Integrated decision support for assessing chemical liabilities.

Area of Science:

• Computational chemistry
• Drug discovery
• Toxicology

Background:

• Chemical liabilities significantly impact drug discovery, necessitating efficient prediction methods.
• Current in silico approaches lack integrated, extensible decision support systems with interpretable results.
• Reducing costs and animal testing requires advanced computational tools for liability assessment.

Purpose of the Study:

• To develop an integrated and extensible decision support system for chemical liability assessment.
• To improve the speed and interpretability of in silico chemical liability predictions.
• To provide a platform for hypothesis testing of chemical structures and compound collections.

Main Methods:

• Integration of similarity searches, structural alerts, and Quantitative Structure-Activity Relationship (QSAR) models.
• Utilizing the Bioclipse workbench for computational modeling and data management.
• Highlighting important substructures for prediction interpretability and interactive visualization.

Main Results:

• Demonstrated a method integrating multiple computational approaches for chemical liability prediction.
• Achieved interpretable results by visualizing substructures crucial for predictions.
• Successfully applied the system to predict mutagenicity, carcinogenicity, and aryl hydrocarbon receptor (AhR) activation.

Conclusions:

• The developed system offers an integrated, extensible, and interpretable solution for chemical liability assessment in drug discovery.
• The approach facilitates rapid hypothesis testing and decision-making in the drug development pipeline.
• The free, open-source Bioclipse decision support system is available for broader scientific use.