Hi Jianghai,

I see now. The problem is that if you do individual B-factors refinement (no TLS) then the result will be affected by restraints. Ideally you would need to do this without restraints, but this is not practical.

To obtain the estimate you want I would simply do group B-factors refinement with one B per residue. In this case the restrains are not used. It would be good to compare the results with TLS and individual only refinement. Do it in two steps: re-set all B to Wilson B (or to current average), then do group B refinement:

1) phenix.pdbtools model.pdb set_b_iso=25
2) phenix.refine model_b25.pdb data.mtz strategy=group_adp
Note, the default is one B per residue. You can specify your own selections.

Pavel.


On 5/11/2008 7:00 AM, Jianghai Zhu wrote: 
Pavel,

Thanks.  I actually just want to do a very simple thing.  In the days  
before TLS refinement, we use B-factor distributions to find the  
ordered and less ordered regions in the structure, i.e.  low B- 
factors, ordered region and high B-factors, less ordered region.  I  
just wondered if I can still do that with the isotropic equivalent B- 
factors after a TLS refinement.  Or should I just get a real isotropic  
B-factors distribution by re-refining the model without TLS?

-- Jianghai






On May 11, 2008, at 2:03 AM, Pavel Afonine wrote:

  
Hi Jianghai,

    
If I compare the isotropic equivalent B-factor
after TLS refinement to the isotropic B-factor without TLS  
refinement,
will they have the same trend?
      
yes, only if:
1) the choice of TLS groups is correct;
2) the restraints on isotropic individual B-factors are not too tight.

    
I want to use the B-factor
distribution to describe the ordered and disordered region in the
structure, will the TLS refinement skew that?

      
I'm not sure what exactly you want to do here. But anyway, by using  
TLS
model you better describe atomic ADP: TLS models global domain motion
and individual B-factors model local atomic vibrations; that is each
kind of motion receives dedicated parameterization (which is good).
However, if you model everything with individual B-factors (no TLS)  
the
similarity restraints are always used and they tend to remove TLS
contribution making final B-factors less adequate.

Pavel.


_______________________________________________
phenixbb mailing list
[email protected]
http://www.phenix-online.org/mailman/listinfo/phenixbb

    

_______________________________________________
phenixbb mailing list
[email protected]
http://www.phenix-online.org/mailman/listinfo/phenixbb