In silico modelling of hazard endpoints: current problems and perspectives

Area of Science:

• Computational chemistry and toxicology
• Predictive modeling for chemical risk assessment

Background:

• Quantitative structure-activity relationships (QSAR) are crucial for regulatory chemical assessment but face challenges.
• Current QSAR models often oversimplify chemical structures and lack comprehensive databases for toxicity endpoints.
• Defining the applicability domain of QSAR models remains a significant obstacle for regulatory acceptance.

Purpose of the Study:

• To address major scientific hurdles hindering the regulatory acceptance of QSAR models.
• To propose and demonstrate approaches for enhancing QSAR by considering molecular complexity and biological transformations.
• To improve the predictability and interpretability of QSAR models for risk assessment.

Main Methods:

• Developing mechanistic models of chemical structure, including conformational flexibility and metabolic activation.
• Utilizing systematic databases for risk assessment endpoints and clustering reactive chemicals by toxicity pathways.
• Creating computational tools to define the applicability domain of QSAR models within regulatory chemical spaces.

Main Results:

• Improved QSAR predictability for chemical mutagenicity by incorporating molecular flexibility and metabolic activation.
• Demonstrated enhanced QSAR interpretations through mechanistic considerations.
• Presented an applicability domain for a QSAR model predicting estrogen receptor binding within a mechanistically-defined chemical space.

Conclusions:

• Addressing molecular complexity and metabolic activation significantly enhances QSAR predictability and interpretation.
• Systematic databases and robust applicability domain definitions are essential for regulatory QSAR acceptance.
• A strategic approach to model development and validation is necessary to meet regulatory criteria for risk assessment.