Tutorial: End-to-End Deep Learning Pipeline for RNA-RNA Interaction Prediction

Bridging methodological advances and practical applications in computational biology requires clear, reproducible tutorials. This chapter provides an end-to-end guide for constructing deep learning pipelines for RNA-RNA interaction (RRI) prediction. Starting from raw RNA sequences, it covers dataset curation, negative sampling, feature engineering, model selection, training, evaluation, and interpretability using explainable AI techniques. Multiple modeling strategies are illustrated, including sequence-based Transformers for capturing long-range dependencies and structure-aware Graph Neural Networks (GNNs) that encode secondary structures and topological relationships. Hybrid architectures integrating sequence and structural information are also demonstrated to improve predictive performance on complex RNAs. Emphasis is placed on reproducibility, with recommendations for workflow management, containerization, version-controlled datasets, and FAIR-compliant pipelines to facilitate sharing and reuse. Transparent discussion of design choices, preprocessing, and evaluation enables readers to develop robust, interpretable, and transferable RRI prediction models. This chapter equips students, researchers, and practitioners with practical tools to apply advanced deep learning to real-world RNA biology problems.