Date Published: September 01, 2017
Publisher: International Union of Crystallography
Author(s): Andrea Thorn, James Parkhurst, Paul Emsley, Robert A. Nicholls, Melanie Vollmar, Gwyndaf Evans, Garib N. Murshudov.
AUSPEX is a new software tool for the statistical analysis of single-crystal X-ray diffraction data. It can be used to identify problems in the data resulting from the experiment itself, image processing, data scaling or conversion.
In a preliminary study, 156 integrated and scaled data sets from the JCSG measured using PILATUS detectors and deposited in the PDB between 2011 and 2015 were evaluated (test set A; see Supporting Information). The observed amplitudes Fobs were plotted against resolution. In these plots, ice rings were visibly identifiable in 15 of the 156 data sets (for an example, see Fig. 3 ▸). This indicates that the background estimation used in processing these data sets was insufficient to correct for the presence of ice diffraction. Two further data sets were found to have significant portions of data removed prior to processing owing to the presence of ice rings (see Fig. 2 ▸). It was clear that these plots held some diagnostic value, and as a result AUSPEX was developed.
In the presence of an ice ring, negative intensities can result from incorrect estimation of the background (see Fig. 3 ▸). When intensities are subsequently converted to amplitudes, they must have values equal to or greater than zero (French & Wilson, 1978 ▸). Consequently, it was found that plots of observed intensities Iobsversus resolution, as shown in Fig. 3 ▸, are more useful than plots of observed amplitudes Fobs against resolution to identify ice rings in X-ray data sets.
In the following subsections, the use of AUSPEX to identify unusual data behaviour is demonstrated and the cause of the observed behaviour is investigated.
Even after more than 20 years of specific research to minimize the influence of ice diffraction in macromolecular crystallography (Mitchell & Garman, 1994 ▸), ice-ring artefacts were present in roughly 20% of 400 data sets (test sets B and C) chosen randomly from the PDB (as found by visual inspection of plots of Iobsversus resolution). A similar percentage (19%) was obtained when 28 895 data sets from the PDB for which intensities had been deposited were evaluated with the automatic ice-ring detection implemented in AUSPEX.