[phenixbb] Statistics for pdb deposition

Pavel Afonine pafonine at lbl.gov
Wed Aug 11 13:01:55 PDT 2010

  Hi Elwood,

> I'm trying to deposit a model to the pdb database, but there are a few required values in the Refinement Statistics section that I'm not sure where to find.
> 1,2) Cut-off sigma (F), Number unique reflections (observed)
> Looking at the log file, it appears that the cut-off is 0, which results in 330 reflections being rejected. The pdb header, however, lists the number of reflections at a value corresponding to 352 reflections being rejected. Why are these numbers different and which should I report?

The phenix.refine log file reports the number of reflections found in 
your input data file and the number of reflections used in refinement.

The latest version of phenix.refine uses all Fobs>=0 (in older versions 
it was Fobs>0). Also, phenix.refine automatically detects and removes 
Fobs outliers. Typically it is a small number between 0 and 10-20, 
sometimes slightly more. This is all reported in log file. Look for 
"Outliers rejection" section of log file.

The information in PDB file header (REMARK 3) is what was actually used 
in refinement and this is what should be used for deposition.

If you send me the files (off list) I will be able to comment on 
specific numbers.

> 3) R-factor (all)
> R-factor (all) includes both the working and test sets in addition to reflections rejected by the sigma cut-off. Does PHENIX calculate this value? I know R-factor (observed), which includes the working and test sets but not rejected reflections, is calculated.

Although it is a meaningless value, phenix.refine still reports it, for 

REMARK   3   R VALUE     (WORKING + TEST SET) : 0.1783

I presume this is what you mean by "R-factor (all)".

Let me know if you have any questions.


More information about the phenixbb mailing list