RobPicker: a meta-learning framework for robust identification of macromolecules in cryo-electron tomograms

Accurate particle picking of macromolecules of diverse types, shapes and sizes in cryo-electron tomography (cryo-ET) is critical for understanding the molecular architecture of biological systems in their native state. Current deep learning methods have shown potential for identifying macromolecules in tomograms; however, they are vulnerable to annotation noise in the training data and to imbalanced distributions of macromolecule species. Here, to address these limitations, we developed RobPicker, a meta-learning framework that effectively mitigates these issues by automatically learning to correct label errors and by placing greater emphasis on smaller and rarer macromolecule species. In evaluations across diverse cryo-ET datasets with noisy labels and imbalanced species distributions, RobPicker substantially outperforms state-of-the-art methods, particularly in identifying small and rare macromolecules. The robustness and efficiency of RobPicker led to improved resolution in 3D density maps of protein structures, enabling high-resolution cellular structural biology analysis.