KGOT: Unified Knowledge Graph and Optimal Transport Pseudo-Labeling for Molecule-Protein Interaction Prediction
The framework integrates diverse biological datasets into a large knowledge graph to overcome data scarcity in molecule-protein interaction prediction. It uses an optimal transport approach to generate high-quality pseudo-labels for unlabeled pairs, effectively bridging the gap between sparse true labels and rich biological context. This method significantly improves prediction accuracy and zero-shot performance.