Hi Jianghai, phenix.refine writes out the total B-factors after TLS refinement (this is why each atom participating in a TLS group receives ANISOU record regardless the resolution). The total B is the sum of individual local B-factors (isotropic) + TLS contribution (anisotropic). Now if you take a model previously refined with TLS and refine just individual B-factors for this model (no TLS) then you don't need to do anything special to preserve the TLS information from the previous runs or read in the TLS matrices. This is simply because this TLS information is "stored" in total atomic B-factors (ANISOU records). One thing here you must be careful about. Like I said above, the model after TLS refinement contains ANISOU records. If you take such a model for individual B-factors refinement, phenix.refine will think that you are going to refine atoms with ANISOU records as individual anisotropic. To prevent phenix.refine from doing this you need to explicitly say that all atoms must be refined as isotropic (no matter what is in you PDB file). Here is how: % phenix.refine model_after_tls.pdb data.mtz adp.individual.isotropic=all For more information, details and examples, see new updated phenix.refine documentation here: http://phenix-online.org/new-site/documentation/refinement.htm We will work out something more automatic for this in future (including possibility to adjust wxu_scale when TLS is used). Thanks for your question and, as always, let me know if you have any other questions! Pavel. Jianghai Zhu wrote:
I will chime in. Is it possible for phenix.refine to get the TLS parameters from a previous run? then we can change wxu_scale value during an individual_adp refinement in the presence of fixed TLS parameters.
Jianghai