Large Language Models in RNA-RNA Interaction Prediction

This chapter explores the transformative role of large language models (LLMs) in predicting RNA-RNA interactions (RRIs). By leveraging the sequence-based learning capabilities of LLMs, this chapter illustrates how these models can capture complex patterns in nucleotide sequences, structural motifs, and functional dependencies without explicit feature engineering. We discuss state-of-the-art architectures adapted for RNA data, including Transformer-based models and protein-inspired sequence embeddings, highlighting their performance in both supervised and self-supervised settings. Practical considerations for training, fine-tuning, and evaluating LLMs on RRI datasets are provided, alongside strategies to interpret model predictions using explainable AI approaches. Case studies demonstrate how LLMs can accelerate discovery in non-coding RNA regulation, RNA therapeutics design, and the broader transcriptomic landscape. Finally, challenges such as data sparsity, generalization to novel RNA families, and computational efficiency are examined, offering guidance for future research at the intersection of natural language processing and RNA biology.