Progression and prospects of machine learning techniques in nanotoxicology: riding the AI-driven wave

The widespread application of nanoparticles (NPs) has led to an increasing number of NPs being distributed in the ecological environment. This has raised concerns about human health and promoted the development of nanotoxicology. Traditional toxicity assessments, limited by high costs and time consumption, make machine learning (ML) an attractive alternative. ML models, particularly deep learning (DL) networks, can predict NP toxicity by analyzing extensive datasets, providing a more efficient and ethical method compared to animal testing. This review systematically summarizes the applications and challenges of ML in nanotoxicology. It discusses the importance of NPs properties in toxicity prediction and the difficulties in modeling the dynamic interactions with biological systems. The potential of integrating ML with other computational approaches to improve toxicity assessment is also considered. Despite progress, ML faces challenges such as limited training data, issues with model interpretability, and the complexity of nanomaterial-biological interactions. Overcoming these challenges requires enhanced data collection, interdisciplinary collaboration, and more directed ML models. Looking forward, the integration of ML with nanotoxicology is poised to revolutionize toxicity assessments, facilitating the development of safer nanotechnology applications.