RiboSphere: Learning Unified and Efficient Representations of RNA Structures

2026-03-20

AI Summary

RNA three-dimensional structure modeling faces significant challenges due to high backbone flexibility, the prevalence of non-canonical interactions, and the scarcity of experimentally determined structures. This work proposes a novel approach that integrates vector quantization with flow matching: a geometric Transformer encoder extracts SE(3)-invariant features, which are then discretized via finite scalar quantization (FSQ) into a codebook enriched with RNA structural motif information. A flow-matching decoder subsequently reconstructs atomic coordinates from these discrete representations. By uniquely combining discrete geometric representations with the modular nature of RNA, the method achieves state-of-the-art performance in structure reconstruction, yielding an RMSD of 1.25 Å and a TM-score of 0.84. Furthermore, it demonstrates strong transferability and robust generalization under data-scarce conditions in downstream tasks such as inverse folding and RNA–ligand binding prediction.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: 3D Computer Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Accurate RNA structure modeling remains difficult because RNA backbones are highly flexible, non-canonical interactions are prevalent, and experimentally determined 3D structures are comparatively scarce. We introduce \emph{RiboSphere}, a framework that learns \emph{discrete} geometric representations of RNA by combining vector quantization with flow matching. Our design is motivated by the modular organization of RNA architecture: complex folds are composed from recurring structural motifs. RiboSphere uses a geometric transformer encoder to produce SE(3)-invariant (rotation/translation-invariant) features, which are discretized with finite scalar quantization (FSQ) into a finite vocabulary of latent codes. Conditioned on these discrete codes, a flow-matching decoder reconstructs atomic coordinates, enabling high-fidelity structure generation. We find that the learned code indices are enriched for specific RNA motifs, suggesting that the model captures motif-level compositional structure rather than acting as a purely compressive bottleneck. Across benchmarks, RiboSphere achieves strong performance in structure reconstruction (RMSD 1.25\,Å, TM-score 0.84), and its pretrained discrete representations transfer effectively to inverse folding and RNA--ligand binding prediction, with robust generalization in data-scarce regimes.

Problem

• RNA structure modeling
• non-canonical interactions
• 3D structure scarcity
• backbone flexibility

Innovation

• discrete representation
• geometric deep learning
• RNA structure modeling
• vector quantization
• flow matching