[phenixbb] getting started with anisotropic refinement

[Pavel Afonine]

Hi Mark,

> I need some help advice regarding anisotropic refinement, which I've 
> never done before. My current Rfactors are R-work = 0.1587, R-free = 
> 0.1726, resolution 1.26 Å. 

two things I would do right away without thinking long:

- add H atoms;
- refine all macromolecular atoms with anisotropic ADPs (probably, 
except alternative conformations), and solvent with isotropic ADPs;

Also, please make sure to use recent Phenix version. That may have 
impact in your case, since latest phenix.refine have improved handling 
of hydrogen atoms (*), and a new bulk-solvent and anisotropic scaling 
procedure (**).

If you still want to define by hand which atoms to refine 
isotropically/anisotropically, then a parameter file like this should 
allow to do this:

refinement {
   refine {
     adp {
       individual {
         isotropic = (chain A and resseq 1:100) or (chain X and resseq 
300:900) or water
         anisotropic = (chain Z and resseq 123 and name CG1) or (element Au)
       }
     }
   }
}

Make sure you cover all atoms with

isotropic = ...
anisotropic = ...

selections, because the default behavior is that if you change these 
parameters then phenix.refine assumes you know what you do. If you do 
not select all atoms with these selections, then B-factors of atoms that 
are not selected will not be refined. I don't like this and I will 
change this once I get a chance.

For your specific example:

refinement {
   refine {
     adp {
       individual {
         isotropic = not (chain A and (resseq 7:44 or resseq 61:149 or 
resseq 173:182 or resseq 202:254 or resseq 263:290))
         anisotropic = chain A and (resseq 7:44 or resseq 61:149 or 
resseq 173:182 or resseq 202:254 or resseq 263:290)
       }
     }
   }
}

I've never tried but this might work too:

refinement {
   refine {
     adp {
       individual {
         aniso_sel = chain A and (resseq 7:44 or resseq 61:149 or resseq 
173:182 or resseq 202:254 or resseq 263:290)
         isotropic = not $aniso_sel
         anisotropic = $aniso_sel
       }
     }
   }
}

Phil experts may comment on this more.

Let me know if you have any questions or need any help with this!

Good luck,
Pavel

(*) "On contribution of hydrogen atoms to X-ray scattering"
http://www.phenix-online.org/newsletter/

(**)
http://www.phenix-online.org/presentations/PhenixSantaFe2012_PA.pdf

> I have TLS groups, and know I need to exclude these from individual 
> anisotropic refinement.
>
> Questions:
>
> 1) how to decide my atom selection for anisotropic refinement. I've 
> identified segments with MC B below 20, but what to do for surface 
> residues which have well ordered MC but B factors for side chains 
> increase (e.g. along a well ordered Lys side chain) - should (can?) I
>
> a) refine only MC anisotropically,
> b) refine whole residue anisotropically,
> c) refine isotropic
>
> 2) How to write a a parameter file for phenix.refine defining the atom 
> selections.
>
> The residues I'd like to use for anisotropic refinement are: A7-A44 
> A61-A149 A173-A182 A202-A254 A263-A290 all in chain A
>
> I found following on phenix.refine online docs, but am not sure to to 
> extend this for a more complex selection needed above, which Ideally 
> I'd put into a file to be read in when I start refinement. Could yo 
> show me how to do for A7-A44 and A61-A149, and I'll follow example for 
> the rest.
>
> adp.individual.anisotropic="resid 1:2 and not element H" 
> adp.individual.isotropic="not (resid 1:2 and not element H)
>
> Thanks for help.
>
> Mark
> .