TY - GEN
T1 - Advancing Drug-Drug Interaction Prediction using Multi-Modal Feature Integration with Graph Neural Networks
AU - Chianumba, Ernest C.
AU - Liu, Hao
AU - Varde, Aparna S.
AU - Zhuang, Yan
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Pharmaceutical treatments are essential for managing medical conditions, but drug-drug interactions (DDIs) pose significant risks. This research integrates Knowledge Graphs and Graph Neural Networks to predict DDIs by exploring drug relationships. We construct a knowledge graph using DrugBank data (1,000 drugs, 155,774 interactions) enriched with PubChem features, then enhance our approach by integrating transformer-based embeddings (ChemBERTa, SPECTER, and SBERT) to create 1152-dimensional feature vectors. Formulating DDI prediction as link prediction, we compare three GNN architectures: Graph Convolutional Network (GCN), GraphSAGE, and Graph Attention Network (GAT). With basic molecular features, GCN achieved 75.65% accuracy (80.17% F1). After multimodal integration, performance improved across all models, with GAT showing the greatest enhancement (80.61% accuracy, 82.57% F1). These results highlight the value of integrating diverse data modalities for DDI prediction and the potential for enhancing medication safety in polypharmacy scenarios.
AB - Pharmaceutical treatments are essential for managing medical conditions, but drug-drug interactions (DDIs) pose significant risks. This research integrates Knowledge Graphs and Graph Neural Networks to predict DDIs by exploring drug relationships. We construct a knowledge graph using DrugBank data (1,000 drugs, 155,774 interactions) enriched with PubChem features, then enhance our approach by integrating transformer-based embeddings (ChemBERTa, SPECTER, and SBERT) to create 1152-dimensional feature vectors. Formulating DDI prediction as link prediction, we compare three GNN architectures: Graph Convolutional Network (GCN), GraphSAGE, and Graph Attention Network (GAT). With basic molecular features, GCN achieved 75.65% accuracy (80.17% F1). After multimodal integration, performance improved across all models, with GAT showing the greatest enhancement (80.61% accuracy, 82.57% F1). These results highlight the value of integrating diverse data modalities for DDI prediction and the potential for enhancing medication safety in polypharmacy scenarios.
KW - Clinical Decision Support
KW - Drug-Drug Interaction
KW - Graph Neural Networks
KW - Health Informatics
KW - Knowledge Graphs
KW - Multi-Modal Learning
KW - Transformer Embeddings
UR - https://www.scopus.com/pages/publications/105033543027
U2 - 10.1109/BIBM66473.2025.11357147
DO - 10.1109/BIBM66473.2025.11357147
M3 - Conference contribution
AN - SCOPUS:105033543027
T3 - Proceedings - 2025 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2025
SP - 6559
EP - 6566
BT - Proceedings - 2025 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2025
A2 - Liu, Juan
A2 - Huang, Jingshan
A2 - Wang, Xiaowo
A2 - Zhang, Fa
A2 - Zou, Xiufen
A2 - Tian, Tian
A2 - Hu, Xiaohua
A2 - Hu, Bin
A2 - Xiong, Yi
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2025 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2025
Y2 - 15 December 2025 through 18 December 2025
ER -