Using Pareto points for model identification in predictive toxicology.

Area of Science:

• Computational toxicology
• Cheminformatics
• Machine learning in drug discovery

Background:

• Predictive toxicology models are crucial for accelerating chemical compound discovery in pharmaceuticals, cosmetics, and food safety.
• Current methods for selecting appropriate predictive models rely on manual user decisions, which is inefficient and time-consuming.
• Managing and automatically identifying relevant models from large collections remains a significant challenge in the field.

Purpose of the Study:

• To introduce and evaluate an automated algorithm for identifying the most suitable predictive toxicology model from a collection.
• To address the open problem of automated model management and selection in predictive toxicology.
• To facilitate faster and more reliable chemical safety assessments.

Main Methods:

• Development of a novel algorithm based on Pareto optimality for mining model collections.
• Application of the algorithm to identify predictive models for specific toxicological endpoints.
• Verification of the algorithm's performance using IGC50 and LogP endpoints.

Main Results:

• The Pareto optimality-based algorithm successfully mines model collections to identify relevant predictive models.
• The approach demonstrated effective performance for predicting IGC50 and LogP values.
• The automated identification method shows significant potential for improving predictive toxicology workflows.

Conclusions:

• Automated model identification using Pareto optimality offers a promising solution for selecting reliable predictive toxicology models.
• This approach can significantly enhance the efficiency and accuracy of chemical safety assessments.
• The findings pave the way for more streamlined drug design and chemical product development processes.