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


Hi,

Dirk as always has a good point - we've had the same issue once or twice
when trying to do molecular replacement on beta-barrels (GFP relatives),
symmetrical helix bundles, beta-helices, and other structures that have
internal quasi-symmetry. A hallmark of this kind of problem is the presence
of multiple close solutions, usually clustered so that the search model is
rotated or rotated/translated around the quasi-symmetry axis. There are two
things that help: 1. Generate packing based on multiple 'close' solutions
and compare packing efficiency (visually) - usually the correct packing
solution will not have many (sometimes none) bad intermolecular contacts and
2. Automatically refine all of the top 30-40 solutions because chances are
that one of them is probably ok. The third option is to do a molecular
replacement incorporating real space elements - specifically a rotation
search around the main axis of any of the 'close' solutions.

Artem
Ex-PGP

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Dirk
Kostrewa
Sent: Tuesday, April 04, 2006 3:38 AM
To: Kerr, Iain; CCP4BB
Subject: Re: [ccp4bb]: Molecular replacement in C2/I222

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


Dear Iain Kerr,

this sounds very interesting. From what you say, it is quite obvious 
that your true space group is I222 with probably one monomer in the 
asymmetric unit (estimated) and that you have a clear molecular 
replacement solution. You are, however, stuck with the refinement, 
probably in some false local minimum. It is very difficult to make an 
analysis from the distance, but I would like to point you to 1-2 
potential pitfalls with TIM-barrels. Although your search model has 75% 
identity, from my experience (with DHPS, a TIM-barrel), there are two 
problems with TIM barrels. The first problem is that you have a kind of 
pseudo-8-fold internal symmetry with TIM-barrels that might got you 
stuck into an overall correct placement, but rotated by n/8 (n=1,7) with 
respect to the true solution. The other problem lies in the variability 
of TIM-barrels: by the time a tried to solve DHPS, David Banner (who 
solved the very first TIM-barrel structure) analysed the PDB for 
TIM-barrels, and found ~50 different entries, _all_ of which had 
different relative arrangements of beta-strands and alpha-helices when 
superimposed! Therefore, I would have two suggestions for you:
To get out of a potential false local "8-fold" minimum, I would try to 
refine a a poly-ala model and see which side chain electron densities 
come back. It could be advantagous, to reduce the resolution to 3 A, 
first, to allow for a larger radius of convergence and then increase the 
resolution again. You could also try torsion-angle simulated annealing 
refinement with CNS. Maybe, this solves your problem already.
If this doesn't help, you could try several different related 
TIM-barrels from the PDB as search models for MR. Maybe, you will get an 
even better solution that helps you out.

This is all that comes to my mind at the moment.

Good luck,

Dirk.

Kerr, Iain wrote:
> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
> 
> 
> I have a dataset that merges very nicely in C2 and I222 (Rmerges ~5% and
~8% resp), 100% complete to 1.75A. The protein is a TIM-barrel enzyme, a
couple of loops are expected to be in different orientations or missing with
respect to the search model. The diffraction is very clean and appears to
represent a single lattice. The crystal is from a cluster with the same
symmetry but which presented multiple lattices upon diffraction...I just
recently managed to separate a single crystal and was very excited when I
saw the diffraction at the synchrotron....
> 
> The search model and target sequence have 75% identity and so I was
expecting this to be a routine MR problem. I have mutated the sequence of
the search model anyway to match the target sequence (I got tired of
mutating the 25% non-identical whilst fitting MR solutions..). Two loops are
deleted (~30 resdiues). I find oustanding solutions easily in both
spacegroups (no solutions in I212121), using a number of programs including
MOLREP, PHASER and CNS. None of these refine (REFMAC5 and CNS). For the best
solutions I have fit the model to the density and the refinement does not
improve (currently stuck at Rfree ~44%, Rfactor 40%). There are no obvious
large errors excepting some missing loops residues (~20 still). The best
solution (MOLREP) shows some difference density for one of the missing loops
and waters and overall the maps are good. I have checked for merohedral
twinning (C2) and pseudo-translational NCS; neither are present.
> 
> The indexing output from MOSFLM:
> 
>  PENALTY
> 7 158     mP    55.55    82.57    82.27    90.1 109.9 109.5   P2,P21
> 6  12     oI    55.55   102.35   116.68    89.6  89.8  90.0   I222,I212121
> 5  10     mC   116.44   116.68    55.55    89.8 118.5  89.8   C2
> 4   8     mC   129.04   102.35    55.55    90.0 115.3  90.3   C2
> 3   5     mC   154.73    55.55   102.35    90.0 131.1  89.8   C2
> 2   2     aP    55.55    82.27    82.57    90.1 109.5 109.9   P1
> 1   0     aP    55.55    82.13    82.57    76.8  70.5  70.4   P1
> 
> I have tried solutions 6 and 4 (solution 4 is found by HKL2000 and I tried
this first after I222). I have still to try 3 and 5 (3 looks non-standard
with the large beta angle though, ..), which I just noticed when I
re-indexed using MOSFLM today...at least with a similar crystal I have tried
P1 and didn't get an MR solution. 
> 
> In lieu of running another hundred MR jobs with multiple spacegroups I was
wondering if anyone had encountered anything like this before ?. Is there
something about C2 I am discounting ? I was expecting this problem to
disappear, having separated a single crystal from the cluster and therefore
removed the additional lattices from the diffraction pattern, but this is
apparently not the problem...
> 
> Thank you very much in advance for any help.
> 
> Cheers,
> Iain
> 
> 
> 
> 
> 
> 
> 


-- 

****************************************
Dirk Kostrewa
Paul Scherrer Institut
Biomolecular Research, OFLC/110
CH-5232 Villigen PSI, Switzerland
Phone:  +41-56-310-4722
Fax:    +41-56-310-5288
E-mail: [EMAIL PROTECTED]
http://sb.web.psi.ch
****************************************

Reply via email to