Progress in computational toxicology.

Area of Science:

• Toxicology
• Computational Science
• Data Science

Background:

• Computational methods are increasingly vital in toxicology for pharmaceutical, consumer product, and environmental safety assessments.
• A decade of progress has been made in developing and applying computational toxicology approaches.
• The field leverages big data for more comprehensive safety evaluations.

Purpose of the Study:

• To review the progress in computational toxicology over the past decade.
• To evaluate the performance of machine learning models in predicting toxicological endpoints.
• To compare Bayesian and Support Vector Machine (SVM) learning methods using established datasets.

Main Methods:

• A comprehensive literature review was conducted on computational models for key toxicities: hepatotoxicity (including drug-induced liver injury - DILI), cardiotoxicity, renal toxicity, and genotoxicity.
• Machine learning methodologies were highlighted and analyzed.
• Bayesian and SVM algorithms were compared using existing computational toxicology model datasets.

Main Results:

• The availability of extensive toxicological data has facilitated the development and application of machine learning models for predictive toxicology.
• Cross-validation data indicates that Bayesian and SVM methods exhibit comparable performance across various predictive toxicology models.
• Machine learning models are increasingly effective for predicting toxicological outcomes.

Conclusions:

• Significant advancements in computational toxicology have occurred over the last decade, marked by improved model development and the emergence of big data approaches.
• Future toxicological data generation will yield vast compound libraries suitable for machine learning models.
• These enhanced models will cover a broad chemical space for diverse applications in pharmaceuticals, consumer products, and environmental science.