Nanomaterials (NMs) are an ever-increasing field of interest, due to their wide range of applications in science and technology. However, despite providing solutions to many societal problems and challenges, NMs are associated with adverse effects with potential severe damages towards biological species and their ecosystems. Toxicity testing is an essential step for assessing the potential risks of NMs, but the experimental assays are often very expensive and usually too slow to flag the number of NMs that may cause adverse effects. Alternative in silico modelling approaches such as the one proposed here, Quantitative Structure-Toxicity Relationships (QSTR), are increasingly used to guide the selection of suitable nanomaterials with a desired level of safety. In this communication we move a step forward and present results of a unified QSTR-perturbation modelling approach aimed at simultaneously predicting the general toxicity of metal oxide NMs under various experimental assay conditions. The QSTR-perturbation models derived, using a dataset of ca. 70 unique NMs, exhibited accuracies higher than 96% for both training and validation sets. In order to demonstrate the practical applicability of such models, the latter were then employed to predict the toxicity of some NMs not included in the original modelling dataset. The results for these independent sets were found to be strongly consistent with the experimental reports gathered in the literature. Overall, that thus suggest that the present QSTR-perturbation modelling approach can be viewed as a highly promising in silico tool for the fast and costefficiently assessment of the toxicity of metal oxide NMs. Finally, and most importantly, the developed models are able to provide important insights regarding the mechanism of the toxicity triggered by these NMs.