Integrating Multi-Modal Biological Knowledge via Contrastive Dual-View Graph Learning for Phosphorylation Site-Disease Association Prediction.

CDVGL-PDA achieves high accuracy in predicting phosphorylation site-disease associations by integrating multi-modal biological data with contrastive dual-view graph learning.

• Why it matters: Understanding phosphorylation site-disease links is crucial for uncovering disease mechanisms and developing targeted therapies, but current methods lack comprehensive multi-modal data integration.
• What they did: The authors developed CDVGL-PDA, a framework that constructs a multi-modal heterogeneous graph with nine node types and ten edge types, incorporating sequence embeddings and semantic features, then applies contrastive learning and attention-based fusion for prediction.
• The result: The model demonstrates strong predictive performance across various datasets, effectively captures biological relationships, and can identify potential PDAs, supporting advances in disease research and therapeutic discovery.