KGCN_NFM

KGCN_NFM integrates knowledge graph convolutional networks (KGCNs) and neural factorization machines (NFMs) to identify drug-drug interactions (DDIs) by combining KG-derived embeddings with Morgan molecular fingerprints for multimodal prediction.

Key Features:

• Knowledge Graph Convolutional Networks (KGCNs): KGCNs extract and learn high-order structural and semantic information from knowledge graphs (KGs) to generate comprehensive embeddings.
• Neural Factorization Machines (NFMs): NFMs consume KGCN-generated embeddings alongside Morgan molecular fingerprints to model feature interactions and predict DDIs.
• Morgan Molecular Fingerprints: Morgan molecular fingerprints provide detailed drug characterizations used as input features for NFMs.
• Multimodal Data Integration: Integrates heterogeneous data sources, including knowledge graphs and molecular fingerprints, to enhance DDI prediction accuracy.
• Scalability with Real-world Datasets: Processes real-world datasets of varying sizes to accommodate different data scales and complexity.
• Experimental Validation Techniques: Predictions have been validated using MTT assays, apoptosis experiments, cell cycle analysis, and molecular docking.
• Performance Evaluation: Evaluated against state-of-the-art algorithms and reported to outperform them in accuracy and effectiveness.

Scientific Applications:

• Drug Development: Predicts potential DDIs to inform safety and efficacy assessments during drug development.
• Synergistic Drug Combinations: Identifies synergistic effects between drugs, exemplified by validation of the topotecan and dantron combination showing synergistic anticancer activity against lung carcinoma.

Methodology:

Knowledge graphs encode biomedical information; KGCNs learn high-order embeddings from these KGs; embeddings are combined with Morgan molecular fingerprints and input to NFMs for DDI prediction while processing datasets of varying sizes.