Topology-Driven Negative Sampling Enhances Generalizability in Protein-Protein Interaction Prediction

Unraveling the human interactome to understand biological processes and uncover disease-specific patterns hinges on accurate protein-protein interaction (PPI) predictions. However, challenges persist in machine learning (ML) models due to a scarcity of quality hard negative samples, shortcut learning, and limited generalizability to novel proteins. Here, we introduce ComPPlete (Completing the Protein-Protein Interaction Network), a novel ML pipeline utilizing PPI network topology for strategic sampling of protein-protein non-interactions (PPNIs) by leveraging higher-order network characteristics that capture the inherent complementarity-driven mechanisms of PPIs. Integrating unsupervised pre-training in protein representation learning with topological PPNI samples, ComPPlete improves PPI prediction generalizability and interpretability, particularly in identifying potential binding sites locations on amino acid sequences. ComPPlete strengthens the prioritization of screening assays, facilitates the transferability of ML predictions across protein families and homodimers. ComPPlete establishes the foundation for a fundamental negative sampling methodology in graph machine learning by integrating insights from network topology.