Newsletter contents... UP

MOSFLM - Recent changes and future developments.

Andrew G. W. Leslie

MRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, UK.

MOSFLM has undergone a number of improvements since the last CCP4 Newsletter described the implementation of the STRATEGY option in 1996. The algorithms employed in the STRATEGY option, which facilitates the design of an efficient data collection scheme, have been revised to avoid occasional problems with higher symmetry space groups that were present in the original implementation. The option is still most valuable when collecting data at a synchrotron, and particularly if multiple crystals are required to collect a complete data set. It proved invaluable when collecting data from crystals of Hepatitis B core particle at the ESRF, where a total of twenty crystals were used. Irrespective of how much data had already been collected, it only one to two minutes to determine what rotation range should be used for a new crystal in order to maximise data completeness. This includes auto-indexing the new image and running the strategy option. Contrary to a recent report in the literature (Ravelli, R.B.G., Sweet, R.M., Skinner, J.M., Duisenberg, A.J.M. and Kroon, J., J. Appl. Cryst., 30, 551-554, 1997) the facility to complete a partial dataset has always been part of the STRATEGY option.

The program can now process images from a wide variety of image plate and CCD based detectors. Recent additions have been the Mar 345 detector (both compressed and uncompressed formats), the Mar CCD detector (in use at the ESRF) and the ADSC Quantum 4 detector (soon to be installed on station 9.6 at Daresbury, and already in use at SSRL, CHESS, ALS and APS). These complement the existing formats for Mar, Raxis II, Raxis IV and Mac Science DIP scanners. The flexibility of the keyworded input allows a wide range of possible image formats to be accommodated without making any changes to the code (the SCANNER keyword). A previous limitation that the two-theta axis (for a swung-out detector) had to be parallel to the fast changing direction in the image has been removed, in order to deal with images collected at non-zero two-theta angles at the D2AM beamline at the ESRF and with the ADSC Quantum 4 detector.

A new feature introduced at the request of users is the ability to write a "savefile" containing the current processing parameters. This is designed to make it easier to set up a background processing job having auto-indexed and integrated a number of images interactively. The file contains keywords defining the current values of the processing parameters, including, for example, the name of the file containing the orientation matrix. It should only be necessary to add the appropriate PROCESS keyword defining the images to be processed when submitting a background job. The "savefile" can be written at any stage by using the SAVE keyword, or the option is provided by default when "Exiting" from an interactive session.

Several limitations present in previous versions have now been removed. These include the previous maximum reflection index of 255 (there is now no limit), and the maximum reflection width of ten images for partially recorded reflections (this is now 100 images). A significant number of minor bugs have also been corrected, and the output to the Xdl_view windows (when running interactively) has been improved. Non-standard FORTRAN code has been removed allowing the program to be compiled under Linux (at least on some systems), although some users have experienced problems when running the Linux version which are not yet understood.

CCP4 has generously provided support for further development of the MOSFLM software. The current priority is to implement the FFT based auto-indexing procedure developed in Michael Rossmann's group (Steller, I., Bolotovsky, R. and Rossmann, M.G., J. Appl. Cryst. 30, 1036-1040, 1998). This code is in the public domain, and is currently in use at CHESS and should provide a more robust auto-indexing procedure. It is hoped that this option will be available within the next six months, and should greatly improve the general usefulness of the program. Other plans include allowing the use of partials extending over several images in the post-refinement. At present, this is restricted to reflections present on two images, which can cause problems when processing data from crystals with large unit cell parameters (necessitating the use of small oscillation angles) and with a large mosaic spread (equal to or greater than the width of two oscillations). Finally, it would be an advantage to use reflection spot positions, in addition to the post-refinement data, in the refinement of crystal cell and orientation parameters. This is particularly true when processing low resolution data (less than 3.5E resolution) where post-refinement does not lead to very precise values.

The source code and executables for SGI and Dec Alpha machines for the latest version (5.51) are available by anonymous ftp:

   cd pub/mosflm

The README file in this directory contains instructions on installation and a listing of known bugs with workarounds where possible.

Newsletter contents ... UP