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

Genetic toxicology is a scientific discipline investigating the impact of deleterious physical and chemical agents on the inheritance process. Genotoxic cellular processes such as chromosomal abnormalities, micronuclei production, sister chromatid exchanges, and cell death contribute to the development of numerous adverse drug reactions (ADRs), which ultimately lead to carcinogenic and mutagenic effects. Recently, there have been instances where drugs used to treat a specific disease have been found to induce genotoxicity in humans. This chapter concentrates on categorizing various assays like the Ames test, which can quantify the degree of genotoxic damage. Furthermore, the studies on artificial intelligence (AI) and machine learning (ML) approaches used to predict genetic damage are elaborated. In the context of genotoxicity prediction, AI models are typically categorized into three main types: Quantitative Structure-Activity Relationship (QSAR), Machine Learning (ML), and Deep Learning (DL). Additionally, a detailed datasheet is made available for the models, studies, and research that specifically focus on genotoxicity prediction. Researchers have used a wide range of molecular descriptors and fingerprints for the predictions, including topological, electrostatic, and intricate quantum descriptors. This chapter provides a fundamental overview of genotoxicity, including its prediction using various tools and software and its evaluation using prediction scores and different metrics. These advancements, as a whole, encourage the use of AI models to assess genotoxicity and more accurately predict potential genotoxic risks in the drug development process.