QSAR and machine learning modeling of toxicity of nanomaterials: a risk assessment approach

The advancement of nanoscience and enormous use of nanomaterials (NMs) in the form of coated nanoparticles, engineered metal oxide nanomaterials (MONMs), single- and multiwalled carbon nanotubes, fullerenes (C60/C70), and silica NMs open up multifaceted possibilities of inflicting toxicity on the environment. With the implications of NMs in medicine, cars, batteries, solar panels, textiles, toys, electronics, etc., one can't control the safe release of NMs into the ecosystem, which directly affects the organisms present in the aquatic and terrestrial environment and indirectly affects humans' lives. The testing of each individual form of NMs on diverse species as well as different response/endpoints by traditional experimental assays is an impossible task. Thus in most cases, environment regulatory bodies, industries, and environmental scientists largely depend on in silico methods like quantitative structure–activity relationships (QSARs) and machine learning approaches. In a relatively short time, these models can be prepared with the expertise of cheminformaticians and can be employed for the prediction of future NMs as well as the already existing ones in the ecosystem. Since the beginning of 2010, a huge number of in silico models have been developed in combination with in vivo and in vitro analysis, and the present number of such models exceeds 150. The successful models cover eukaryotic to prokaryotic organisms and cell lines including cytotoxicity, genotoxicity, enzymatic inhibition, egg hatching, cellular viability, and cellular uptake. Most of the models can identify the mechanistic interpretation behind the toxicity of NMs toward specific organisms/cell lines. This is helpful for the safe design of NMs for the future along with risk assessment.