[phenixbb] advice for refinement with anomalous information

[Pavel Afonine]

  Hi Vincent,

- I would add that ion and refine its occupancy and anisotropic ADPs... 
Oh, I see, the resolution is 3.4A, well, then I would try first 
isotropic ones.
- if the data set you are using contains Fobs(+) and Fobs(-), (and not 
merged Fobs_merged = 0.5*(Fobs(+) + Fobs(-))), I would refine f' and f'' 
of the anomalous scatterer.

After having done this, check if the residual density is clear, and keep 
in mind that unbalanced syntheses may have Fourier truncation ripples 
around heavy atom, see relevant discussion here:
http://www.mail-archive.com/ccp4bb@jiscmail.ac.uk/msg14835.html

and some pictures here:
http://www.phenix-online.org/presentations/latest/pavel_maps.pdf

In general, getting rid of an unmodeled density is good in a sense that 
it has a potential to improve the density everywhere else and you may 
better model other some other parts of your structure, so you get a 
better model overall.

Let me know if you need any help with trying the above suggestions.

Good luck!
Pavel.


On 8/11/10 5:56 PM, vincent Chaptal wrote:
> Hi,
>
> the refinement of my structure brings me to a new point where I would
> appreciate any input.
> - I solved the structure by molecular replacement.
> - The refinement looked good, I was almost ready to deposit but there
> always was a fairly big blob of density that i couldn't explain, loosely
> coordinated to the protein and far from the active site.
>
> To try to identify what it could be, i calculated an anomalous
> difference map (data collected at 1A) and I see a peak right in the
> middle of this density.
> I have BaCl2 in my crystallization solution so it is the likely guilty
> ion! (and with further look, it looks like a loose metal site)
>
> So now, I want to refine with Barium in the density and I was wondering
> what was the best way to do it, and since there is anomalous signal, why
> not use it?
> * I updated to the new version of phenix and it reads I+ and I- (while
> the previous version didn't) but under the refinement settings, the
> target function ML-SAD is still for developers only. Is it a good idea
> to use the anomalous Is for refinement, or:
> * should I re-run Phaser-EP to have phases to use in phenix.refine, or:
> * should I not bother, the signal is weak anyway (1 barium/ ~400
> residues) and continue with my non-anomalous scaled dataset and use the
> anomalous signal only for a figure to explain why I put the metal here.
>
> (resolution of 3.4A, quite anisotropic)
>
> I can also try everything but I'd like to understand what would be best.
> Thanks very much for your help
> vincent