[phenixbb] Unusually low B factors with phenix 1.8, but not with phenix 1.7 (data set has ice rings)

[Pavel Afonine]

Hi Oliv,

this was discussed in the past indeed, and I believe I addressed the 
problem. So what you describe surprises me.
To be on the same page, could you please send me the inputs necessary to 
reproduce the problem. Once I have the data I will have a look. Please 
send files to my email directly (not to the mailing list).

Pavel

On 2/27/13 10:44 AM, Oliv Eidam wrote:
> Hi,
>
> I would like to continue the B-factor discussion started by Demetres 
> D. Leonidas on Nov 2, 2012 on CCP4BB:
> http://www.mail-archive.com/ccp4bb@jiscmail.ac.uk/msg29012.html
>
> I make the same observations like Demetres and Mingzhu Wang: unusually 
> low B-factors when refining with Phenix 1.8-1069 and 1.8.1-1168, but 
> not with Phenix 1.7.1-743. In numbers: the average B-factor is 10 A2 
> but it should be 20 A2 (based on Wilson B and similar structures). A 
> few atoms (~20) have Bf = 0.00.
>
> The following is important: I refined dozens of similar structures 
> without that problem and I observe this phenomenon only for a single 
> data set. That specific data set is different because 1) I excluded 
> regions with ice rings in XDS and 2) I limited the resolution to 2.28 
> Angstroem, although the crystal diffracted to ~1.8 Angstroem. While 
> the preparation of the data set might be questionable in itself, I 
> would prefer that this would not be the topic of this thread.
>
> My questions are: What can I do to refine my structure with Phenix 
> 1.8.X using this truncated data set? What causes the difference in 
> average B-factor values between Phenix 1.7 and 1.8 for this data set?
>
> Your advice is much appreciated.
>
>   Oliv
>
>
>
> On 11/04/12 19:21, mingzhu wang wrote:
>> Pavel,
>>
>> 1.8.1-1168 also has this problem.
>>
>> I found this problem existed with some dataset, but not all dataset.
>>
>> 1.8.1 give better map, so I built model with 1.8.1 map and go back to
>> 1.7.3 before I deposit the structure to PDB.
>>
>> When I compared the B factor from 1.8.x and 1.7.3, I found
>> B(1.7.3)=B(1.8.x)+Boverall(1.8.x). Perhaps this is the problem.
>>
>>
>>
>> 2012/11/3, Pavel Afonine<pafonine at gmail.com>:
>>> Joao,
>>>
>>> I am using the latest version of phenix 1.8.1-1168
>>>
>>>
>>> 1.8.1-1168 should not have that "problem".
>>>
>>> If you suspect there is still a problem, you can send me the data 
>>> and model
>>> files off list, explain what exactly the problem is, and I will have 
>>> a look
>>> right away.
>>>
>>> FYI: there is Phenix mailing list for Phenix-specific questions, 
>>> where I
>>> can also explain in great details the behavior with the B-factors you
>>> observe.
>>>
>>> All the best,
>>> Pavel
>>>
>