[ccp4bb]: Summary: refine occupancies with Refmac

[Joerg Mueller <J.Mueller@Uni-Koeln.de>]

***  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 All,

here is a short summary concerning my last question to the CCP4 BB.

Question:
> is it possible to refine occupancies for alternate conformations using
> REFMAC and how does it work?
> Or is there an other CCP4 program which can do this?

The general answer is no (at the moment).
But there was a discussion if this refinement is necessary or really
possible.




Eleanor Dodson:
 No - it isnt, but the B values of the 2 conformations give you a very
good indication..
 Do people think it is likely that the data can support such refinement
unless you have very high resolution? 


Harry Powell:
Even with data going out to 0.7A there's a strong correlation between
B-factor and occupancy, so I wouldn't refine occupancies and B-factors
in
the same refinement run without the use of something like similarity
restraints for the B's and/or constraining the sum of the occupancies to
unity (for small molecule structures this is exactly what I do); SHELXL
should allow you to do this easily.


Bart Haze:
Individual occupancy refinement for all atoms indeed does not make sense
as a value of 1.00 is a very good estimate for the fulk of our models. 
However, you normally only want to refine occupancies for small groups
of
atoms for which you have reason to believe that they have non-zero
occupancies. That is for sidechains build in multiple conformations and
bound ligands. There remains the problem that B-value and occupancy are
highly correlated in refinement, but if B-values increase to
unreasonable
values it may be better to fix the B-value to a reasonable value and
refine the occupancy. 


S. Ravichandran:
        I totally agree that the occupancy refinement remains
*meaningless* unless you have a very high (say < 0.8 A) resolution
dataset with good completeness and R-merge.

        But, these days there were several examples of high resolution
datasets being collected (see for example, the Feb. issue of ACTA D..
viz.
the location of valence electron at 0.96 A resolution !) where one will
be tempted easily for testing the occupancy refinement. 

        One way of handling this case (may be) to fix the B-values prior
to the occupancy refinement, as there is a strong dependency of B with
the atomic positions (occupancies). SHELXL can handle this (bit tricky).
Why can't there be an option in REFMAC now?   (for those 0.1% of people
who are successfull in collecting the high(est) resolution data)..



David J. Schuller:
i think it should be possible if one maintains a few reasonable
constraints:
a) the occupancies for all alternates should sum to 1.0
b) the B factors for all alternates should be the same

perhaps you could allow manual override for the adventurous



Regards
Joerg

-- 
Joerg Mueller                 Department of Biochemistry
Phone +49-221-4706443         University of Cologne
                 6455         Zuelpicher Strasse 47
Fax   +49-221-4705092         D-50674 Koeln