Discriminating toxicant classes by mode of action: 3. Substructure indicators

Area of Science:

• Computational toxicology
• Cheminformatics
• Predictive modeling

Background:

• Quantitative Structure-Activity Relationships (QSARs) are crucial for predicting chemical toxicity.
• Selecting appropriate QSAR models is challenging due to varying applicability domains and prediction reliability.
• Substructure indicators offer a potential method for pre-filtering chemicals based on toxicity mechanisms.

Purpose of the Study:

• To develop and evaluate a stepwise procedure for selecting suitable QSARs for predictive toxicology.
• To assess the effectiveness of substructure indicators in pre-filtering compounds for baseline versus excess toxicity.
• To improve the reliability and applicability range of QSAR-based toxicity predictions.

Main Methods:

• A stepwise procedure was proposed, starting with a pre-filtering tier based on substructure indicators.
• A test set of 115 chemicals across 9 Mechanism of Action (MOA) classes was used.
• Contingency table statistics evaluated the performance of various classification schemes for discriminatory power.

Main Results:

• No single substructure-based classification scheme provided sufficient applicability and reliability for pre-filtering chemical inventories.
• The discriminatory power of different schemes varied significantly.
• Combining three classification schemes demonstrated major improvements in performance.

Conclusions:

• A combined approach using three classification schemes enhances the protective assignment of baseline toxicants and acceptable recognition of excess toxicants.
• The combined strategy favorably increases the overall applicability range for QSAR-based toxicity prediction.
• This refined procedure offers improved decision support for selecting QSARs in predictive toxicology.