Supporting data for “Utilizing AlphaFold Predictions for Enhancing X-ray Crystallographic Structure Determination”

High-resolution X-ray diffraction remains central to macromolecular structure determination; however, traditional phase determination methods, such as molecular replacement with homologous structures or experimental techniques like single-wavelength anomalous dispersion, possess inherent limitations. The emergence of accurate de novo structural models generated by AlphaFold, even for proteins without existing homologs, promises significant advancement in crystallographic phasing by providing reliable initial models. Concurrently, technological innovations, including rapid pixel array detectors and automated beamline systems, have enabled high-throughput crystallography, necessitating the development of similarly automated structure determination pipelines.

AutoPD, an open-source meta-pipeline leveraging AlphaFold predictions for automated crystallographic analysis, is presented. This pipeline streamlines the entire workflow, progressing from raw diffraction images and protein sequences to refined structural models. AlphaFold-assisted molecular replacement is integrated to establish initial phasing, followed by a dual-space iterative refinement method combining real-space and reciprocal-space approaches. Adaptive decision-making dynamically selects optimal processing pathways based on data quality and interim results. Collectively, these innovations allow structures to be solved automatically, significantly reducing manual intervention and addressing critical bottlenecks in conventional methods.