RiboSphere: Learning Unified and Efficient Representations of RNA Structures

Zhou Zhang, Hanqun Cao, Cheng Tan, Fang Wu, Pheng-Ann Heng, Tianfan Fu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:154263-154278, 2026.

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 RiboSphere, a framework that learns 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 trained using mean-centered coordinates and random rotation augmentation to produce geometry-aware 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.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-zhang26h, title = {{R}ibo{S}phere: Learning Unified and Efficient Representations of {RNA} Structures}, author = {Zhang, Zhou and Cao, Hanqun and Tan, Cheng and Wu, Fang and Heng, Pheng-Ann and Fu, Tianfan}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {154263--154278}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/zhang26h/zhang26h.pdf}, url = {https://proceedings.mlr.press/v306/zhang26h.html}, 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 RiboSphere, a framework that learns 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 trained using mean-centered coordinates and random rotation augmentation to produce geometry-aware 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.} }
Endnote
%0 Conference Paper %T RiboSphere: Learning Unified and Efficient Representations of RNA Structures %A Zhou Zhang %A Hanqun Cao %A Cheng Tan %A Fang Wu %A Pheng-Ann Heng %A Tianfan Fu %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-zhang26h %I PMLR %P 154263--154278 %U https://proceedings.mlr.press/v306/zhang26h.html %V 306 %X 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 RiboSphere, a framework that learns 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 trained using mean-centered coordinates and random rotation augmentation to produce geometry-aware 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.
APA
Zhang, Z., Cao, H., Tan, C., Wu, F., Heng, P. & Fu, T.. (2026). RiboSphere: Learning Unified and Efficient Representations of RNA Structures. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:154263-154278 Available from https://proceedings.mlr.press/v306/zhang26h.html.

Related Material