wanted to use TLS groups as well as individual B-factors (data/parameter ratio should be fine with ncs of 8) and after several cycles of refinement i get the Rfac of 20 and Rfree of 24 which is quite good (i also use simulated annealing, Secondary structure restraints and NCS restraints). Re-reading the refinement manual I realized that by using a PDB file which comes from a phenix.refine with TLS in the next refinement i am probably doing an anisotropic B-factor refinement for each atom, which in my case is not justified. I found a post on ccp4bb from 19 Fev 2009 discussing this issue but for me it is still not very clear. By reading the PDB i got from phenix.refine (version 1.7.3) it is not very clear (at least for me) if the refinement was with TLS groups and individual isotropic B-factors or with anisotropic B-factors. Basically my question is: if the input PDB has already anisoU records from a previous phenix.refine with TLS groups is the new phenix.refine going to use isotropic individual B-factors or anisotropic ones? (Even if in my case one can assume that i can probably use anisotropic B-factor refinement (24000 unique reflections with ncs of 8 vs. 6000 atoms in total) this is rarely the case for non-atomic resolution datasets) I would be very happy if someone gives an explanation. Many thanks Cheers Filip