Dear Charles,
Eleanor is right: look at the images, and look also at the
processing output produced by XDS. If you find that too ASCII-looking,
then look at the graphs produced from these numbers by programs like
XDSGUI, XDSAPP and others. Our own autoPROC produces an extensive html
output (called "summary.html") that will contain all these graphs.
In general you have to look at several of them simultanously to
figure out what went wrong. Your type of problem is often produced by
a bad crystal centring that causes the crystal to move out of the beam
in some ranges of images. In that case you can see that the number of
indexed spots as a function of image number goes down to worryingly
low values in certain ranges, and that the image scale factors and
B-factor have large excursions for those same ranges, that are also
the ranges where the Rmerge values end up quite high. In this case it
is just because spots on images made systematically weaker (and
therefore noisier) by the loss of centring had to be scaled up by
CORRECT/AIMLESS.
If such is the case with your dataset, there is not much you can
do. Your only hope is to be lucky enough to have enough redundancy in
your raw data that the reflections on the useless images will have
been measured through symmetry equivalents on images for which the
crystal centring was good.
It could also be that your crystals are very thin plates.
With best wishes,
Gerard.
--
On Thu, Nov 09, 2017 at 07:53:40PM +0000, Eleanor Dodson wrote:
> I think you need to look at the images.
> We found one case where the overall Rmerge didnt look too bad but there
> were horrendous streaks across many images. No idea what had happened but
> other crystals from the same batch were much better.
>
> Eleanor
>
> On 9 November 2017 at 19:26, CPMAS Chen <cpmas...@gmail.com> wrote:
>
> > That is right. I had the data already and did not want to throw it away.
> >
> > On Thu, Nov 9, 2017 at 2:09 PM, Eleanor Dodson <eleanor.dod...@york.ac.uk>
> > wrote:
> >
> >> I think you need to worry about why that has happened, rather than get an
> >> automated rejection criteria!
> >> There must be some problem in the data collection for that to happen..
> >>
> >> Eleanor
> >>
> >> On 9 November 2017 at 19:04, CPMAS Chen <cpmas...@gmail.com> wrote:
> >>
> >>> Hi All,
> >>>
> >>> Is there a way to reject diffraction images based on Rmerge?
> >>>
> >>> When I processed my data with XDS, I use AIMLESS in CCP4 to get merged,
> >>> truncated data. However, there is quite some images with high Rmerge, say
> >>> larger than 1. Is there a keyword I can use to reject these images based a
> >>> Rmerge cut-off, say 0.6?
> >>>
> >>> Thanks!
> >>>
> >>> Charles
> >>>
> >>> --
> >>>
> >>> ***************************************************
> >>>
> >>> Charles Chen, Ph. D
> >>>
> >>> Research Associate
> >>>
> >>> University of Pittsburgh School of Medicine
> >>>
> >>> Department of Anesthesiology
> >>>
> >>> ******************************************************
> >>>
> >>>
> >>
> >
> >
> > --
> >
> > ***************************************************
> >
> > Charles Chen
> >
> > Research Associate
> >
> > University of Pittsburgh School of Medicine
> >
> > Department of Anesthesiology
> >
> > ******************************************************
