Making sense of SFX data: standards for data and structure validation for a non-standard experiment that has come of age

Abstract

Innovation and revolution are paramount to the advancement of science and have shaped the ways in which we do research today. The field of structural biology has had multiple drivers of change, the two most recent being the use of X-ray free-electron laser (XFEL) sources for crystallographic data collection (Martin-Garcia et al., 2016), and the 'resolution revolution' in cryo-EM, stemming from advances in detectors and image processing (Kü hlbrandt, 2014). In protein crystallography, several advancements have moved this technique forward since the first structures of myoglobin and hemoglobin. These advances include (i) the use of synchrotron radiation (Dauter & Wlodawer, 2016; Phillips et al., 1976); (ii) the application of multiwavelength data collection to the solution of the phase problem (Guss et al., 1988; Hendrickson & Teeter, 1981; Hendrickson et al., 1990); (iii) the advent of cryocrystallography (Hope, 1988; Pflugrath, 2015); (iv) automation (Cohen et al., 2002; Snell et al., 2004) and remote access (Smith et al., 2010; Soltis et al., 2008); and (v) the application of hybrid photon-counting (HPC) detectors (Fö rster et al., 2019; Brö nnimann & Trü b, 2018). Additional advances in data analysis and validation , including the use of the free R factor during refinement (Brü nger, 1992), the introduction of R meas and R p.i.m. statistics during data processing (Weiss, 2001) (which have effectively replaced R merge in 'Table 1'), and the use of CC 1/2 and CC* (Karplus & Diederichs, 2012), have all contributed to the robustness of the modern protein crystallography experiment. The most recent advance, the application of high-brilliance, time-structured XFEL beams to problems in structural biology, has disrupted the way in which the protein crystallography experiment is carried out at these fourth-generation light sources. The intense microfocus beams opened new experimental possibilities with micro-and nanocrystals, hitherto deemed too small for conventional data collection at synchrotrons, and the unique time-structure of the beams in the femtosecond regime, sparked a resurgence in the use of time-resolved (TR) crystallography to study the reaction mechanisms of enzymes in action (Tenboer et al., 2014; Schmidt, 2017; Barends et al., 2015). It also sparked the concomitant development of novel sample delivery methods including injectors, fixed target and hybrid methods (Martiel et al., 2019). Moreover, the use of XFELs for protein crystallography has given rise to a new data collection paradigm , serial femtosecond crystallography (SFX), whereby a series of still images are collected from randomly oriented crystals intersecting the XFEL beam at a rate determined by the repetition rate of the beam and/or the readout rate of the detector. Because the methods for data collection and data processing in conventional synchrotron crystallography were so robust, having been continuously developed over the preceding 50 or more years, it seemed obvious to attempt to apply these 'standard' methods to the data sets collected at XFELs. In this issue of IUCrJ, Gorel and colleagues (Gorel et al., 2021) suggest that in SFX experiments the distinct features of the XFEL beams and the various ways in which samples are delivered into the beam give rise to issues unique to these types of experiments, particularly with respect to the determination of the quality of the data, the validity of the derived structure, and the extrapolated biological results. In order to fully validate the results from these experiments, the scientific community needs to be able to visualize and analyze the experimental data rather than relying on a 'Table 1' type of approach which, although completely adequate for conventional synchrotron-based diffraction experiments, falls short in the case of XFEL experiments.