[ccp4bb]: Summary - MAD data with radiation damage

[Tina Bakolitsa <bakolitsa@burnham.org>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

Thanks to all those who replied. Here's a summary of what I received on the
subject:

* Frank von Delft suggested getting in touch with Zbyszek Otwinowski
(author of DENZO) who is now working on incorporating a radiation decay
model into scaling (to include in SCALEPACK). He also suggested a single
wavelength (peak) and shorter exposures explaining that resolution is not
important since high redundancy will bring out the signal. In addition,
changing the crystal orientation in the beam can help scale out systematic
error and help improve data quality. He also referred me to the local
scaling option in SCALA.

* Savvas N. Savvides suggested trying SAD phasing with inflexion or peak
data alone using CNS and SOLVE as he has found SHARP not to perform so well
in SAD.

* Stephen M. Soisson suggested systematically cross-checking the Se sites
using difference Fouriers and local scaling using SOLVE.

* Jeff Vargason referred me to

"Single-wavelength anomalous diffraction phasing revisited"
Acta Cryst. 2000 D56, 1413-1420

and suggested collecting a complete peak first.

* Steve Ernst reported a similar situation, a very weak anomalous signal
for 3 or 4 Se's out of a possible 10, with the sites found by SOLVE
resulting in a FOM that was in the twenties. He achieved useful improvement
by changing to SCALEPACK v.1.96 and using the undocumented option
"absoprtion". This improved the signal significantly and produced maps
that, using FFFEAR, showed several helices and some beta structure.

* Felix Vajdos suggested culling the data back from the end of the data
sweep (I've tried that by the way, but don't have sufficiently high
redundancy to really afford it and anyway, the data were collected in
10-degree wedges one wavelength after the other). He quotes Shelx-D as the
best program for finding the sites (I agree) and SHARP as the best
refinement program (ditto). His suggestion is to pick the top 10 heavy atom
sites using the single best wavelength available (the top sites will give
sufficient preliminary phase information to orient the heavy atom vector
roughly correctly), refine in SHARP and run the default 50 cycles of
Solomon solvent flipping. If the maps are still uninterpretable, repeat
data trimming. The point about picking the top ten sites (or five or twelve
or whatever) is that you just need phases that are good enough for the
density modification (SOLOMON in this case) to be able to work.

* Kris Tesh (Rigaku) generously offered to test my crystals on Rigaku's new
Cr K-alpha anode/FRE detector/R-AxisIV generator/osmic optics/helium cone
setup that has already helped phase several structures using the sulfur
anomalous signal. For Se, the anomalous signal would be 1.14 electrons on
Cu K-alpha and 2.28 electrons on Cr K-alpha. He stated the advantages of an
in-house source (with respect to a synchrotron) as less flux, more stable
beam and more sensitive detection; in addition, the newest optics allow for
higher brilliance (resolution obtained was only 0.2 Å less than at
SPRING-8). And since I'm not promoting Rigaku, I'll stop here :) For more
information on these results, he referred me to this year's ACA and IUCR
presentations from Chen Yang and BC Wang (C67 in IUCR abstracts).

Kris also explained that radiation damage can have different profiles
depending on whether it is the main-chain or the side-chains that are being
clipped off (in the former case both lower and higher resolutions will be
affected, in the latter, mainly the higher) requiring different correction
models. He suggested trying the B-factor option available in d*TREK but
expressed reservations regarding its effectiveness.

* Martin Hallberg suggested using the 10-degree wedges as separate batches
in SHARP thus making it possible to partly compensate for the
nonisomorphism caused by crystal deterioration. For getting rid of outliers
he suggested using the MAD-scaling protocol found in SCALA modified to
output each 10-degree wedge as a separate dataset, then merge them all in
CAD and start SHARP. He also suggested SAD phasing with the wavelength
having the largest anomalous differences introducing the other one once
things have started looking stable/good; this strategy was crucial in one
of his cases for MAD-phasing in SHARP.

To summarise the summary:

If you're stuck with these data:

* Try local scaling (SOLVE, SCALA) and absorption corrections (SCALEPACK
v.1.96)

* Providing you have sufficiently high redundancy, try leaving out some of
the data then pick the top/best sites using a single wavelength, run
SHARP/SOLOMON and check for map interpretability.

* Systematically cross-check all Se sites using difference Fouriers

* Try and see if SAD can get you anywhere

If you can recollect:

* Use shorter exposure times, high redundancy and go for the peak wavelength.

* If resolution and crystal number/quality allows, try phasing by S or Se
using Cr K-alpha (2.29 Å) and the latest state of the art in-house
optics/detectors.


Thanks again for your time and help - Loretta especially ;)

	Tina

----------------------------------------------------------------------
Tina Bakolitsa (PhD)

Lab 5107
The Burnham Institute			Tel: (858) 646 3100 ext. 3939
10901 North Torrey Pines Road		Fax: (858) 646 3196
La Jolla, CA 92037
-----------------------------------------------------------------------