Problems with Phasing

[Chris Hosfield <chris@crystal.biochem.queensu.ca>]

***  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,

I am having some trouble producing interpretable electron density maps 
and was wondering if anyone can offer some suggestions:

I am trying to combine information from two MAD data sets (both on 
Selenium), a gold-derivative, and a native data set.  The MAD data and 
gold-data are to 2.8 A resolution, and the native data is to 2.6 A.

To this point, I have been able to determine the selenium positions both 
by anomalous patterson methods (to locate a few sites) or by direct 
methods (which locates all the sites).  The gold sites were found both by 
isomorphous difference pattersons and by cross-phasing with Se-derived 
phases.  

I have only been able to get stable MAD refinements using SHARP or CNS, 
where I typically get Rcullis ~40-70%, Phasing Power ~1.5-2.8 for the 
isomorphous and anomalous contributions, with an overall FOM 
approximately 0.55 to 0.6.  

There is an appearance of a solvent boundary with one hand and not 
the other prior to solvent flattening with these FOM's, which is 
encouraging, and there is some connectivity within these regions, but the 
overall map quality is very poor, with no secondary structure elements 
detectable.   Solvent flattening (solomon/dm) does improve things somewhat, 
but not nearly enough.  Analyzing the maps with ESSENS produces the 
same results I see by eye...nothing that great.
  

Addition of the gold-native contribution in the SHARP refinement also 
seems to help slightly (at least in terms of producing a slightly cleaner 
protein/solvent boundary), even though the refinement statistics are 
somewhat poorer (which is surprising). 

The gold-native SIR statistics are roughly Rcullis ~75%, Phasing power 
~1.2, FOM ~0.3 on one major site and two minor sites.

I am at a loss to explain why the map quality is not better, considering 
the statistics that we have seen.  The raw data collected is generally 
90% complete, with Rsyms between 4-5%, with redundancy about 6.   
Data were processed using DENZO/SCALEPACK, and were then run through 
TRUNCATE and SCALEIT.


I don't think it is a problem with the sites (17 Se) because they were 
found both by patterson (~3) and direct methods (all 17).  The Se-phases 
were good enough to cross phase the gold data (12 sigma peak for the 
strongest site) which agreed with the isomorphous difference pattersons 
(gold-native).  Additionally, the SIR phases (using SOLVE) from gold-native 
were good enough to cross-phase out 11 of the 17 Se sites, and the 
remaining 6 were readily identified in residual maps after SHARP refinement.
Additionally, SHARP refinement with bad "test" sites clearly identifies 
them as incorrect (occupancies refined to zero).

There is about 900 residues (100 kDa) per a.s.u. and the space group is 
P21. 

Can anyone suggest any way I might be able to improve my map 
quality...perhaps in terms of phase combination to include both MAD data 
sets, the native and the gold derivative.  I have tried this in SHARP but 
the resultant maps were again not interpretable.


I thank you for your time and consideration,

Sincerely,

Chris Hosfield

-------------------------------------------------------------------
Chris Hosfield                     chris@crystal.biochem.queensu.ca
Graduate Student	   
Dep't of Biochemistry
Queen's University		   6cmh2@qlink.queensu.ca
-------------------------------------------------------------------