Learning the Shared Latent Structure of Cryo-EM Inverse Problems
Abstract
Understanding biomolecular function requires characterizing three-dimensional structure and conformational variability, for which cryogenic electron microscopy (cryo-EM) is a powerful tool. Cryo-EM produces noisy two-dimensional particle images in which molecular structure, particle pose, and imaging conditions are jointly encoded. Recovering biological information from these observations spans inverse problems including pose estimation, heterogeneity analysis, and reconstruction. These tasks are usually addressed independently despite shared protein structure and image formation, motivating a common representation across inverse problems. Since biologically meaningful differences may be spatially localized, we study self-distillation with complementary global and masked patch objectives. Because experimental dataset discrimination can reflect acquisition-specific cues, we develop controlled synthetic and experimental evaluations and find that structural discrimination benefits from dense spatial features. We introduce the EMber family of vision transformers, pre-trained on >50 million experimental particle images from 214 EMPIAR datasets together with simulated projections from 268,000 molecular assemblies. EMber improves upon existing methods at conformational and compositional discrimination by 14.85 and 14.26 percentage points, estimates conformational populations to within 1.59pp error with a linear probe, and produces particle assignments yielding reference-quality reconstructions in fewer epochs. We demonstrate how to learn representations that capture biological information for use across multiple cryo-EM inverse problems.
Related
- LNODE: latent dynamics reveal the shared spatiotemporal structure of amyloid-β progression
- Learning Interpretable Switching Dynamics in Shared Neural-Behavioral Latent Space
- Shared latent representations of speech production for cross-patient speech decoding
- From Arm to Leg: A Shared Primitive Framework for Solving Inverse Kinematics in Human Movement
- Shared Latent Decision Strategies Underlie Reward-Guided Behavior Across Species
BibTeX
@misc{chalapathi2026learning,
title = {{Learning the Shared Latent Structure of Cryo-EM Inverse Problems}},
author = {Nithin Chalapathi and Theo Sternlieb and Ethan Han and Sonya M. Hanson and Pilar Cossio and Aditi S. Krishnapriyan and Ritwik Gupta},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.22523},
eprint = {2610.22523},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.22523},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Learning the Shared Latent Structure of Cryo-EM Inverse Problems
AU - Nithin Chalapathi
AU - Theo Sternlieb
AU - Ethan Han
AU - Sonya M. Hanson
AU - Pilar Cossio
AU - Aditi S. Krishnapriyan
AU - Ritwik Gupta
PY - 2026
DA - 2026/10/15
PB - arXiv
DO - 10.48550/arxiv.2610.22523
UR - https://arxiv.org/abs/2610.22523
AB - Understanding biomolecular function requires characterizing three-dimensional structure and conformational variability, for which cryogenic electron microscopy (cryo-EM) is a powerful tool. Cryo-EM produces noisy two-dimensional particle images in which molecular structure, particle pose, and imaging conditions are jointly encoded. Recovering biological information from these observations spans inverse problems including pose estimation, heterogeneity analysis, and reconstruction. These tasks are usually addressed independently despite shared protein structure and image formation, motivating a common representation across inverse problems. Since biologically meaningful differences may be spatially localized, we study self-distillation with complementary global and masked patch objectives. Because experimental dataset discrimination can reflect acquisition-specific cues, we develop controlled synthetic and experimental evaluations and find that structural discrimination benefits from dense spatial features. We introduce the EMber family of vision transformers, pre-trained on >50 million experimental particle images from 214 EMPIAR datasets together with simulated projections from 268,000 molecular assemblies. EMber improves upon existing methods at conformational and compositional discrimination by 14.85 and 14.26 percentage points, estimates conformational populations to within 1.59pp error with a linear probe, and produces particle assignments yielding reference-quality reconstructions in fewer epochs. We demonstrate how to learn representations that capture biological information for use across multiple cryo-EM inverse problems.
ER -