Ok, Graeme basically described the way I usually do things as
well. That's the beauty of a modular program like XDS and the passing
of information in ASCII format. Old-fashioned, but very, very useful
;-)

I've created a short page with a recipe for using XDS in that way -
there might be some typos still in there, but it might give you the
idea:

  http://strucbio.biologie.uni-konstanz.de/xdswiki/index.php/Indexing

Cheers

Clemens

On Mon, Feb 18, 2008 at 01:43:43PM -0000, Winter, G (Graeme) wrote:
> Hmm...
> 
> Thinking about this it may not be too hard to do with XDS. May involve a
> certain amount of fiddling though.
> 
> If you do the first stages of XDS processing, namely XYCORR, INIT,
> COLSPOT & IDXREF you will end up with a spot file (SPOT.XDS) from
> colspot and with indices attached from IDXREF. Once you have processed
> the strongest lattice, you could remove the spots indexed for that "run"
> from the SPOT.XDS file from COLSPOT and rerun IDXREF - this should the
> index the remaining reflections and give you an XPARM file you can use
> for integration. As Clemens says, the measurements may be suboptimal,
> but it is certainly worth a go and should not really be that hard. I
> suspect a certain amount of plaing with sed and grep here will be
> necessary.
> 
> XDS *will* complain a lot though saying that it can index < 70 % of
> reflections. This is to be expected.
> 
> You can give XDS the unit cell and symmetry if known and it will use
> these for clustering the difference vectors generated from the spot
> positions. In your case this is a good idea, as there will be a lot of
> noise (which difference vector indexing is quite sensitive to) from the
> spots from the other lattices.
> 
> What I would definitely do is run once with the strongest lattice, then
> copy the files to another directory for the processing of the second
> strongest lattice and so on.
> 
> Good luck,
> 
> Graeme 
> 
> -----Original Message-----
> From: CCP4 bulletin board [mailto:[EMAIL PROTECTED] On Behalf Of
> Clemens Vonrhein
> Sent: 18 February 2008 13:35
> To: [email protected]
> Subject: Re: [ccp4bb] indexing multiple lattices
> 
> On Mon, Feb 18, 2008 at 01:58:59PM +0100, Jason Greenwald wrote:
> > I have a crystal that grows with multiple lattices despite the fact 
> > that the crystals look more or less like single crystals under the 
> > microscope.
> > 
> > With a lot of guesswork I am able to index with mosflm by carefully 
> > selecting spots that appear to be in the same lattice.
> > Now I would like to use the known cell and symmetry to index more 
> > datasets that have the same "twinning" problem.
> 
> This is real fun to do - you can even index/integrate these several
> lattices: gives you higher multiplicity etc. But (there's always a
> but): be careful in using the data. It is very likely that a very large
> proportion of reflexions overlap - either completely or partially. There
> are ways of classifying these reflections, but a better way is to use a
> program that can process these twinned crystals (and no: neither XDS nor
> MOSFLM can ... at least not without hacking the source code or doing
> some other weird stuff).
> 
> >  Is there any way to do a search of crystal orientation matrices with 
> > a known cell to find the best fit to the diffraction pattern?
> > The data were collected on the Pilatus6M detector so I am limited to 
> > mosflm and XDS for processing.  Both packages always trying to index 
> > the crystal without using the known cell and symmetry to get the 
> > orientation matrix.
> 
> Not quite true: in both packages you can give the cell and symmetry. I'm
> not sure how it is handled internally exactly by those programs - but it
> seems to me that MOSFLM will use this information only after
> autoindexing to pick a solution closests to the user-given
> cell/symmetry. I _think_ XDS might use that info a bit earlier during
> indexing. But I could be wrong on both accounts.
> 
> > I thought of writing a script to do a full search (testing all 
> > orientations sequentially) in mosflm but I thought that I should 
> > submit to the experts before I try anything.
> 
> If you want some more detailed 'recipe' for processing non-merohedrally
> twinned crystals with XDS, please let me know directly. However, if
> you're planning to use that data for experimental phasing on a medium to
> small signal: don't get your hopes up. It might be enough for molecular
> replacement (not that MR wouldn't be helped with good data as well ...).
> 
> Cheers
> 
> Clemens
> 
> -- 
> 
> ***************************************************************
> * Clemens Vonrhein, Ph.D.     vonrhein AT GlobalPhasing DOT com
> *
> *  Global Phasing Ltd.
> *  Sheraton House, Castle Park
> *  Cambridge CB3 0AX, UK
> *--------------------------------------------------------------
> * BUSTER Development Group      (http://www.globalphasing.com)
> ***************************************************************
> 

-- 

***************************************************************
* Clemens Vonrhein, Ph.D.     vonrhein AT GlobalPhasing DOT com
*
*  Global Phasing Ltd.
*  Sheraton House, Castle Park 
*  Cambridge CB3 0AX, UK
*--------------------------------------------------------------
* BUSTER Development Group      (http://www.globalphasing.com)
***************************************************************

Reply via email to