KGANSynergy: knowledge graph attention network for drug synergy prediction

Area of Science:

• Computational Biology
• Pharmacology
• Bioinformatics

Background:

• Combination therapy is crucial for complex diseases, especially when monotherapy fails.
• Drug combinations can overcome resistance and enhance treatment efficacy, particularly in cancer.
• High-throughput screening for synergistic drug combinations is costly and challenging due to vast combinatorial possibilities.

Purpose of the Study:

• To develop an effective computational approach for predicting drug synergy.
• To address the challenges of high-throughput synergistic drug combination screening.
• To leverage biomedical information and knowledge graphs for improved drug combination identification.

Main Methods:

• Proposed a novel end-to-end Knowledge Graph Attention Network (KGANSynergy).
• Utilized hierarchical propagation in knowledge graphs to identify multi-source neighbor nodes for drugs and cell lines.
• Employed a multi-attention mechanism to weigh neighbor importance and aggregate entity information for enriched embeddings.

Main Results:

• KGANSynergy effectively utilizes neighbor information of drugs and cell lines.
• The method demonstrated superior performance compared to existing competing methods in predicting drug synergy.
• Learned drug and cell line embeddings proved effective for synergy prediction.

Conclusions:

• KGANSynergy offers an effective computational solution for identifying synergistic drug combinations.
• The approach enhances the prediction of drug synergy by leveraging knowledge graph attention mechanisms.
• This method holds potential for advancing the development of effective combination therapies.