RNA Design via Conditioned Flow Matching and Finite-Policy Reinforcement Learning
Abstract
RNA design aims to identify sequences that fold into specified secondary structures. Existing methods formulate the task as target-specific search or conditional generation. However, natural RNA evolution proceeds through sequence variation and selection, with compensatory substitutions, whereas these methods do not explicitly model this process. To address this limitation, we propose a two-stage framework comprising RNA Inverse-Folding Flow (RNA-IFlow) and RNA-IFlow-RL. RNA-IFlow uses structure-conditioned Dirichlet Flow Matching to model coordinated variation across the sequence, while RNA-IFlow-RL maps the learned flow to a pairing-preserving finite policy and refines it with thermodynamic feedback. Our framework achieves leading performance on multiple benchmarks, reaching 85.19% Pass@1 on Rfam-27. Further analyses reveal thermodynamic gains, policy dynamics, and robustness across settings. Our work couples coordinated variation with thermodynamic selection, offering a novel paradigm for RNA design.
Related
- MORAD: Reinforcement Learning with Multi-Objective Rewards for RNA Inverse Design
- RiboBridge: protein-conditioned de novo RNA binder structure design
- Environment-Conditioned Design of α-Helical Peptides.
- PeTFlow: Flow-Matching Based All-Atom Peptide Design through Side-Chain State Tokenization
- La-Ribo: RNA Co-Design via Geometry-Latent Flow Matching
BibTeX
@misc{lin2026design,
title = {{RNA Design via Conditioned Flow Matching and Finite-Policy Reinforcement Learning}},
author = {Zefeng Lin and Xianyong Fang and Tianfan Fu and Xiaohua Xu},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.36885},
eprint = {2609.36885},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.36885},
}
RIS
TY - UNPB
TI - RNA Design via Conditioned Flow Matching and Finite-Policy Reinforcement Learning
AU - Zefeng Lin
AU - Xianyong Fang
AU - Tianfan Fu
AU - Xiaohua Xu
PY - 2026
DA - 2026/09/29
PB - arXiv
DO - 10.48550/arxiv.2609.36885
UR - https://arxiv.org/abs/2609.36885
AB - RNA design aims to identify sequences that fold into specified secondary structures. Existing methods formulate the task as target-specific search or conditional generation. However, natural RNA evolution proceeds through sequence variation and selection, with compensatory substitutions, whereas these methods do not explicitly model this process. To address this limitation, we propose a two-stage framework comprising RNA Inverse-Folding Flow (RNA-IFlow) and RNA-IFlow-RL. RNA-IFlow uses structure-conditioned Dirichlet Flow Matching to model coordinated variation across the sequence, while RNA-IFlow-RL maps the learned flow to a pairing-preserving finite policy and refines it with thermodynamic feedback. Our framework achieves leading performance on multiple benchmarks, reaching 85.19% Pass@1 on Rfam-27. Further analyses reveal thermodynamic gains, policy dynamics, and robustness across settings. Our work couples coordinated variation with thermodynamic selection, offering a novel paradigm for RNA design.
ER -