Multi-task Learning for Macromolecule Classification, Segmentation and Coarse Structural Recovery in Cryo-Tomography
This work addresses the difficulty of macromolecule analysis in cryo-tomography for structural biology, offering a multi-task approach that improves performance and generalizability, though it is incremental as it builds on existing single-task methods.
The paper tackled the challenge of recognizing and recovering macromolecular structures in cellular electron cryo-tomography (CECT) by proposing a multi-task 3D convolutional neural network for simultaneous classification, segmentation, and coarse structural recovery, which outperformed single-task methods on simulated and experimental data and demonstrated generalization to novel structures.
Cellular Electron Cryo-Tomography (CECT) is a powerful 3D imaging tool for studying the native structure and organization of macromolecules inside single cells. For systematic recognition and recovery of macromolecular structures captured by CECT, methods for several important tasks such as subtomogram classification and semantic segmentation have been developed. However, the recognition and recovery of macromolecular structures are still very difficult due to high molecular structural diversity, crowding molecular environment, and the imaging limitations of CECT. In this paper, we propose a novel multi-task 3D convolutional neural network model for simultaneous classification, segmentation, and coarse structural recovery of macromolecules of interest in subtomograms. In our model, the learned image features of one task are shared and thereby mutually reinforce the learning of other tasks. Evaluated on realistically simulated and experimental CECT data, our multi-task learning model outperformed all single-task learning methods for classification and segmentation. In addition, we demonstrate that our model can generalize to discover, segment and recover novel structures that do not exist in the training data.