Research Article: TakeTwo: an indexing algorithm suited to still images with known crystal parameters

Date Published: August 01, 2016

Publisher: International Union of Crystallography

Author(s): Helen Mary Ginn, Philip Roedig, Anling Kuo, Gwyndaf Evans, Nicholas K. Sauter, Oliver Ernst, Alke Meents, Henrike Mueller-Werkmeister, R. J. Dwayne Miller, David Ian Stuart.


A novel indexing method is presented that is well suited to the minimal information on a still image diffraction pattern and can achieve indexing rates of over one lattice per image.

Partial Text

Indexing, or deducing the specimen orientation from crystalline diffraction patterns, can potentially be performed with high accuracy and precision owing to the integral nature of the Miller indices at which Bragg reflections are located. Indexing algorithms implemented in programs such as XDS (Kabsch, 1993 ▸), iMosflm (Powell et al., 2013 ▸), DENZO (Otwinowski & Minor, 2006 ▸), LABELIT (Sauter et al., 2004 ▸) and DIALS (Gildea et al., 2014 ▸) are well established for data-collection strategies that involve crystal rotation, which are typically employed at synchrotron sources. For data collected at X-ray free-electron laser (XFEL) sources, where each image represents diffraction from a separate nonrotating specimen, the measure of success is less well defined. Data-analysis pipelines sometimes include a preprocessing ‘hitfinder’ step, which distinguishes images without diffraction (blanks) from hits that exceed a certain threshold number of candidate Bragg spots, which is typically set to around 20. The indexing rate is therefore defined as the percentage of hits for which crystal orientations can be determined. Reported indexing rates are often quite low (Barends et al., 2013 ▸; Liu et al., 2013 ▸; Johansson et al., 2013 ▸; Ginn, Brewster et al., 2015 ▸; Chapman et al., 2011 ▸), although the rate depends on a number of factors, including the strength of diffraction on each image and the diffraction resolution. Our own two analyses of diffraction data from Cypovirus type 17 polyhedrin (CPV17) yielded mediocre indexing rates of 53% (Ginn, Messerschidt et al., 2015 ▸) and 36% (Ginn, Brewster et al., 2015 ▸). Here, we investigate whether alternate algorithms can help improve the indexing rate, with the overall goal of producing good-quality structures while consuming a minimal amount of crystalline sample and beam time.

For the reliable indexing of diffraction patterns, experimental parameters such as the direct beam position and crystal-to-detector position need to be known reasonably accurately. Powder patterns aggregated from a large number of diffraction patterns are useful in establishing these parameters. We have found that in cases where only a small number of images are available, or the unit cell is large enough to obscure the low-resolution rings or prevent adequate separation of rings, a pseudo-powder pattern can still be generated by considering the projected three-dimensional vectors between spots on the Ewald sphere. We find that such pseudo-powder patterns are generally useful tools for detecting and correcting significant errors in experimental parameters such as the crystal-to-detector distance, which can still be poorly estimated at XFEL beamlines.




Leave a Reply

Your email address will not be published.