Wednesday, 29 April 2009

CCTBX, going for it

Sometime in the next couple of releases I will probably move to using tools from the CCTBX library. This is free and available for all of the xia2 supported platforms, so should not present a problem, and will allow the xia2 program itself to have access to much more in the way of crystallographic functionality.

It can also replace a fair amount of the code I have written already, which can't be a bad thing!

Monday, 27 April 2009

New feature on the way

A common complaint I get is that xia2 does not use e.g. all 4 cores in a 4 core machine. Well, it does if you tell it with -parallel N where N is e.g. 4. So anyway, xia2 should be clever, so the next release will figure out how many cores (if you don't tell it) and use all of them by default.


xia2 released

Hi Folks,

I have just put on the server - this is the one which includes the following new features:

Changes since

- Fixed merging statistics bug - I/sigma output were very slightly different.
- Now extend FreeR column if copying from lower resolution input file.
- Detector limits now correctly specified around the detector centre for Mosflm, rather than the beam centre. Only useful if your beam centre is a fair way from the image centre...
- Parallelised integration with Mosflm - divides the sweeps into an appropriate number of chunks and then sorts together the resulting reflection files. -parallel N, remember.
- Added a putative 'small molecule' mode, which will autoindex with Mosflm from a modest number of images, for smaller molecules where Labelit is unhappy with the number of good Bragg reflections.

As the parallelisation was a major feature, have bumped the version to 0.3!

Thanks as always to everyone who provides feedback - many of the developments above result from bug reports and xia2 development would go nowhere without help from you the users!

All the best,


Monday, 20 April 2009 Progress

Quickly hammering out some of the remaining dents which are preventing the release of xia2 - this will be the one which supports parallel integration, has the fix to the statistical summary calculation from XDS, works with Python 2.6 &c.

Obviously there are a few changes, which by extension raises the likelihood of bugs - however as always I will welcome any comments.

Small molecules

Had a request to get xia2 working with a small molecule (mP 6.18 22.39 11.65 90.00 99.10 90.00) which Labelit did not like as there were too few reflections to be useful. Have accordingly hacked together a putative small molecule mode - this will use quite a few images for autoindex with Mosflm, but is not yet clever enough for small molecule spacegroups (i.e. knows noting of inversions!)

Will be in

Friday, 3 April 2009

Parallel Mosflm's

So, this is on it's way, for those really big data sets you want to process in lots of chunks... it is now done "by magic" - or will be in the next release:

29681 gw56 R 92.1 0.5 1:16.95 ipmosflm-7.0.3
29683 gw56 R 92.1 0.5 1:18.64 ipmosflm-7.0.3
29684 gw56 R 90.3 0.5 1:18.66 ipmosflm-7.0.3
29685 gw56 R 90.3 0.5 1:20.13 ipmosflm-7.0.3
29687 gw56 R 88.5 0.5 1:18.30 ipmosflm-7.0.3
29680 gw56 R 84.9 0.5 1:21.23 ipmosflm-7.0.3
29686 gw56 R 84.9 0.5 1:18.59 ipmosflm-7.0.3
29682 gw56 R 83.0 0.5 1:17.74 ipmosflm-7.0.3

Just benchmarking now, but for big data sets with lots of images it's got to be a fair amount faster. Numbers will follow...


Just for kicks, to keep up with these modern times, have started a twitter feed so people can see what's happening. Are your bugs being fixed? Best way to find out.

xia2 and CCP4 6.1 reprise

So, it is clear that the dependence on xia2 being told the version of CCP4 it is using is a pain, and is in no way future proof, so I have now fixed this. No more '7_Sigma' errors, then.

So you no longer need export XIA2_CCP4_61=1 or -ccp4_61...

Any other annoyances like this, please drop me a line!