Re: Significance of atomic B's

[Eleanor Dodson <ccp4@yorvic.york.ac.uk>]

***  For details on how to be removed from this list visit the  ***
***    CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html    ***

On Oct 26,  9:00am, Edward Berry wrote:
> Subject: Significance of atomic B's
> ***  For details on how to be removed from this list visit the  ***
> ***    CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html    ***
>
> Dear CCP4'ers,
>
> I have a question about the significance of the atomic B-factors
> in a refined coordinates file. Are the absolute values significant,
> or are only the relative values significant and all values can be
> shifted up or down by a constant depending on whether and how
> the data were scaled during data reduction?
>

 This is of course true; usualy you scale to the first image, and all others
tend to be scaled up to match it. And if that first image is somehow deviant
you will do something damaging. For instance - it is never a good idea to take
the low resolution sweep from a synchroton as the basis; that over all B factor
will not be as well determined as that from the highest resolution sweep.

> If the latter, would it be better to avoid scaling the overall
> B-factor of the data so that the atomic B-factors in the final
> coordinates file reflect the situation in the crystal? It seems to
> me that if you have a dataset with a high B-factor but you scale
> the data, either by absolute Wilson scaling (e.g. in truncate) or
> by scaling against an initial model with all B-factors set to 20
> (B-overall in Xplor), then refinement against that dataset will
> lead to artificially low B-factors. Or do refinement programs
> have some trick to get "true" B-factors from arbitrarily scaled
> data?


 Truncate does not APPLY the overall B factor - it just estimates it.

 And I cannot believe that XPLOR does this either - most scaling algorithms
apply the value of k to the Fobs, and correct the Fcalcs ( and the model) by
the OverallB correction.


 But it is possible to do something ugly at the data processing stage!

 And also when you merge two data sets; eg for MIR or MAD phasing, there must
be an overallB correction applied to one..


 But no amount of scaling can "invent" data, and the B factors will be no
better determined whatever you do..

Eleanor Dodson

-- 
------------------------------------------------------------------
Eleanor J.Dodson, Chemistry Department, University of York, U.K.
Tel: Home +44 (1904) 42 44 49, work:  +44 (1904) 43 25 65
------------------------------------------------------------------