An Overview of Machine Learning and Big Data for Drug Toxicity Evaluation

Area of Science:

• Pharmacology and Toxicology
• Computational Chemistry
• Bioinformatics

Background:

• Drug toxicity evaluation is critical, causing ~30% of candidate attrition.
• In silico models offer early toxicity assessment to reduce costs and time.
• Machine learning (ML) is increasingly used but limited by small datasets.

Purpose of the Study:

• Review common ML methods for drug toxicity assessment.
• Provide an overview of available toxicity data and tools for in silico models.
• Highlight opportunities and challenges in big toxicity data for drug safety.

Main Methods:

• Literature review of ML applications in drug toxicity.
• Analysis of big toxicity data types (e.g., molecular descriptors, toxicogenomics).
• Overview of existing in silico toxicity prediction tools and datasets.

Main Results:

• Common ML methods in toxicity assessment are identified.
• Examples of ML-driven toxicity studies are presented.
• The landscape of big toxicity data and its utility is described.

Conclusions:

• Advances in ML and big toxicity data can improve drug safety evaluation.
• In silico models are beneficial for early-stage toxicity screening.
• Addressing data limitations is key for robust toxicity prediction.