From non-twinned to twinned?

From: Zhiyi Wei
Date: 4 January 2012 13:42


Dear all,

I recently collected a dataset (~2000 frames) from a single crystal.
If merge first 600 frames (sca1) or last 600 frames (sca2), Rmerge
values from scalepack seem to be ok (~10%) though rejection ratios are
high (~5%). But if I merge all frames together, Rmerge value goes up
to ~20% and rejection is extremely high (~20%). Then, I checked sca1
and sca2 by xtriage in phenix. Surprisingly, the logfiles told me that
sca1 is no twining while sca2 is very likely to be twinned. I never
met this case before. So, I am wondering if it is possible from a
non-twinned structure to a twinned structure just due to radiation
damage. If the answer is yes, does it mean that I should not collect a
large number of frames to amplify anomalous signals by using this
crystal?

Thanks a lot!

Best,
Zhiyi

----------
From: Jacob Keller


Need more info: how many degrees per frame? Also, on integration, do
various stats change over the 'sets?

JPK

--
*******************************************
Jacob Pearson Keller
Northwestern University

----------
From: Jens Kaiser


What are your cell constants and space group? It sounds to me you
misindexed and then artificially "twinned" your structure by
integrating/merging in too high of a symmetry. I've seen that happen for
primitive hexagonal which was actually C-centered monoclinic.
Also, in my experience this is more likely to happen with
denzo/scalepack as it refines every image and does postrefinement in
scalepack. XDS will decide for you on a symmetry after integration, and
in MOSFLM you should see earlier that something is wrong.

HTH

Jens

----------
From: Zhiyi Wei


Thank you for all the quick reply. Here are the additional information
about the crystal. The crystal was frozen and shot in synchrotron. The
collection was 1degree/frame. The space group and unit cell are P21
and  a=70 b=165 c=170 alpha=90 beta=90.1 gamma=90. I processed the
data in P1 first (I did not see any abnormal changes during
integration), and then merge them using P21 or P2221. P21 gives
reasonable Rmerge of ~10% while P2221 has much higher Rmerge of ~30%.
I reprocessed the data with mosflm using P1. And Pointless also
suggested the space group of P21.

Zhiyi

----------
From: Zhiyi Wei

Thanks. Your suggestion remind me that I forgot to mention one more
information. Actually, I did try to shot different regions of crystal
to reduce radiation damage. Say, the sca1 set used one region and the
sca2 set used another. Is it possible that this two regions has
different crystal domain arrangement (one is normal and another is
twinned)? The crystal looks like a nice single crystal.

Zhiyi

On 1/5/12, Carlos Frazao wrote:
> Hi
>
> If the beam cross section is significantly smaller then the crystal
> dimensions, when you rotate the crystal you will be constantly
> illuminating new volumes of the crystal. If the crystal contains
> differently oriented "single crystal domains" (that is indeed a twinned
> crystal) then you may have orientations where only one of those "single
> crystal domains" are being measured, and other orientations where
> multiple "single crystal domains" are collected. The same applies if you
> apply translations at the crystal (to collect data from  fresh crystal
> regions to decrease the radiation damage effect).
>
> Hope it helps, cheers,
>
> Carlos

> --
> **************************************
> Dr. Carlos Frazao

----------
From: Ed Pozharski


Absolutely.  This will, of course, vary for different crystal systems,
but from what I have seen it appears that a different crystal section
removed by ~50 micron behaves statistically as if it is a different
crystal.  This is with "high quality" crystals, so for more regular
situation this can be even shorter.

As far as "nice single crystals" go, looks may be deceiving. Twinned
crystals often look great.

Cheers,

Ed.

--
Oh, suddenly throwing a giraffe into a volcano to make water is crazy?
                                               Julian, King of Lemurs

----------
From: ccp4


I wouldnt expect radiation damage to cause twinning!

1) maybe you have different indexing for the 2 passes? Have you used
pointless to check if this could happen (depends on SG and cell dimensions)


Ask for "match reference set in input and give it sca1 as input 1, then
check sca2

2) use pointless to examine quality of symmetry matches before merging -
maybe you have the wtrong SG?

Eleanor