Classification nano-SAR modeling of metal oxides nanoparticles genotoxicity based on comet assay data
Résumé
In nearly a decade of vigorous attempt in the toxicology and exposure research carried out to provide evidence for the assessment of health and environmental risks of nanomaterials (NMs), some progress has been made in generating the health effects and exposure data needed to perform risk assessment and develop risk management guidance. Quantitative Structure Activity Relationship ((Q)SAR) models are a powerful tool for rapid screening of large numbers and types of materials with advantage of saving time, funds and animal suffering. In this work we present first (Q)SAR models developed to predict genotoxicity of metal oxide NMs by using large initial sets of nano descriptors. We used a dataset containing in vitro comet assay genotoxicity for 16 nano metal oxides with different chemical core composition. This multi-source data was retrieved from genotoxicity profiles collected in our previous work. To properly analyse the data, we used a weight of evidence approach for evaluation of quality of the comet in vitro data for (Q)SAR modelling. Subsequently, based on the quality of checked dataset, we assigned genotoxic or non-genotoxic property to each metal core composition. By employing orthogonal partial least squares–discriminant analysis (OPLS-DA) method, nano-(Q)SAR models were derived with significant predictive power: accuracy 0.83 and 1. Conventional molecular descriptors and quantum chemical descriptors together with descriptors based on metal-ligand binding properties have been analysed to discuss the key factors affecting genotoxicity of metal oxide NMs. All derived models involve descriptors that describe possible structural factors influencing genotoxic behaviour of metal oxide NMs.