A Statistical Exploration of QSAR Models in Cancer Risk Assessment: A Case Study on Pesticide-Active Substances and...

Area of Science:

• Environmental chemistry
• Toxicology
• Computational chemistry

Background:

• New Approach Methodologies (NAMs) are crucial for chemical hazard identification.
• (Quantitative) structure-activity relationship ((Q)SAR) models estimate toxicity based on chemical structure.
• Metrics for assessing chemical congruence with training sets and risk categorization are vital.

Purpose of the Study:

• To investigate the coherence of different (Q)SAR models for pesticide carcinogenicity assessment.
• To evaluate the importance of applicability domain definitions in (Q)SAR modeling.
• To explore the potential and limitations of NAMs in chemical safety evaluation.

Main Methods:

• Methodological study focusing on (Q)SAR models.
• Assessment of carcinogenicity for pesticide-active substances and metabolites.
• Analysis of congruence metrics and risk categorization.

Main Results:

• Significant potential of (Q)SAR models for toxicity assessment was confirmed.
• Inconsistencies in results across different (Q)SAR models were observed.
• The importance of a priori defined and user-transparent applicability domains was highlighted.

Conclusions:

• (Q)SAR models are valuable tools for chemical hazard assessment, but limitations exist.
• Careful definition of the applicability domain is critical for reliable (Q)SAR predictions.
• Transparent strategies for applicability domain definition are essential for integrating NAMs.