[phenixbb] Failed to refine B factors of some residues and ligands

Christian Roth christianroth034 at gmail.com
Mon Oct 31 10:55:12 PDT 2016


Not more than 5 I'm afraid. I don't know if it would find the right B's 
using 10 cycles and more, but that also increases the running time quite 
considerably even on some of our faster machines.

Best wishes

Christian


Am 31.10.2016 um 16:46 schrieb Dorothee Liebschner:
> Hi Christian,
>
> OK, it looks you found a way to prevent the positive density by using 
> "realistic" starting B values (of the order of that of surrounding 
> protein atoms).
> Did you also try to increase the number of macro-cycles?
>
> Best wishes,
>
> Dorothee
>
> On Mon, Oct 31, 2016 at 9:11 AM, Christian Roth 
> <christianroth034 at gmail.com <mailto:christianroth034 at gmail.com>> wrote:
>
>     Hi Dorothee
>
>     to answer your questions
>
>     The resolution is about 2.0 Ang and the average B around 55. The
>     Bfactor from the ligand was set to the default of 20 by Coot when
>     placing the ligands.
>
>     It seems that setting the B-value to the average B of the
>     structure nearly solves the problem (most of the ligand fine, just
>     parts with unusuallly high values). Setting it to a bit higher
>     values 70 for a start and the ligand refines fine.
>
>     I noticed that in the first macrocycle with and initial factor of
>     20, the B's are nearly unchanged (20 +- 2) Red diff density
>     appears as expected. In the second cycle The minimizer let the
>     B-factors "explode" to about 200 and that basically is the
>     overkill and it never recovers.
>
>     HTH
>
>     Cheers
>
>     Christian
>
>     Am 31.10.2016 um 15:59 schrieb Dorothee Liebschner:
>>     Hi Christian,
>>
>>     Could you answer the following questions:
>>
>>     - what is the resolution of the data set?
>>     - what is average B of the protein?
>>     - what are the starting B values for the ligand?
>>     - did you try lowering starting B for the ligand and repeat the
>>     refinement? Is this behavior reproducible with different
>>     refinement options?
>>
>>     Best wishes,
>>
>>     Dorothee
>>
>>     On Sun, Oct 30, 2016 at 11:06 AM, Christian Roth
>>     <christianroth034 at gmail.com <mailto:christianroth034 at gmail.com>>
>>     wrote:
>>
>>         Hi Phenix Team,
>>
>>         regarding that ligand issue in this thread, we observed the
>>         same with a ligand in one of our structures now. After
>>         refinement the ligand is surrounded by pos diff density and
>>         the B-Factors are around 200 (way above any useful range in
>>         that case).
>>
>>         The cif was generated using elbow and phenix runs with
>>         options Real space refinement, TLS on, individual B, and
>>         optimizing weights. Nothing crazy.
>>
>>         Could you find a solution for the previous case and give us
>>         some tips?
>>
>>
>>         Cheers
>>
>>
>>         Christian
>>
>>
>>
>>
>>         Am 06.09.2016 um 01:44 schrieb Pavel Afonine:
>>>         Hi Tongqin,
>>>
>>>         I second Dorothee's point that we need more information to
>>>         resolve this problem. If you still have this problem please
>>>         send me inputs files (data, model, .eff from last
>>>         refinement, as well as ligand .CIF files if any) and I will
>>>         look into this once I have files.
>>>
>>>         Pavel
>>>
>>>         On 8/31/16 09:29, Dorothee Liebschner wrote:
>>>>         Hi Tongqin,
>>>>
>>>>         It is a bit difficult to diagnose with the information you
>>>>         provided. Could you please answer the following questions?
>>>>
>>>>         - Did you also try to reset the B-factors to similar values
>>>>         than neighboring atoms?
>>>>         F.ex. the average B-factor in the model could be 50 A**2,
>>>>         but in the ligand region, it could be lower, let's say 20
>>>>         A**2. Then the starting value using average B is quite far
>>>>         from the likely B-factor of the ligand.
>>>>
>>>>         - What refinement strategy do you apply? How many
>>>>         macro-cycles? Do you use any non-default parameters for
>>>>         B-factor refinement?
>>>>
>>>>         - Which Phenix version are you using?
>>>>
>>>>         Best wishes,
>>>>
>>>>         Dorothee
>>>>
>>>>         On Wed, Aug 31, 2016 at 7:12 AM, Zhou, Tongqing (NIH/VRC)
>>>>         [E] <tzhou at mail.nih.gov <mailto:tzhou at mail.nih.gov>> wrote:
>>>>
>>>>             Dear All,
>>>>
>>>>             I am refining a structure with diffraction to 2.2A
>>>>             resolution and 95 % overall completeness. Now the Rs
>>>>             are at 18% and 23%, respectively. But the program
>>>>             failed to refine B factors for several ligands, atoms
>>>>             had high B factors while showing positive Fo-Fc map
>>>>             around them. Occupancy was set to 1 and I also tried to
>>>>             reset B to mean B of the whole PDB. Any of you seeing
>>>>             this phenomenon? Thanks!
>>>>
>>>>             Shown below is HEPES that has high B factor and
>>>>             positive density around it after refinement:
>>>>
>>>
>>>
>>>
>>>         _______________________________________________
>>>         phenixbb mailing list
>>>         phenixbb at phenix-online.org <mailto:phenixbb at phenix-online.org>
>>>         http://phenix-online.org/mailman/listinfo/phenixbb
>>>         <http://phenix-online.org/mailman/listinfo/phenixbb>
>>>         Unsubscribe:phenixbb-leave at phenix-online.org
>>>         <mailto:phenixbb-leave at phenix-online.org>
>>         _______________________________________________ phenixbb
>>         mailing list phenixbb at phenix-online.org
>>         <mailto:phenixbb at phenix-online.org>
>>         http://phenix-online.org/mailman/listinfo/phenixbb
>>         <http://phenix-online.org/mailman/listinfo/phenixbb>
>>         Unsubscribe: phenixbb-leave at phenix-online.org
>>         <mailto:phenixbb-leave at phenix-online.org> 
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20161031/fd2cca0a/attachment.htm>


More information about the phenixbb mailing list