Risk assessment of heterogeneous TiO2-based engineered nanoparticles (NPs): a QSTR approach using simple periodic table based descriptors.

Nowadays, the risk assessment of engineered nanoparticles (NPs) on human health and animals is of great importance. We have used here simple periodic table based descriptors for mixture compounds to predict the cytotoxicity for the heterogeneous NPs. We have developed mono parametric quantitative structure-toxicity relationship (QSTR) models for 34 TiO2-based NPs modified with (poly) metallic clusters of noble metals (Au, Ag, Pt) to assess the cytotoxicity (-log EC50) towards Chinese Hamster Ovary cell line. After critical statistical analysis of the developed five linear regression (LR) models, we found that the derived models are close to each other in terms of different metric values (R2 = 0.922-0.926; Q2 = 0.907-0.911; R2adj = 0.918-0.922; Q2F1 = 0.930-0.938; Q2F2 = 0.924-0.932). Thus, we have developed a partial least squares (PLS) model using the five descriptors obtained from the five LR models. The developed PLS model showed good predictivity and robustness in terms of both internal (R2 = 0.925; Q2 = 0.911) and external validation (Q2F1 = 0.944; Q2F2 = 0.938) parameters. The descriptors, Electrochemical equivalent (Eq), 2nd ionization potential (2χpi), covalent radius (Rc), amount of Ag (Agamt) and thermal conductivity (Tc) obtained from the final PLS model well explained the cause of cytotoxicity of the heterogeneous NPs without requiring any computationally expensive descriptors. The insights obtained from the developed models suggested that higher electronegativity, lower oxidation state, and release of metal cation from its oxide increase cytotoxicity through various mechanisms. Thus, these models can be used as efficient tools to assess the toxicity with physiological property of the new heterogeneous NPs in the future.