pDILI_v1: A Web-Based Machine Learning Tool for Predicting Drug-Induced Liver Injury (DILI) Integrating Chemical...

Area of Science:

• Computational chemistry
• Pharmacology
• Machine learning applications in drug safety

Background:

• Drug-induced liver injury (DILI) poses significant risks in drug development and clinical practice.
• Predicting human DILI risk is crucial, yet the chemical space associated with DILI remains underexplored.

Purpose of the Study:

• To systematically investigate the structural determinants of DILI risk using computational methods.
• To explore the chemical space and scaffold diversity linked to DILI.
• To identify structural alerts (SAs) influencing DILI risk and develop predictive models.

Main Methods:

• Leveraged machine learning (ML) for a comprehensive computational analysis of DILI risk.
• Employed fragment-based approaches to identify DILI-associated structural alerts.
• Developed supervised ML models utilizing molecular fingerprints for DILI risk prediction and structural significance elucidation.

Main Results:

• Identified key structural features and scaffold diversity associated with DILI.
• Developed predictive ML models demonstrating potential for early hepatotoxicity risk identification.
• Introduced pDILI_v1, a user-friendly web application for DILI risk prediction and visualization.

Conclusions:

• The study advances drug safety evaluation by integrating ML predictions with chemical space analysis.
• The developed models and tools facilitate early identification of hepatotoxic risks in drug candidates.
• This research contributes to the development of safer pharmaceuticals and mitigation of DILI risks.