Programmable design of functional RNAs through multimodal generative learning
- 1State Key Laboratory of Membrane Biology, Beijing Advanced Innovation Center for Structural Biology & Frontier Research Center for Biological Structure, School
Abstract
RNA engineering is bottlenecked by the tight coupling among sequence, structure, and function, leaving the field reliant on heuristics or screening. We introduce RNAverse, a multimodal generative model that learns a conditionable design grammar from RNA sequences, experimental structural probing profiles, and functional annotations. Beyond setting a new state of the art across 16 predictive modeling tasks, RNAverse inverts into generative design across four distinct RNA classes through three regimes, requiring experimental testing of no more than 20 candidates per-task. Zero-shot in-painting generates snoRNAs that promote premature termination codon (PTC) readthrough, and chimeric aptazymes for ligand-controlled gene regulation, exceeding rational design or large-scale screens. In both settings, RNAverse outperforms family-adapted sequence-only baselines. Fine-tuned in-painting extends to an unseen functional class, yielding IGFBP3-binding aptamers with picomolar affinity that outperform SELEX winners. Iterative optimization engineers 5' UTRs that outperform clinical benchmarks, enhancing protein production in cell-based assays for both the Zolgensma gene therapy and the BNT162b2 COVID-19 vaccine, and boosting antibody responses to BNT162b2 in mice. Together, these results establish multimodal generative learning as a route from predictive RNA modeling to programmable specification of functional RNAs.
Related
- Metabolite-centric Knowledge Graph Unveils Predictive and Mechanistic Insights through Self-supervised Multimodal Learning
- VarNexus: an integrative database for interpreting disease-associated regulatory variants through multimodal predictions and functional evidence
- ImmuneLens: linking transcriptional states and TCR clonotypes through disentangled multimodal learning
- How thermostable direct haemolysin (TDH) diverges from TDH-related haemolysin (TRH)? Reassessing haemolysins in Vibrio parahaemolyticus through functional and structural representation learning
- SEME enables programmable transcript reconstruction through synthetic microexon incorporation
BibTeX
@misc{ma2026programmable,
title = {{Programmable design of functional RNAs through multimodal generative learning}},
author = {J. Ma and J. Zhu and W. Huang and K. Mu and K. Huang and Z. Dong and F. Guo and X. Xie and Y. Shao and R. Huber and Y. Wan and L. Zhang and Q. Wu and Q. C. Zhang},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.14.759035},
url = {https://www.biorxiv.org/content/10.64898/2026.10.14.759035v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Programmable design of functional RNAs through multimodal generative learning
AU - J. Ma
AU - J. Zhu
AU - W. Huang
AU - K. Mu
AU - K. Huang
AU - Z. Dong
AU - F. Guo
AU - X. Xie
AU - Y. Shao
AU - R. Huber
AU - Y. Wan
AU - L. Zhang
AU - Q. Wu
AU - Q. C. Zhang
PY - 2026
DA - 2026/10/14
PB - bioRxiv
DO - 10.64898/2026.10.14.759035
UR - https://www.biorxiv.org/content/10.64898/2026.10.14.759035v1
AB - RNA engineering is bottlenecked by the tight coupling among sequence, structure, and function, leaving the field reliant on heuristics or screening. We introduce RNAverse, a multimodal generative model that learns a conditionable design grammar from RNA sequences, experimental structural probing profiles, and functional annotations. Beyond setting a new state of the art across 16 predictive modeling tasks, RNAverse inverts into generative design across four distinct RNA classes through three regimes, requiring experimental testing of no more than 20 candidates per-task. Zero-shot in-painting generates snoRNAs that promote premature termination codon (PTC) readthrough, and chimeric aptazymes for ligand-controlled gene regulation, exceeding rational design or large-scale screens. In both settings, RNAverse outperforms family-adapted sequence-only baselines. Fine-tuned in-painting extends to an unseen functional class, yielding IGFBP3-binding aptamers with picomolar affinity that outperform SELEX winners. Iterative optimization engineers 5' UTRs that outperform clinical benchmarks, enhancing protein production in cell-based assays for both the Zolgensma gene therapy and the BNT162b2 COVID-19 vaccine, and boosting antibody responses to BNT162b2 in mice. Together, these results establish multimodal generative learning as a route from predictive RNA modeling to programmable specification of functional RNAs.
ER -