Integrating surface chemistry properties and machine learning to map the toxicity landscape of superparamagnetic iron oxide nanoparticles

Abstract

The relationship between Superparamagnetic Iron Oxide Nanoparticles (SPIONs) surface chemistry and their toxicological outcomes is crucial for biomedical applications, including drug delivery and imaging diagnostics. SPIONs' surface properties—such as size, shape, type of coating agents, and charge—are directly linked to their interactions with the biological environment, significantly affecting their toxicity. Surface chemistry plays a significant role in determining biocompatibility, cellular uptake, and the potential for adverse reactions. This study focuses on building a classification and prediction model based on the experimentally obtained properties and linked with the calculated molecular descriptors to describe the nature of the various coatings used for SPIONs in such a combined mode. The predictive model helps identify how specific surface modifications, including coating types and functional groups, influence toxicity responses. The results that were obtained, which correlate well with the existing literature, confirm the effects of surface chemistry on toxicity. For instance, the model accurately predicts that chitosan derivative coatings with a higher positive charge exhibit toxic potential, which aligns with previous findings. Incorporating these experimentally obtained surface features into a predictive framework enables the design of safer SPION formulations, enhancing therapeutic efficacy while managing surface chemistry's effects on toxicity.