Machine Learning-Based Models to Predict Drug-Induced Liver Injury (DILI) to Assist Medicinal Chemistry.

Area of Science:

• Pharmacology
• Computational Biology
• Toxicology

Background:

• Drug-induced liver injury (DILI) is a major reason for drug attrition and market withdrawal.
• Current preclinical methods miss a significant percentage of clinical hepatotoxicity cases.
• Machine learning (ML) and deep learning (DL) show promise for DILI prediction in drug discovery.

Purpose of the Study:

• To review the evolution of DILI datasets and computational prediction methods.
• To highlight advancements in integrating causality, pharmacogenomics, and mechanistic models.
• To guide the development of more reliable DILI prediction strategies.

Main Methods:

• Review of DILI annotation datasets, from small to large-scale resources.
• Analysis of computational methods, from descriptor-based models to advanced DL and ensemble techniques.
• Examination of efforts to integrate causality frameworks, pharmacogenomics, and mechanistic insights.

Main Results:

• DILI datasets have grown significantly in size and comprehensiveness.
• Computational methods have advanced from basic models to sophisticated DL and ensemble approaches.
• Integration of diverse data types (causality, genomics, mechanisms) is enhancing predictive power.

Conclusions:

• Computational approaches, particularly DL, are crucial for improving DILI prediction accuracy.
• Integrating mechanistic and clinical data with computational models is key for real-world applicability.
• Further development of robust and consensual DILI prediction strategies is essential for drug safety.