[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ccp4bb]: Monoclinic twinning?
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Dear Matthew,
another program that you might want to try and which correctly indicated all
twinned crystals that I had so far is Todd Yeates's Crystal Twinning Server,
http://www.doe-mbi.ucla.edu/Services/Twinning/, which also points to Todd
Yeates's excellent paper about twinning. In my experience, the single
intensity statistic numbers that you reported are not very meaningful, better
are plots of the intensity statistics against resolution, what you have
already done. For monoclinic space groups, merohedral twinning can only occur
if the monoclinic angle is 90 degrees. From the denzo indexing, I think the
angle of 90.24 degrees is close enough to really consider 90.0 degrees. If
the monoclinic angle in fact deviates from 90 degrees, you will have
non-merohedral twinning, which means that the spots do not overlap exactly in
reciprocal space, and thus you should observe splitted spots at higher
resolution. In the case of merohedral twinning in a monoclinic space group,
you won't observe this. The symmetry R-factors for the hypothesis
"orthorhombic space group" are only considerably higher than for the
hypothesis "monoclinic space group" if the twin-fraction is much smaller than
0.5, whereas in the case of perfect merohedral twinning, i.e.
twin-fraction=0.5, they will be (almost) as good as for the monoclinic space
group. If you are confident enough that you have really twinned crystals, I
can think of three ways how to proceed:
(1) the usual advice is to grow untwinned crystals by either adding additives
(beta-octa-glucoside?, metal ions?) or by switching to different
crystallization conditions; in practice, this is often not easy/possible. In
my experience it might be worth to measure many crystals from the same
conditions and even from the same drop - individual crystals may differ a lot
in their twin-fractions
(2) if the twin-fraction is smaller than 0.5, you might try to de-twin the
data. For that you have to know the twin law and the twin fraction, both of
which are reported by the twin server mentioned above. The most crucial
parameter is the twin-fraction: here you might also want to inspect the
Britten-Plot produced by the CCP4 program DETWIN: the twin-fraction at which
the number of resulting negative intensities after detwinning starts to grow
rapidly is a good estimate and should be close to the number that the twin
server reports. But you must be aware that de-twinning does a very "brutal"
manipulation of both your intensities and your sigmas.
(3) If the twin fraction is small (~ 0.05-0.15) you might try to "ignore"
twinning at all. Maybe, the results for phasing and refinement will be better
than for de-twinned data. I had one case, where refinement of a model against
twinned data (twin-fraction ~ 0.15) gave better results than both refinement
against de-twinned data and twin-refinement against twinned data. But there,
I never tried to refine the twin-fraction. Personally, I think a program that
would correctly refine both the model together with the twin-fraction in a
maximum-likelihood way would be very helpful.
Anyway, I wish you good luck!
Dirk.
On Wednesday 05 February 2003 21:00, matthew.c.clifton.1 wrote:
> *** For details on how to be removed from this list visit the ***
> *** CCP4 home page http://www.ccp4.ac.uk ***
>
> Hi all,
> I may have come across an example of monoclinic twinning in my data that
> appears to be orthorhombic.
>
> The story:
> I am working on a MAD data set that appears to be in the space
> group P21221. However, upon running the data through SOLVE, the electron
> density maps are very difficult if not impossible to follow even though
> the data set was to 2.4 angstroms. I thought that I had the wrong space
> group and I proceeded to process the data as monoclinic.
>
> During the indexing of the data as P2, a strange problem occurred. DENZO
> had problems with the identification of the correct lattice (they varied
> between data sets taken from the same crystal in the exact same
> orientation).
>
> Example indexing in P2:
> Peak_minus data set 57.38 91.51 112.37 90 90.24 90
> Inflection_minus data set 91.53 57.4 112.36 90 90.24 90
>
> The data from the inflection wavelength could be "corrected" such that it
> found the same lattice as the Peak_minus data set by adding ~200 more
> peaks to the indexing, but it changed the unit cell parameters to:
> 57.15 91.36 111.06 90 91.04 90
>
> Following this, I decided to investigate the possibility of twinning. I
> ran the data through the CNS program detect_twinning where it output the
> following information:
> <|I|^2> / (<|I|>)^2 = 1.6711 (average value over all bins, where 2.0 is
> untwinned and 1.5 is for twinned).
> (<|F|>)^2 / <|F|^2> = 0.8602 (average value over all bins, where 0.785 is
> untwinned and 0.865 is twinned).
>
> When the data was processed through TRUNCATE, the Cumulative Intensity
> Distribution looked almost exactly like the example given on the CCP4
> website for a twinned crystal. The data appears to be clean with no large
> streaked reflections or the presence of a secondary lattice.
>
> Is the data truly twinned? If so, how is this happening geometrically? (I
> am having a hard time drawing a picture of this). The only example I can
> find that is similar is an abstract from the bioinformatics-structure
> conference in Jerusalem, Israel from November 17-21, 1996 (HiPIP II
> structure to 1.7 angstroms resolution from a twinned monoclinic crystal).
> Any suggestions on how to proceed from here would also be helpful.
>
> If not, why is all the data indicating that it may be twinned?
>
> Any suggestions or input would be appreciated.
>
> Thanks in advance,
> Matt
--
****************************************
Dirk Kostrewa
Paul Scherrer Institut
Life Sciences, OSRA/007
CH-5232 Villigen PSI, Switzerland
E-mail: dirk.kostrewa@psi.ch
Phone: +41-56-310-4722
Fax: +41-56-310-4556
WWW: http://www.sb.psi.ch
****************************************