[ccp4bb]: MR summary, ZN2+ as anchoring point

[Jeroen Mesters <jmesters@imb-jena.de>]

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

Hi ccp4bb community,

thanx for all the suggestions! Besides working on overproduction
and crystallization of the Se-Met protein, I am in the process of
trying several of the stategies suggested.

Thanx again, Jeroen.

--
Interested in crystallogenesis?
Visit http://www.conventus.de/iccbm9/
--

Here is the question with the suggestions:

The question:

> Hi ccp4bb community,
> 
> our protein contains two zinc ions for which we are able to pick up
> the signal. However, the phasing power is to low to solve the structure.
> With MR we also failed because the search model is less then half of
> the molecule with about 30 % seq. identity but also containing two zincs.
> 
> Is it possible to use the zinc ions as an anchoring point and rotate the
> search model around this axis?! Which program will do so?
> 
> THANX, JEROEN
> --

The answers from the community:

Juergen Bosch:
run SOLVE, RESOLUTION_STEPS 3 from let's say 20 to 4 A
Then run molrep in combination with those phases.

Fred Vellieux:
Ask Renaud Morales at IBS (renaud.morales@ibs.fr). We published a paper
concerning such an operation. Rotation about an axis defined by 2 Fe
sites in that case. Note that you must place the model in two opposite
positions, i.e + direction and - direction, and carry out the
search. Monitoring is with Rfree, and you eliminate all solutions with bad
packing.

Bart Hazes:
In the old XPLOR/CNS you could specify your rotations
explicitly and I guess that is still true. You'll have to find out the
direction of your rotation and then rotate around it in lets say 2.5 degree
steps.
An alternative is to generate the set of models yourself and use each one of
them for a translation search in for instance amore. This will require
a bit of scripting but could be more sensitive.
The real question is whether a successful solution of this problem is going to
give you an interpretable map. At < 30% identity and only 50% of the full
structure I think that's going to be tough.

Pete Dunten:
Did you try soaking a crystal in EDTA to remove the Zn, and
generate a set of isomorphous differences that way?  It's 
worked for me.
And if you do the rotation search, don't forget you need to consider
the two cases of ZnA from your model superimposed with Zn site 1
and site 2 from your data.

Bernhard Rupp:
If you want, I can try epmr followed by shake&warp to 
salvage a weak solution. we rebuilt complete structures from
less than 50%, but admittedly from reasonable models and 
decent data.

G. Birrane:
I had the same problem with a structure I am working on right now.
I can see two Ca2+ ions in the map but the solution does not give 
good phases to solve the whole structure. 
I agree with Bart, use Se-Met or other heavy atom methods to 
solve it.

Ho-Leung Ng:
If your data is of fairly high resolution and quality, try direct
methods. In particular, try the new OASIS program in ccp4.

Robert:
have you already tried to run MOLREP with your starting phases?
Theres another program BRUTEPTF which sounds interesting.
http://russel.bioc.aecom.yu.edu/server/NYSGRC.html

Yong TANG:
If you could compute an electron density map with the Zn-derived phases 
and the map is just good enough for you to identify the molecule boundaries, 
you could try to get the approximate center-of-gravity of you own molecule. 
In this way, you get three anchoring points and would, together with the two 
zincs and the c-o-g of the model (could be easily found by a simple run of 
moleman2), be able to determine the RT matrix by Site2RT in RAVE.

Rene de Jong:
Since you have only < 50% of a model, with 30% seq identity, MR is going to be
tough... Probably (or most likely...?) Randy Read's BEAST program will give you
better results than conventional MR progs, since it uses maximum likelyhood theory
in its MR functions. At a workshop in Como last June, Randy presented some
promissing figures on test cases, which were outstanding compared to results from
Amore, especially in non-trivial cases.  Ask him for program and details:
rjr27@cam.ac.uk

Eleanor:
Hmmm - you lose the chance to use FFT search functions then. Best to
verify your solutions by checking the Zn positions are consistent

James Naismith:
We have found BEAST from Randy Read to find solutions when we all but lost
hope! If you want to go down the random search method. A simple automated method
would be work out the rotations & translation to take your zincs to lie
along one axis (put a couple of points along this line and AMORE will do
this during it centre of mass). The simply apply an incremental rotation
around the axis (5 degrees around z if you do AMORE), ie cycle = 0,
cycle_now=cycle + 1, rot_now=cycle_now x 5, in lsqkab apply rot_now to your
centred model (which has had the same centre of mass centering rot & trans
as the zinc atoms in a line applied). The add another lsqkab which would
return the zinc atom line back to the correct position, this will move your
model to a point in the cell rotated around the two zinc atoms. You can then
test the model by some criteria (I would think packing first), then test it
by calculating R-factors or correlation coeffs. The longest part of the
whole business will be writing the script. My experience is that it almost
never works (partly because I only do stuff like this when things are
hopeless) but its very satisfying to get the script running. I would use
BEAST or do phased translation searches.


-- 
Jeroen Raymundus Mesters, Ph.D.                           Alias Geronimo
Dept. of Structural Biology and Crystallography,  Institute of Molecular
Biotechnology, P.O.Box 100 813, D-07708 Jena, Germany.
Tel: +49-3641-656063, Fax: +49-3641-656062, E-mail: jmesters@imb-jena.de
http://www.imb-jena.de/www_sbx/home.html
--
If you can look into the seeds of time and say
which grain will grow and which will not - speak then to me    (Macbeth) 
--