[phenixbb] Phenix TLS refined model feeding into next round of refinement

Pavel Afonine pafonine at lbl.gov
Wed Jul 20 14:39:47 PDT 2011


> I used to think after one round of Phenix TLS refinement, if I need to 
> do further refinements after correcting couple residues, I need to 
> take out the ANISOU records from the pdb file. The reason is if not, 
> all protein atoms might be applied anisotropic individual B factor 
> refinement in the next round. Yesterday, I made a "mistake" directly 
> feeding the pdb file from the previous TLS refinement into the next 
> one. Seems it is fine. Is that true Phenix automatically recognize 
> what I wanted, that is, TLS only, not anisotropic individual B factor 
> refinement?

- it seems to me extremely weird if you do one refinement run with TLS 
and then the next one without TLS. I can't see a reason for this.

- by removing ANISOU records you completely invalidate the results of 
previous TLS refinement;

- phenix.refine tries to decide automatically where the input ANISOU 
come from. If input file contains ANSIOU and resolution is low enough 
and TLS refinement is not requested then phenix.refine simply removes 
ANISOU and continues. Obviously, this means the previous results of TLS 
refinement are all completely invalidated.

More details:

1) http://www.phenix-online.org/newsletter/
articles "TLS for dummies" and "On atomic displacement parameters (ADP) 
and their parameterization in PHENIX".


3) http://www.phenix-online.org/presentations/latest/pavel_phenix_refine.pdf


More information about the phenixbb mailing list