Three-dimensional electron diffraction (3D ED), also known as microcrystal electron diffraction (microED), is an emerging method for determining structures from submicron-sized crystals. With the development of rapid and convenient data collection protocols, acquiring dozens of datasets in a single 3D ED/microED session has become routine. A fast and automated workflow for processing, scaling and merging a large number of 3D ED/microED datasets can significantly accelerate the structure determination process. Herein, we present an XDS-based pipeline with a graphical user interface for automated real-time and offline batch 3D ED/microED data processing (AutoLEI). We demonstrate the functionality and applications of the pipeline through four examples, using both offline and real-time data processing capabilities. The samples include small organic molecules, metal–organic frameworks (MOFs) and proteins, showcasing the versatility and efficiency of AutoLEI in various applications.

3D electron diffraction (3D ED) has undergone impressive development in the last decade. However, its accuracy and reproducibility have never been tested, up to now, in different laboratories on the same batch of samples. This paper reports a round robin on three test structures, two inorganic and one organic, solved and refined with 3D ED in seven different laboratories employing different transmission electron microscopes, with different acceleration voltages, different methodologies and different detectors. The results of the round robin show a remarkable accuracy of the technique that, in the case of kinematical refinement, is around 0.05 Å error on atomic positions for the inorganic samples and 0.15 Å for the beam-sensitive organic crystal. Dynamical refinement further improves the accuracy. The analysis of diverse samples and numerous data sets again confirms that dynamical refinement is a well established procedure, significantly reducing the refinement R factors, improving the accuracy of the structure models in most cases, and providing fine structural details, such as hydrogen-atom positions and the absolute structure, for both inorganic and organic samples.

Prussian blue analogues (PBA) are a large family of functional materials with diverse applications such as in electrochemical fields. However, their use in the emerging two-electron oxygen reduction reaction for clean production of hydrogen peroxide (H₂O₂) is lagging. Herein, a general solvent exchange induced reconstruction strategy is demonstrated, through which an abnormal NiNi-PBA superstructure is synthesized as a high-performance electrocatalyst for H₂O₂ generation. The resultant NiNi-PBA superstructure has a stoichiometric composition with saturated lattice water, and a leaf-like morphology composed of interconnected small-size nanosheets with identical orientation and predominate {210} side surface exposure. Our studies show that the Ni−N centers on {210} facets are the active sites, and the saturated lattice H₂O favors a six-coordinated environment that results in high selectivity. The “perfect” structure including stoichiometric composition and ideal facet exposure leads to a high selectivity of ~100 % and H₂O₂ yield of 5.7 mol g⁻¹ h⁻¹, superior to the reported MOF-based electrocatalysts and most other electrocatalysts.

Macromolecular crystallography provides high-resolution structural information of proteins and protein-ligand complexes, offering essential insights for structure-based drug discovery. Electron diffraction methods, including three-dimensional electron diffraction (3D ED), also known as micro-crystal electron diffraction (MicroED), and serial electron diffraction (SerialED), enable structure determination from nanocrystals and are particularly useful for macromolecules that cannot form large crystals or require long crystallization times. However, the application of electron diffraction to macromolecular structure determination remains challenging, owing to limitations in sample preparation, beam-induced damage, and the efficiency of data processing. To overcome these difficulties, multiple methods and workflows were developed and systematically optimized in this thesis. 

For sample preparation, a rapid mixing protein crystallization (RaMiC) method combined with a surfactant-assisted grid preparation strategy was established. Through this method, cryogenic electron microscopy (cryo-EM) grids containing well-ordered sub-micron-sized crystals suitable for electron diffraction can be reproducibly prepared. To simplify and streamline MicroED data processing, a graphical platform named Automated Real-time and Offline Batch MicroED Data Processing Graphical User Interface (AutoLEI) was developed. It enables high-throughput offline batch processing of large numbers of datasets within minutes and supports online real-time processing. To further reduce electron beam damage, a continuous SerialED (c-SerialED) method was developed, and as a benchmark, the triclinic lysozyme structure was determined at 0.83 Å resolution. From the generated difference map, potential positions for hydrogen atoms were directly identified. 

Based on these developments, a structure-based drug discovery workflow integrating RaMiC, rapid soaking, real-time MicroED ligand screening and high-resolution c-SerialED structure determination was established. To validate the workflow, crystals of the human protein MutT homolog 1 (MTH1) were soaked individually with four ligands, and the method was applied to determine whether they bind to the protein’s active site. With soaking times as short as 3 seconds, binding could be preliminarily evaluated by real-time MicroED within 25-50 minutes. For three ligands that passed pre-screening, c-SerialED data were collected and structures of protein-ligand complexes were solved at 1.7 Å resolution. The electrostatic potential maps clearly revealed the binding modes. 

In summary, this thesis establishes a practical workflow for high-resolution protein and protein-ligand complex structure determination by electron diffraction, highlighting its potential in structure-based drug discovery.

Understanding protein-ligand interactions is a key step for small-molecule drug discovery.  Conventional single-crystal X-ray diffraction provides accurate and high-throughput structural analysis. However, it is limited by the requirement for large crystals. Electron diffraction methods, such as microcrystal electron diffraction (MicroED/3DED) and serial electron diffraction (SerialED), offer ideal alternatives for structure determination of small protein crystals. Here, we present a workflow combining real-time MicroED and continuous SerialED (c-SerialED) to enable fast and accurate ligand binding analysis from microcrystals. By using the real-world cancer target human MutT homolog 1 (MTH1), we performed ligand soaking with four compounds of varying binding affinities. Each soaking event was completed within 3 seconds. Real-time MicroED analysis enabled confirmation of ligand binding within 25–50 minutes. High-resolution (1.7 Å) structures of the protein–ligand complexes were obtained through subsequent c-SerialED data collection. This workflow demonstrates the potential of using electron crystallography for structure-based drug discovery.

X-ray protein crystallography has predominantly centred on growing large (> 5 µm) single crystals for structural analysis. While fragments of these large-grown, as well as small-grown crystals have been utilized in electron diffraction (3D ED/MicroED), a universal method for protein crystallisation tailored to sub-micrometre thickness is missing. In this study, we present a method wherein rapid mixing and fragmentation of initial crystals yield well-diffracting nanocrystals in seconds or minutes. In suitable space groups, these nanocrystals can form thin plates, providing ideal samples for (3D ED/MicroED) without requiring focused ion beam milling. The large number of suitably sized crystals also allows for the preparation of densely populated grids suitable for serial ED. We further introduce a streamlined manual grid preparation protocol using polysorbate 20 as a wetting agent, eliminating the need for glow discharging or plunge freezers. By applying these methods, micro- to nano-sized crystals of four different protein samples were generated and the resolution of each sample reached beyond 1.5 Å. Our high-resolution diffraction data demonstrate that slow and undisturbed growth is not necessary for obtaining well-diffracting protein crystals. This study offers a general and feasible sample preparation approach that enables high-resolution protein structure determination by electron diffraction.