Explainable Artificial Intelligence for RNA-RNA Interaction Models

As deep learning models become increasingly used to RNA-RNA interaction (RRI) prediction, concerns regarding their interpretability, reliability, and biological validity have grown. High-performing models such as Transformers and Graph Neural Networks often function as black boxes, limiting their acceptance in biological discovery and therapeutic applications. Explainable Artificial Intelligence (XAI) seeks to address this challenge by providing tools and methodologies that make model predictions transparent, interpretable, and biologically meaningful. This chapter presents a comprehensive overview of XAI techniques applicable to RRI prediction models. We discuss model-agnostic and model-specific explanation methods, attention-based interpretability in Transformers and GNNs, feature attribution techniques, and biologically grounded validation strategies. Case studies from recent RRI and RNA-interaction prediction studies illustrate how explainability can highlight sequence regions, interaction sites, structural features, and other model-derived signals that can support biological interpretation and hypothesis generation.