DISCA: high-throughput cryo-ET structural pattern mining by deep unsupervised clustering

Cryo-electron tomography directly visualizes heterogeneous macromolecular structures in complex cellular environments, but existing computer-assisted sorting approaches are low-throughput or inherently limited due to their dependency on available templates and manual labels. We introduce a high-throughput template-and-label-free deep learning approach that automatically discovers subsets of homogeneous structures by learning and modeling 3D structural features and their distributions. Diverse structures emerging from sorted subsets enable systematic unbiased recognition of macromolecular complexes in situ.