Correct solution to this would be to standardize tls records. Current PDB policy creates ambiguity that can only be resolved by looking for tls record in the header. So how exactly is the need to analyze header tls record avoided?
ANISOU records should only be used when individual Uij's were actually refined. As a pre-clarification, to my taste there is no value in exactly reproducing R-values. Cheers, Ed. -------- Original message -------- From: Nat Echols <[email protected]> Date: 04/21/2013 1:20 AM (GMT-05:00) To: [email protected] Subject: Re: ANISOU error when atom number exceeds 99999 On Sat, Apr 20, 2013 at 6:51 PM, Jinzhong Lin <[email protected]> wrote: On 04/20/2013 09:03 PM, Ethan A Merritt wrote: I really disapprove of the idea of writing out ANISOU records to describe TLS refinement. Notwithstanding the fact that the PDB recommends this, I strongly suggest that you tell refmac_not_ to write out ANISOU records. The TLS model is adequately described in the header, the extra ANISOU records are a less good way of describing the same thing. This would make sense if there was a standardized format for TLS which all programs that might care had implemented - which there isn't, as far as I know. As a result there are many entries in the PDB whose TLS records cannot be easily interpreted. A further problem is that not every program outputs REMARK records where TLS information is usually dumped, but most will preserve ANISOUs. Finally, most molecular graphics programs don't have an option to visualize TLS parameters, but most do support anisotropic ellipsoids. I agrees with you. But I am using phenix for the refinement, unfortunately there is no option for it not to write the ANISOU records. Since I assume you don't actually care about the ANISOUs for the purpose of adjusting your model in Coot, I suggest this: grep -v ANISOU model.pdb > model_iso.pdb You're going to start the next round of refinement with higher R-factors than if the ANISOUs were left alone, but in my experience they'll usually recover if you run TLS refinement again. -Nat
