Automated, optimized, intelligent data collection for cryo-EM - PROJECT SUMMARY Cryo-electron microscopy (cryo-EM) has become a transformative technique for determining high-resolution structures of macromolecules, providing insights into complex biological systems that drive our understanding of disease and inform drug design. Despite its success, the level of quality control and automated guidance of high- throughput necessary for the widespread adoption of structural studies requires further improvements. This grant aims to continue the development of Magellon, a next-generation software suite that advances cryo-EM data collection through enhanced automation, real-time analysis, and improved sample preparation workflows. We have established a foundational codebase that will be used to develop advanced web-based tools to enable users to monitor, visualize, and analyze cryo-EM data in real time. Initial development of Magellon will focus primarily on single-particle workflows and will subsequently integrate tomography workflows. Emphasis on the ability to accommodate customizable plugins for various data acquisition functions, such as particle picking, focusing, and CTF estimation, will be at the heart of development. This modular approach allows external developers to add tools directly to Magellon, creating an ecosystem where the community can contribute and test new algorithms without altering core software. These capabilities are especially valuable at regional and national cryo-EM centers, where Magellon’s web-based accessibility and customizable plugins will expand utility across institutions and improve high-quality data acquisition for diverse research needs. However, since the goal of this project is to enable hands-free, fully automated data acquisition, we will leverage our modular framework to integrate new algorithms within Magellon for fully automated data acquisition. To this end, we will largely focus on a standalone platform, CryoEM CoPilot, that will serve as an intelligent data collection assistant. This assistant will use machine learning to dynamically optimize microscope parameters and grid navigation based on ongoing data quality. By leveraging machine learning for real-time decision-making, CryoEM CoPilot will adapt to each sample’s unique features, focusing on high-quality regions and minimizing time spent on suboptimal areas. This approach represents a significant advance over traditional static acquisition strategies, enabling more efficient and precise data collection that reduces the need for expert intervention. Lastly, we will also develop tools for screening and optimizing buffer conditions to identify conditions that yield the most stable samples for cryo-EM. By analyzing protein-protein interactions in various buffer environments, this method will help identify conditions that reduce aggregation and prevent preferred orientation, common issues that can compromise cryo-EM data quality. Together, these aims will establish Magellon as a versatile, modular, and highly extensible platform for cryo-EM, designed to serve as a central resource for the cryo-EM community. By enhancing automation, data feedback, and sample quality, Magellon will make cryo-EM more accessible, reliable, and efficient, enabling scientists to fully exploit the potential of cryo-EM in biomedical research.
