[phenixbb] Average B-factor and ML error

[Leo Sazanov]

Dear Pavel,
Thanks for comprehensive reply - since refinement itself was done some 
time ago, I forgot to look now into "REMARK stage"...
Leonid

Pavel Afonine wrote:
> Hi Leo,
>
>> Hi, phenix.refine outputs "WILSON B" value into PDB header after 
>> refinement, which is not the same as average B-factor of all the atoms 
>> in the structure, and not exactly the same as Wilson B-factor from SCALA 
>> scaling. How exactly is Wilson B-factor for phenix header calculated 
>
> Wilson B is computed using Peter's procedure as described in:
>
> Zwart, P.H., Grosse-Kunstleve, R.W. & Adams, P.D. (2005). CCP4 
> newsletter. No. 42.
>
> or in one of there references in this paper. Peter may comment on this 
> some more.
>
>> and 
>> would it be possible to output there also the average B-factor of all 
>> the atoms in the structure?
>>   
>
> It is reported by phenix.refine many times in many different places:
>
> - in PDB file header:
>
> REMARK            
> |-----overall-----|---macromolecule----|------solvent-------|
> REMARK   stage    b_max  b_min  b_ave  b_max  b_min  b_ave  b_max  
> b_min  b_ave
> REMARK    0    : 100.80   7.48  28.20 100.80   7.48  27.43  78.39  
> 12.05  40.43
> REMARK    1_bss: 101.17   7.85  28.57 101.17   7.85  27.80  78.76  
> 12.42  40.80
> REMARK    1_xyz: 101.17   7.85  28.18 101.17   7.85  27.80  58.46  
> 14.33  35.27
> REMARK    1_adp: 113.33  11.63  27.68 113.33  11.63  27.44  55.19  
> 16.55  32.10
> REMARK    2_bss: 113.47  11.78  27.82 113.47  11.78  27.58  55.33  
> 16.70  32.24
> REMARK    2_xyz: 113.47  11.78  27.93 113.47  11.78  27.58  55.28  
> 14.61  34.43
> REMARK    2_adp: 114.56  11.66  27.71 114.56  11.66  27.46  52.95  
> 16.40  32.23
> REMARK    3_bss: 114.64  11.73  27.79 114.64  11.73  27.54  53.03  
> 16.48  32.30
> REMARK    3_xyz: 114.64  11.73  27.86 114.64  11.73  27.54  57.12  
> 13.65  33.94
> REMARK    3_adp: 115.09  11.32  27.64 115.09  11.32  27.42  53.24  
> 16.30  31.94
> REMARK    3_bss: 115.15  11.38  27.68 115.15  11.38  27.48  53.31  
> 16.37  31.63
>
> - in log file:
>
> |-ADP 
> statistics--------------------------------------------------------------|
> |    Atom    | Number of   | Isotropic or equivalent| Anisotropy 
> lmin/max     |
> |    type    |iso    aniso | min     max     mean   | min   max    
> mean       |
> |    - - - - |- - - - - - -| - - - - - - - - - - - -| - - - - - - - - 
> - -     |
> |    all     : 3549   0      11.38   115.15  27.68    None  None   
> None       |
> |    all(noH): 3549   0      11.38   115.15  27.68    None  None   
> None       |
> |    Sol.    : 169    0      16.37   53.31   31.63    None  None   
> None       |
> |    Mac.    : 3380   0      11.38   115.15  27.48    None  None   
> None       |
> |    Hyd.    : 0      0      None    None    None     None  None   
> None       |
> |    - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 
> -      |
> |       Distribution of isotropic (or equivalent) ADP for non-H 
> atoms:        |
> |    Bin#      value range     #atoms | Bin#      value range     
> #atoms      |
> |      0:    11.384 -  21.760: 1339   |   5:    63.266 -  73.643:   
> 30        |
> |      1:    21.760 -  32.137: 1407   |   6:    73.643 -  84.019:   
> 24        |
> |      2:    32.137 -  42.513:  442   |   7:    84.019 -  94.396:   
> 18        |
> |      3:    42.513 -  52.890:  200   |   8:    94.396 - 104.772:   
> 18        |
> |      4:    52.890 -  63.266:   68   |   9:   104.772 - 115.149:    
> 3        |
> |                               
> =>continue=>                                  |
> |-----------------------------------------------------------------------------|
>
> - finally, you can type:
>
> phenix.pdbtools model.pdb --show-adp-statistics [optionally, give a 
> CIF file]
>
>> Also, deposition into PDB requires only Luzzati coordinate errors, while 
>> phenix produces only ML based ones. Presumably ML estimate is better 
>> than Luzzati?, 
>
> ML error estimate is better than Luzzati one, indeed. That's exactly 
> why phenix.refine outputs ML error estimate. And even ML error 
> estimate has it's own known problems (not just in phenix, but in 
> general), so it shouldn't be taken too literally.
>
> phenix.refine outputs PDB file almost ready for PDB deposition, and 
> "almost" means here that you have to remove the upper part of the 
> header till where the REMARK 3 section starts.
>
> Pavel.
>


-- 
Dr. Leonid A. Sazanov
Research group leader
Medical Research Council
Mitochondrial Biology Unit
Wellcome Trust / MRC Building
Hills Road
Cambridge
CB2 0XY
WEB: www.mrc-mbu.cam.ac.uk
Tel: +44-1223-252910
Fax: +44-1223-252915