AutoPD: an integrated meta-pipeline for high-throughput X-ray crystallography data processing and structure determination

Abstract

The advent of hybrid pixel array detectors and fully automated data acquisition workflows has revolutionized synchrotron light sources, enabling high-throughput collection of diffraction data from biological macromolecular crystals. However, these advancements have also created an urgent need for efficient and fully automated data processing pipelines. To address this challenge, we introduce AutoPD, an open-source high-throughput meta-pipeline for automated data processing and structure determination. Developed for the biological macromolecular crystallography beamline at the High Energy Photon Source in Beijing, AutoPD is also accessible to other academic and synchrotron users. By integrating cutting-edge parallel computing strategies, AlphaFold-assisted molecular replacement, a direct-method-based dual-space-iteration approach for model building, and an adaptive decision-making strategy that dynamically selects the optimal modeling pathway based on data quality and intermediate results, AutoPD streamlines the process from raw diffraction data and sequence files to high-precision structural models. When benchmarked against 186 recently deposited X-ray diffraction datasets from the Protein Data Bank, AutoPD successfully determined structures for 92% of cases, achieving map–model correlation values of at least 0.5 between density-modified electron density maps and the generated models. These results highlight the robustness and efficiency of AutoPD in addressing the challenges of modern structural biology, setting a new standard for automated structure determination.