[sharp-discuss] SHARP with multiple SAD datasets and ignore isomorphous differences?

[Clemens Vonrhein]

Hi Francis,

On Wed, Jun 01, 2011 at 04:44:45PM -0600, Francis E Reyes wrote:
> I'm curious about using multiple SAD datasets with poor isomorphism  
> among them. I imagine I can specify multiple C-sites and multiple G- 
> sites for each respective compound.
> 
>   However, how do I tell sharp to ignore the isomorphous differences?

You can't. After all, each dataset needs to see each other for scaling
etc anyway.

What I would do:

  * try with SAD of each dataset to see how good each one is on its
    own

  * try to see if some are more isomorphous to each other than
    others. 

  * Also: maybe the non-isomorphism is a function of radiation damage:
    so maybe also process your datasets more restrictive (only the
    first half or so); for phasing the completeness isn't that crucial
    - especially if this way you still have reasonable isomorphism and
    you can 'stich' a complete set of phases together.

  * maybe you can get a few 2-wvl double SAD jobs going?

  * if non-isomorphism is too big (what R-factors between datasets are
    we talking about for the 40-6A range? 20%, 30% or higher?): take
    the HLA-D columns of one SHARP run (eden.mtz) and use them as
    'external phase information' in a separate SHARP run

  * if non-isomorphism is still too large: congratulations! You might
    be able to do multi-crystal averaging ;-)

As usual with those cases, the book-keeping is crucial: all datasets
indexed consistently, find groups of datasets that match better with
each other etc.


> Also, should scaling parameters be estimated/refined at all?

Yes: you might need to not estimate/refine scaling B-factors if you
have very low resolution though.

> I imagine that each compound will have its own Multiplier and
> isotropic scale factor.

Yes.

> Also, I imagine NANO_BGLO and NANO_CLOC should be refined as well?

Yes ... and watch them to see if non-isomorphism isn't completely
killing your signal (and making parameter refinement instable).

Combining different datasets in real space (density
modification/averaging) is often the only way here ...

Cheers

Clemens

> 
> Thanks!
> 
> F
> 
> ---------------------------------------------
> Francis E. Reyes M.Sc.
> 215 UCB
> University of Colorado at Boulder
> 
> 
> 
> 
> 
> _______________________________________________
> sharp-discuss mailing list
> sharp-discuss at globalphasing.com
> https://www.globalphasing.com/mailman/listinfo/sharp-discuss

-- 

***************************************************************
* Clemens Vonrhein, Ph.D.     vonrhein AT GlobalPhasing DOT com
*
*  Global Phasing Ltd.
*  Sheraton House, Castle Park 
*  Cambridge CB3 0AX, UK
*--------------------------------------------------------------
* BUSTER Development Group      (http://www.globalphasing.com)
***************************************************************