Date Published: April 01, 2018
Publisher: International Union of Crystallography
Author(s): Igor Melnikov, Olof Svensson, Gleb Bourenkov, Gordon Leonard, Alexander Popov.
A method and software program, MeshBest, for the detection of individual crystals based on two-dimensional X-ray mesh scans are presented.
In X-ray crystallography, samples vary in size, in shape and in diffraction strength. The experiments carried out for data collection in macromolecular crystallography (MX) can be optimized (Bourenkov & Popov, 2006 ▸) via pre-interrogation of the sample(s). Such pre-interrogation should provide essential information concerning the shape, size, position and diffraction strength of the crystal.
The MeshBest workflow is presented in Fig. 1 ▸ and comprises three major steps. Firstly, any mesh-scan images that contain multi-crystal diffraction patterns are detected. This is performed by analysing diffraction vector statistics for each diffraction image (§2.1). Secondly, diffraction images containing diffraction from only one crystal are analysed and those belonging to the same crystal are grouped (§2.2). Finally, the sizes and dispositions of all single crystals contained in the sample loop are described using an elliptical shape approximation (§2.3). Note that rotation of the sample during a mesh scan is not preferred; acquiring still images would simplify the analysis. However, the mesh scans presented here were acquired with slight rotation, which was caused by the need to trigger detector readout on some beamlines.
The program MeshBest was implemented as a Python module using standard (NumPy, SciPy) libraries. In order to demonstrate the applicability of MeshBest and its performance, we present four different experiments here. These include MeshBest analyses of a large, homogeneous crystal (§3.1) and of a disordered crystal with satellites (§3.2), the analysis of a mesh scan performed prior to subsequent multi-crystal data collection on a sample holder containing membrane-protein crystals buried in opaque lipidic mesophase (§3.3) and a mesh scan from a sample holder containing a crystal mess with ‘dirty’ diffraction patterns (§3.4).
The results presented above clearly show that MeshBest analysis of X-ray mesh scans provides, in an automated way, useful information concerning the positions, sizes and relative diffraction strengths of crystals of macromolecules mounted in the sample holder. Such information is critical for the proper organization and design of subsequent diffraction data-collection protocols. Where MeshBest indicates that the sample holder contains one (or relatively few) crystal(s), the assessment of crystal size will allow a more precise description of radiation-damage effects arising during measurements (Zeldin et al., 2013 ▸), especially when the size of the sample is smaller than X-ray beam, and thus help to define a more realistic data-collection strategy (Bourenkov & Popov, 2010 ▸). In cases where MeshBest indicates that large crystals are essentially homogeneous (i.e. §§3.1 and 3.2), then such an approach combined with ‘helical’-style data collections (Flot et al., 2010 ▸) might be the preferred mechanism of optimizing data quality. Here, though, the analysis presented in §3.2 shows that care should be exercised to define a protocol in which the direction of the helical scan and or the beam size used avoids illuminating satellite crystals and the production of multi-pattern diffraction images.