Protein complexes (such as oxygen-carrying hemoglobin, or keratin in hair, and thousands of others) are essential for cell function, and understanding their interactions is essential for our health and finding new disease treatments. Cryo-electron tomography (cryoET) creates 3D images—called tomograms—at near-atomic detail, showing proteins in their very complex and crowded natural environment. Therefore, cryoET has immense potential to unlock the mysteries of the cell.


There is a wealth of cryoET tomograms that is yet to be fully mined. A large and growing portion of this published corpus exists in a standardized format in the cryoET data portal (cryoetdataportal.czscience.com). Mining this data requires automatic identification of each protein molecule within these images. This problem has not been solved even for proteins that are identifiable by the human eye. A generalizable solution will reveal the “dark matter” of the cell, and will enable thousands of discoveries contributing to human health.

This competition challenges you to create ML algorithms that automatically annotate five classes of protein complexes within a curated real-world cryoET dataset.