Hi Yuri,
I am refining a liganded structure at 2.1 and I am pretty far into my modeling. Rfree = .265 rwork= .20
did you do Xray target/Restraints weights optimization for both: coordinates and ADP refinement?
I- If I refine individuals sites and real space (XYZ) and individual ADP with aniso/iso selection automatic my Rfree goes up slightly and Rwork drops to about .196. I also select optimize x-ray/ADp weight I should say.
If this is happening after a phenix.refine with Xray target/Restraints weights optimization for both: coordinates and ADP refinement, then 1) make sure you are using the very latest PHENIX version from nightly builds; 2) if you are using the latest version, then you can set weights manually such that it results in smaller gap between Rfree/Rwork.
II- When I choose no XYZ refinement with anisotropic ADP for protein chain only with weight optimization Rwork plummets to 0.15 and Rfree also drops to 0.25. But my gap is now .10 !!
You can't refine individual anisotropic ADPs at 2.1A (if this is what you did), so the result you got is quite expected.
I checked my log file and it looks like in default automatic individual ADP phenix.refine is picking all the residues to be anisotropic even though I would not try that at 2.1 A.
phenix.refine does not do it, unless you ask it to do. I'm confused at this point about what you have done. If you send me the data and model files, as well as any CIF files for ligands (if any), then I will send you back the most optimal refinement strategy for your specific case. Pavel.