Re: aniso scaling

["Eleanor J. Dodson" <ccp4@ysbl.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    ***

naismith@st-andrews.ac.uk wrote:
> 
> ***  For details on how to be removed from this list visit the  ***
> ***    CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html    ***
> 
> Just a comment
> 
> I am always unhappy about removing positional ncs retraints at all.
> Agreed there may be real differences but these can be satisfactorily
> treated by excluding them from restraint. Domain Domain movements are
> dealt with by having two matrices of restraint. Improperly restraining
> NCS will mess things up no doubt. So in that case removing the restraint
> makes things better, but often things could be even better by specifying
> the groups for restraint with more precision.
> 
> Removing NCS restraints however will generate differences but are they
> real? 


 It is easy to find cases where they are real - look at atomic
resolution proteins where there is no need for them. Vrystal packing has
a much greater influence than we like to admit! 
> 
> As to the B-factor, as has been said the Wilson plot tells you what you
> should expect. This can be misleading where you have processed free
> space rather than data but is a good guide.
> I think the problem is that we grew up with an idea high
> res = low B. This was based on our experience of crystals which diffracted
> in capills in the lab 2.0A. These are rocks compared to frozen crystals
> which diffract to 2.1A at the synchrotron (which under equivalent experimental
> conditions may be lucky to get to 3A). I have seen (and a casual trawl of
> the PDB shows this to be far from unique) what in the past would be a
> problem B of 40+ at 2.2A for frozen xtals collected at the synchrotron.



 I dont think the PDB is a reliable source of information about B
factors! the scaling algorithms in XPLOR inevitably mean that most low
resolution structures has totally unrealistic bfactors - a plot of mean
B versus resolution actually indicates a negative correlation between
<B> and <Resln>! 

 Eleanor