Artificial Intelligence and Machine Learning-Based Approaches for Genetic Damage Prediction.

Area of Science:

• Pharmacology and Toxicology
• Computational Chemistry
• Genetics

Background:

• Genetic toxicology examines how physical and chemical agents impact genetic material.
• Genotoxic events like chromosomal abnormalities can lead to adverse drug reactions, mutagenicity, and carcinogenicity.
• Some drugs intended for therapeutic use have been found to induce genotoxicity.

Purpose of the Study:

• To categorize assays used for quantifying genotoxic damage, such as the Ames test.
• To elaborate on artificial intelligence (AI) and machine learning (ML) approaches for predicting genetic damage.
• To provide an overview of genotoxicity prediction tools, models, and evaluation metrics.

Main Methods:

• Categorization of genotoxicity assays (e.g., Ames test).
• Review of AI/ML models for genotoxicity prediction, including Quantitative Structure-Activity Relationship (QSAR), Machine Learning (ML), and Deep Learning (DL).
• Utilization of diverse molecular descriptors and fingerprints (topological, electrostatic, quantum) for predictive modeling.

Main Results:

• AI models, including QSAR, ML, and DL, are effective in predicting genotoxic damage.
• Various molecular descriptors and fingerprints are employed in genotoxicity prediction studies.
• A comprehensive datasheet of models and research focused on genotoxicity prediction is available.

Conclusions:

• AI models offer a promising approach for assessing genotoxicity.
• Accurate prediction of genotoxic risks is crucial for safe drug development.
• Advancements in computational methods enhance the evaluation of potential genotoxic agents.