Exploring the Efficiency of Deep Graph Neural Networks for RNA Secondary Structure Prediction

Abstract

Ribonucleic acid (RNA) plays a vital role in various biological processes and forms intricate secondary and tertiary structures associated with its functions. Predicting RNA secondary structures is essential for understanding the functional and regulatory roles of RNA molecules in biological processes. Traditional free-energy-based methods for predicting these structures often fail to capture complex interactions and long-range dependencies within RNA sequences. Recent advancements in machine learning, particularly with graph neural networks (GNNs), have shown promise in enhancing the ability to model the relationships between molecular sequences and their structures. This work specifically explores the efficacy of various GNN architectures in modeling RNA secondary structure. Through benchmarking the GNN methods against traditional energy-based models on standard datasets, our analysis demonstrates that GNN models improves traditional methods, offering a robust framework for accurate RNA structure prediction.