Hi Leo, message from Ralf: Ralf apologizes for the accidental reply.

But the model's B factors in some region are very high, higher than 150 or more. I remembered that the B factor can be limited in the refinement using Refmac. Can I do the similar thing in Phenix

The quick answer - no (and it is intentional). phenix.refine uses restraints in ADP refinement, as all other programs. The upper limit for B-factors that they can refine to is set somewhere in the code to prevent numerical issues and it is ~500 (if I re-call correctly). If your B-factors refine to 150 or more that means there is a reason for this: disorder or misplaced atoms. In this case artificially restricting B-factors doesn't seem like a good idea. Using TLS (with properly selected TLS groups) or reviewing these model regions look like a much better idea. You can also experiment trying to repeat your refinement after re-setting all B-factors to some constant value (Wilson B, for instance), and allowing enough of refinement macro-cycles to achieve the previous R-factor values. Pavel.