Re: refmac for partial p-ala trace / refinements

[Harry Powell <harry@mrc-lmb.cam.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 Wed, 12 Jan 2000, Johan Turkenburg wrote:

> model that was only about 50% of the structure I was trying to solve.
> I had data to 1.9A, but the stats on the data (I/sig(I) and Rsym) would
> indicate that the data beyond 2.3A was barely significant.
> 
> The model was incomplete in various ways: side chains, loops and a
> whole domain.
> 
> An initial map was reasonable for the atoms in the mol rep model, but 
> showed very little density for the missing 50%. Even missing side chains 
> in the model did not come up convincingly. 
> 
> Many cycles (something like 100) of refmac/arp later the map was 
> gorgeous, and a model was build in under two hours!
> 
> Repeating this at 2.3A did not work anywhere near as well, and the 
> model could not have been build without further intermittent
> refinement.
> 
> I'll certainly not try to theoretise on why that was the case. But
> next time you have a case like this, give it a go. And your
> high res data should of course always be brilliant, but when it is not, 
> don't throw it away straight away -:)

Um, it's worth saying that even very weak reflections have real
information content (and there will be plenty of these between 1.9 and
2.3A) - including those with negative intensities. A good published
example of this (in the small molecule world) came last year (sorry, I
don't remember the journal, but it was one of the IUCr's...)  with
structures of Emycin antibiotics, where the only reliable information
about the correct stereochemistry in the X-ray data was in the very weak
reflections. I think one of the authors was someone called George
Sheldrick... 

Harry 
-- 
Dr Harry Powell, MRC Laboratory of Molecular Biology, MRC Centre, Hills
Road, Cambridge, CB2 2QH