Dear All,
Thanks for all the suggestions, on and off the board!
Summary:
Some have asked for the L-statistic in P 4 21 2:
Mean |L| :0.397 (untwinned: 0.500; perfect twin: 0.375)
All programs tried (xtriage, truncate, pointless) agree that the 4Å data
is likely twinned P4 with an estimated twin-fraction ranging from 0.38
to 0.48.
People seem to agree (as was my initial understanding) that twinning by
itself cannot change a primitive lattice to a body-centered lattice.
Thus this change in my low-resolution dataset must be caused by
something else.
Pseudo body centering has been suggested as the likely explanation for
the primitive to body-centered lattice change in the low-resolution
dataset.
- As suggested, I have looked at the self-patterson for the 4Å dataset
and see no peaks, at 1/2,1/2,1/2 or elsewhere.
- As suggested, I have looked at the truncate output of the table
"Analysis of mean intensity by parity for reflection classes" in the
h+k+l column, but see no differences from h+k+l=2n to h+k+l=2n+1 in the
4Å dataset.
- As suggested I have processed the 4Å data at 8Å to see if pseudo body
centering was breaking down at higher resolution. At 8Å there was still
not sign of pseudo body centering.
Thus the 4Å data does not not support pseudo body centering, as far as I
can tell.
Some has suggested that the crystals are simply two different things.
This might be, but since the low-resolution dataset has exactly the same
unit-cell parameters as the 4Å dataset, it seems unlikely to me that the
crystal packing is significantly different, or that the content of the
crystals differ. It seem likely that both crystals contain the same
thing in approx. the same type of packing.
So no clear explanation for the observed behavior so far, but most
likely the low resolution is obscuring the real problem in this
particular instance.
As is almost always the case, the way forward is to get more and better
data to understand the problem.
As one suggested offboard (tongue-in-cheek): "Why not find crystals that
are not twinned, probably with higher resolution?"
I will do that, and thanks again for your input!
All the best,
-Bjørn
On 06/04/2014 03:09 AM, Eleanor Dodson wrote:
It helps to look at the output from the truncate step quite critically.
First is there a non cryst translation of 1/2,1/2,1/2 indicated in the
P4 2i2 data set?
If so then the I centring at lower resolution might just be approximate..
If there is NC translation then other twinning statistics are distorted
and I find the only semi-reliable one is the L test.
But if you say there is no room for your protein with that translation
and 4/mmm symmetry then there must be twinning or you have crystallised
something else!
Eleanor
On 4 June 2014 08:48, <[email protected]
<mailto:[email protected]>> wrote:
Dear Bjørn,
I guess the first step to enlightment is to recognize that we as
mere mortals are not able to deduce the space group from diffraction
data alone. All Aimless, XDS etc. can produce are educated guesses
what the space group might be. Especially when twinning is involved,
the crystal packing may not heed the rules and classifications that
we humans try to impose. In many cases, one might have to go down to
P1 and solve the structure in P1 to find out what the true space
group is.
Here are some comments to your questions:
-the same protein under the same crystallization conditions and even
in the same drop may produce crystals with very different crystal
packings, even with the same unit cell, so your 4 and 7.5Å crystals
may be different.
-If there is no way to fit the protein in the asymmetric unit that
is a very strong indication that you do have twinning.
-There have been some discussions in the CCP4BB, but I do not
believe that twinning can generate body centering.
-You might be barking at the wrong tree and the twinning axis might
be parallel to the 4-fold axis, or even generating the 4-fold. You
may even have 4-fold twinning.
-You may have pseudo body centering, which is perfect at low
resolution, but breaks down at higher resolution. As a test, you
could process your 4Å data only to 7.5Å and see what the statistics
would look like.
What I would do:
If you have more crystals, collect data on them all, maybe there is
one which is not or not perfectly twinned.
If there is a model which could be used for molecular replacement:
process the data in P4, I4, P222 and P1 and run molecular
replacement with all possible space groups for both crystals.
However, at 4Å with unclear twinning, solving your structure will be
tough.
Best,
Herman
-----Ursprüngliche Nachricht-----
Von: CCP4 bulletin board [mailto:[email protected]
<mailto:[email protected]>] Im Auftrag von Bjørn Panyella Pedersen
Gesendet: Dienstag, 3. Juni 2014 21:01
An: [email protected] <mailto:[email protected]>
Betreff: [ccp4bb] possible twinning issue in P4212 / I422
Dear All,
I have a strange potential twinning issue that I cannot understand.
I've searched high and low on all the internets to find an answer
but have come up empty-handed, so I look to the wisdom of The Board
to enlighten me.
I have a 4'ish Å dataset that processes nicely in P 4 21 2 (#90).
However intensity distributions indicate possible almost perfect
twinning (eg. <I^2>/<I>^2 : 1.592 ). So I speculate that the real
space group might be P 4 (#75).
Recently we collected a new fairly low resolution (7.5Å) dataset,
from the same type of crystals (same purification, same conditions).
But the space group in XDS and aimless now comes out very clearly as
either I422 (#97) or I4212 (#98) (screw-axis is unclear given the data).
The unit-cell parameters are exactly the same as in sg #90, which btw.
means that in the body-centered lattice there is no way the protein
can fit in the asym. unit.
So I guess what I don't understand is: Is it possible to go from a
primitive lattice to a body-centered lattice by twinning. Is this
just a low-resolution artifact? Or is this a P4 unitcell that can
appear like
P4212 or I422 depending on small variations (weak dehydration or
similar).
Has anyone experienced something similar? Am I missing a basic facet
of how twinning works, or is something else at play here?
Thanks for any insights or suggestions!
All the best,
/Bjørn
--
Bjørn Panyella Pedersen
Macromolecular Structure Group
University of California, San Francisco
