Vision transformer model improves protein structure analysis from cryo-EM images
Learning the Shared Latent Structure of Cryo-EM Inverse Problems
Computer Vision and Pattern RecognitionMachine Learning
Summary
Cryo-EM is a technique that takes very noisy, flat pictures of tiny molecules to understand their 3D shapes and changes. The authors created a special AI model called EMber that learns to recognize shared features in these images, helping solve multiple problems at once, like figuring out how molecules differ and what shapes they take. They trained EMber on tens of millions of real and simulated images and found it works better and faster than previous methods. This helps scientists study how molecules work by revealing details about their structures and variety.
What this means in practice
- •For computational biologists: Improve analysis of cryo-EM data to better distinguish molecular shapes and compositions faster with fewer computational steps.
- •For pharmaceutical data scientists: Enhance characterization of protein conformations to support drug target discovery using more accurate downstream structural assignments.
Authors
Nithin Chalapathi, Theo Sternlieb, Ethan Han, Sonya M. Hanson, Pilar Cossio, Aditi S. Krishnapriyan, Ritwik Gupta
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.