BAnG: Bidirectional Anchored Generation for Conditional RNA Design

2025-02-28

AI Summary

Protein-specific RNA design is hindered by reliance on extensive experimental data or high-resolution structural priors. Method: This paper proposes a conditional generative framework that requires neither labeled data nor pre-built structural models. We introduce Bidirectional Anchored Generation (BAnG), a novel paradigm jointly modeling the locality and context-dependency of functional motifs. Built upon a deep sequence generation architecture, BAnG integrates bidirectional attention, anchor-aware decoding, and a motif-aware loss function. Contribution/Results: On both synthetic benchmarks and real-world protein–RNA binding tasks, our method substantially outperforms existing approaches. It successfully generates multiple novel RNA sequences computationally validated to exhibit high binding affinity. The framework establishes a new, efficient, general-purpose, and data-light paradigm for targeted RNA design.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Multiagent Systems: Mechanism Design

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Designing RNA molecules that interact with specific proteins is a critical challenge in experimental and computational biology. Existing computational approaches require a substantial amount of experimentally determined RNA sequences for each specific protein or a detailed knowledge of RNA structure, restricting their utility in practice. To address this limitation, we develop RNA-BAnG, a deep learning-based model designed to generate RNA sequences for protein interactions without these requirements. Central to our approach is a novel generative method, Bidirectional Anchored Generation (BAnG), which leverages the observation that protein-binding RNA sequences often contain functional binding motifs embedded within broader sequence contexts. We first validate our method on generic synthetic tasks involving similar localized motifs to those appearing in RNAs, demonstrating its benefits over existing generative approaches. We then evaluate our model on biological sequences, showing its effectiveness for conditional RNA sequence design given a binding protein.

Problem

• Design RNA sequences for protein interactions
• Overcome limitations of existing computational approaches
• Generate RNA sequences without extensive experimental data

Innovation

• Deep learning model for RNA sequence generation
• Bidirectional Anchored Generation (BAnG) method
• Generates RNA sequences without experimental data