Predicting drug-target interactions by measuring confidence with consistent causal neighborhood interventions

Area of Science:

• Bioinformatics
• Computational Chemistry
• Drug Discovery

Background:

• Drug-target interaction (DTI) prediction is vital for drug discovery.
• Knowledge graph embedding (KGE) methods show promise but often lack accuracy in DTI prediction.
• Current methods lead to high misjudgment rates and reduced drug development efficiency.

Purpose of the Study:

• To refine the accuracy of DTI prediction models using KGE.
• To enhance DTI prediction precision through causal intervention (CI) confidence measures.
• To provide insights for guiding future drug development experiments.

Main Methods:

• Developed and applied causal intervention confidence measures to assess KGE triplet scores.
• Conducted comparative experiments on three datasets using nine KGE models.
• Analyzed the embedding of intervention values for deeper insights.

Main Results:

• The proposed CI confidence measure approach significantly improved DTI link prediction accuracy.
• Outperformed traditional DTI prediction approaches in comparative experiments.
• Experimental analysis provided valuable insights into intervention value embeddings.

Conclusions:

• Causal intervention confidence measures enhance the accuracy and reliability of DTI prediction.
• The findings offer valuable guidance for optimizing drug development processes.
• This approach contributes to more efficient and precise drug discovery pipelines.