[phenixbb] -differences in refinement stats between phenix-dev-2219 and -1.9-1692

[Pavel Afonine]

Hi Ezequiel,

just to make it clear: torsion-angle NCS, Optimize X-ray/stereo weight 
are *not* the default settings, therefore you did customize the 
refinement parameters.

At this resolution (1.3A) you do not need to use NCS restraints.

Also, you say you used "Automatically add H's" however the PDB file you 
sent me does not contain any H atoms. Furthermore, the PDB file header 
says H are present:

REMARK MODEL CONTENT.
REMARK  ELEMENT        ATOM RECORD COUNT   OCCUPANCY SUM
REMARK        C                     1275 1047.00
*REMARK        H                     2029 1649.00 *
REMARK        O                      559 488.50
REMARK        N                      315 263.00
REMARK        S                        6 6.00
REMARK       Cu                        4 2.89
REMARK    TOTAL                     4188 3456.39

So overall this make an impression that something has gone wrong.

Way forward:

1) Take the file you sent me;

2) Add H atoms:

phenix.reduce model_no_h.pdb > model_with_h.pdb

3) Use new file with H atoms added to run refinement. The only setting 
different from defaults should be "Optimize X-ray/stereo weight". 
Perhaps you can also enable automated water update.

I just did all this myself and got nicely refined model with

Final: r_work = 0.1624 r_free = 0.1939 bonds = 0.010 angles = 1.030

Corresponding PDB and MTZ are sent in a separate email that goes to you 
only (off mailing list).

Pavel

On 12/6/15 07:59, Ezequiel Noguera wrote:
> Hi Pavel, thank you very much for your help. I attach the data. I 
> didn´t customize parameters. The refinement stategy of the last job 
> was: XYZ coords, Real-space, Individual B-factors, torsion-angle NCS, 
> Optimize X-ray/stereo wght and X-ray/ADP wght, Automatically add H's 
> and correct N/Q/H errors.
>
>
> Hi Ezequiel,
>
> deviations are not unexpected in general but not that large as you 
> report, so I wonder what's going on.. Could you please send me input 
> data and model files, as well as any ligand files you used. If you 
> customized some parameters, telling what were those would be helpful too.
>
> Please send all files off-list to my email directly.
>
> Thanks,
> Pavel
>
>
>> Hello, I'm working in a refinement that I started some time ago. I 
>> noted that repeating the same refinement task using a newer version 
>> of phenix, the refinement statistics Before and After refinement are 
>> different from values of the old version (I used "restore job" from 
>> the Job history to repeat the same task in the two versions installed 
>> in the same PC). Any explanation? I'm worried about the rms 
>> deviation. Thanks in advance.
>>
>> new version:
>> R-work: start=0.1697, final=0.1761
>> R-free: start=0.2011, final=0.2053
>> Bonds: start=0.046, final=0.029
>> Angles: start=1.321, final= 0.631
>>
>> old version:
>> R-work: start=0.1651, final=0.1657
>> R-free: start=0.2018, final=0.2009
>> Bonds: start=0.013, final=0.013
>> Angles: start=1.409, final= 1.411

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20151213/642594a3/attachment.htm>