ArcDFI: Attention regularization guided by CYP450 interactions for predicting drug-food interactions

Area of Science:

• Pharmacology
• Computational Chemistry
• Bioinformatics

Background:

• Cytochrome P450 (CYP450) isoenzymes play a critical role in drug-food interactions (DFI).
• Existing computational models for DFI prediction often lack generalizability and do not account for drug-CYP450 interactions (DCI).

Purpose of the Study:

• To develop a novel computational model, ArcDFI, for predicting drug-food interactions.
• To enhance DFI prediction by integrating drug-CYP450 interactions (DCI) into the model.
• To improve the generalizability and explainability of DFI prediction models.

Main Methods:

• Development of the ArcDFI model utilizing attention regularization.
• Guidance of the attention mechanism by known drug-CYP450 interactions (DCI).
• Evaluation of the model under stringent cold-drug and cold-food settings.

Main Results:

• ArcDFI significantly outperforms ten baseline approaches in DFI prediction.
• The model demonstrates improved predictive performance, especially in cold-drug and cold-food scenarios.
• Analysis of the attention mechanism provides insights into DCI's role in DFI prediction.

Conclusions:

• ArcDFI is the first DFI prediction model to incorporate drug-CYP450 interactions (DCI).
• Incorporating DCI enhances predictive generalizability and model explainability for drug-food interactions.
• The ArcDFI model offers a promising advancement in computational DFI prediction.