Recruiting Outside Talent to Find the World's Smallest Machines

Cryo-electron tomography (cryo-ET) provides a powerful tool for visualizing macromolecular complexes in their native cellular environments, offering unprecedented insights into cellular structure and function. However, automating the identification and classification of these complexes within in vivo tomograms is a significant challenge.We have developed a robust machine learning pipeline that integrates both supervised and unsupervised approaches for accurate detection of macromolecular complexes in complex, real-world cryo-ET data. In addition, we sponsored a $65,000 Kaggle competition, releasing a carefully curated dataset of in vivo cryo-ET tomograms to foster global participation. We will present the initial results, highlighting the most successful deep learning architectures, advanced image processing techniques, and hybrid approaches contributed by a global community. Importantly, the computational challenges of cryo-ET are not limited to experts in biology or microscopy—anyone with a background in machine learning, computer science, or data analysis can make meaningful contributions. The untapped potential within cryo-ET is vast; by shifting the focus of a broader scientific community toward solving these problems, we can unlock biological insights that may revolutionize our understanding of cellular processes and disease mechanisms.