TOXRIC: a comprehensive database of toxicological data and benchmarks

Area of Science:

• Environmental science and toxicology
• Computational chemistry and cheminformatics
• Pharmacology and drug discovery

Background:

• Assessing compound toxicity is crucial for environmental safety, human health, and drug development.
• Computational toxicology methods require extensive, standardized datasets and benchmarks for accurate predictions.
• Existing toxicological data is often fragmented, hindering systematic analysis and machine learning model development.

Purpose of the Study:

• To introduce TOXRIC, a comprehensive database for toxicological data, attributes, and benchmarks.
• To facilitate early-stage identification of compound toxicity and support computational drug discovery.
• To provide researchers with tools for understanding toxicological mechanisms and developing predictive models.

Main Methods:

• Curated and integrated toxicological data from diverse sources.
• Standardized attribute data including structural, target, transcriptome, and metabolic information.
• Development of practical benchmarks and visualization tools for molecular representations.

Main Results:

• TOXRIC houses data for 113,372 compounds across 13 toxicity categories and 1474 endpoints.
• Includes 39 feature types encompassing various molecular and biological descriptors.
• Provides ML-ready datasets, benchmarks, and visualization tools for endpoint prediction tasks.

Conclusions:

• TOXRIC serves as a valuable resource for toxicological investigations and computational method development.
• Facilitates compound/drug discovery by enabling better understanding of toxicity and selection of optimal prediction strategies.
• Enhances the interpretation of toxicological mechanisms through integrated data and visualization capabilities.