MS02-P12 New tools for serial crystallography: SPIND - sparse pattern auto-indexing and DatView - exploring and optimizing large multi-crystal datasetsWe present two new tools for making efficient use of serial femtosecond crystallography (SFX) data. A large fraction of SFX patterns from microcrystals of large macromolecules contain fewer than 15 Bragg spots and are often discarded by existing autoindexing algorithms. SPIND is a reference-based algorithm for auto-indexing sparse, snapshot serial crystallography data with as few as 5 Bragg spots, using a known unit cell . We demonstrate the suitability of SPIND for indexing sparse inorganic crystal data and for improving the quality of SFX data from two G protein-coupled receptors.
We have also developed DatView, a tool for exploring correlations in, and optimizing merged reflection lists from, large serial crystallography datasets. Stochastic fluctuations in microcrystals and X-ray free electron laser pulse parameters lead to orders of magnitude variability in measured intensities from equivalent reflections, which necessitated large SFX datasets for accurate merged structure factors. Numerous improvements in sample delivery, detectors and, critically, data analysis (including geometry optimization, orientation refinement, scaling and post-refinement), have significantly decreased the required number of indexed SFX patterns for structure determination. However, large SFX datasets (collected from thousands of microcrystals at room temperature) now allow users to not only select an optimum subset from potentially anisomorphous microcrystals for more accurate structure factors, but, importantly, it enables the quantification of microcrystal variability and sensitivity to experimental environment such as temperature and humidity, and offer insights into protein flexibility.
SPIND and DatView are written in Python and available via zatsepinlab.atlassian.net
 C. Li, X. Li, R. Kirian, J. Spence, H. Liu, and N. Zatsepin. 2018. IUCrJ. In press. Keywords: serial femtosecond crystallography, dataset optimization, clustering