RIDER: 3D RNA Inverse Design with Reinforcement Learning-Guided Diffusion
This addresses the need for more accurate RNA design in synthetic biology and therapeutics by moving beyond limited sequence-based metrics to directly optimize structural fidelity.
The paper tackles the problem of inverse design of RNA 3D structures by proposing RIDER, a framework that directly optimizes for 3D structural similarity rather than relying on native sequence recovery. The method achieves a 9% improvement in native sequence recovery and over 100% improvement in structural similarity metrics compared to state-of-the-art approaches.
The inverse design of RNA three-dimensional (3D) structures is crucial for engineering functional RNAs in synthetic biology and therapeutics. While recent deep learning approaches have advanced this field, they are typically optimized and evaluated using native sequence recovery, which is a limited surrogate for structural fidelity, since different sequences can fold into similar 3D structures and high recovery does not necessarily indicate correct folding. To address this limitation, we propose RIDER, an RNA Inverse DEsign framework with Reinforcement learning that directly optimizes for 3D structural similarity. First, we develop and pre-train a GNN-based generative diffusion model conditioned on the target 3D structure, achieving a 9% improvement in native sequence recovery over state-of-the-art methods. Then, we fine-tune the model with an improved policy gradient algorithm using four task-specific reward functions based on 3D self-consistency metrics. Experimental results show that RIDER improves structural similarity by over 100% across all metrics and discovers designs that are distinct from native sequences.