Learning Adaptive Topology with FiLM-Guided Distillation for Tertiary Structure-Based RNA Design

Tertiary structure-based RNA design aims to generate RNA sequences that can fold into desired 3D structures, but remains a challenging problem due to the scarcity of annotated data, structural noise, and the intrinsic complexity of RNA topology. Existing structure-to-sequence frameworks largely rely on static k-nearest neighbor graphs and rigid message passing schemes, which fail to capture the flexible and heterogeneous nature of RNA geometry. To address these issues, we propose a unified framework, ATL-FGD, that integrates Adaptive Topology Learning (ATL) and FiLM-Guided Distillation (FGD) for robust RNA design. ATL introduces a differentiable edge gating mechanism to jointly learn topology and representation, enabling the model to construct data-driven, layer-adaptive graphs that better reflect structural dynamics and biochemical consistency. On top of this, FGD bridges structural and sequence representations via feature-wise linear modulation, softly transferring the semantic knowledge from RNA foundation models without relying on them during inference. Extensive experiments on tertiary structure-based RNA design benchmarks demonstrate that our approach achieves significant improvements in both sequence recovery and structural fidelity.

RNA molecules play important roles in biology, and their function is closely related to the three-dimensional shapes they fold into. A key challenge is to design RNA sequences that can fold into a desired 3D structure, which could support future work in RNA therapeutics, synthetic biology, and molecular engineering. However, this is difficult because experimentally measured RNA 3D structures are limited, and RNA molecules often have flexible and complex interactions. We introduce ATL-FGD, a method that learns which structural connections should be kept for each layer, allowing the model to filter noisy contacts while preserving useful local and long-range relationships. Most existing AI methods represent an RNA structure using a fixed kNN graph of nearby nucleotides, but this graph may include misleading connections or miss important long-range interactions. ATL-FGD addresses this by starting from a broad set of possible structural connections and then learning, layer by layer, which connections should actually be used for information processing. To further compensate for the limited data, ATL-FGD also learns from a large RNA sequence foundation model during training, but transfers this knowledge through soft feature adjustments rather than directly alignment. Across several RNA design benchmarks, ATL-FGD generates more accurate RNA sequences than previous methods while keeping inference efficient. This work provides a more adaptive way to model RNA structures and may help support the computational RNA design.