Probing the toxicity of nanoparticles: a unified in silico machine learning model based on perturbation theory

Abstract

Nanoparticles (NPs) are part of our daily life, having a wide range of applications in engineering, physics, chemistry, and biomedicine. However, there are serious concerns regarding the harmful effects that NPs can cause to the different biological systems and their ecosystems. Toxicity testing is an essential step for assessing the potential risks of the NPs, but the experimental assays are often very expensive and usually too slow to flag the number of NPs that may cause adverse effects. In silico models centered on quantitative structure-activity/toxicity relationships (QSAR/QSTR) are alternative tools that have become valuable supports to risk assessment, rationalizing the search for safer NPs. In this work, we develop a unified QSTR-perturbation model based on artificial neural networks, aimed at simultaneously predicting general toxicity profiles of NPs under diverse experimental conditions. The model is derived from 54,371 NP-NP pair cases generated by applying the perturbation theory to a set of 260 unique NPs, and showed an accuracy higher than 97% in both training and validation sets. Physicochemical interpretation of the different descriptors in the model are additionally provided. The QSTR-perturbation model is then employed to predict the toxic effects of several NPs not included in the original dataset. The theoretical results obtained for this independent set are strongly consistent with the experimental evidence found in the literature, suggesting that the present QSTR-perturbation model can be viewed as a promising and reliable computational tool for probing the toxicity of NPs.

Nano-(Q)SAR for Cytotoxicity Prediction of Engineered Nanomaterials

“…The predictions were found to be in very good agreement with the experimental evidence, confirming that Ni-nanoparticles are not ecotoxic when compared with other NPs [67]. Further, a unified in silico machine learning model based on artificial neural networks was developed by Concu and co-authors [68]; the model was aimed to simultaneously predict general toxicity profiles of NPs under diverse experimental conditions. Application of perturbation theory to a set of 260 unique NPs showed higher accuracy of more than 97%.…”

“…Application of perturbation theory to a set of 260 unique NPs showed higher accuracy of more than 97%. Two families of descriptors were used in this study: Physico-chemical and 2D topological [68].…”

Application of Bayesian networks in determining nanoparticle-induced cellular outcomes using transcriptomics

“…QSTR-perturbation models have dealt also with relatively small datasets of cases, while combining multiple endpoints, experimental conditions and toxicities/activities (Concu et al 2017; Speck-Planche 2015). There are three ways that approach could be fused to our study: The dataset in this study can be preprocessed and modeled using the QSTR-perturbation model of Concu et al (2017); the Concu et al (2017) dataset can go through the CFS and BN construction and preprocessing stages to be modeled by BNs; finally, a BN can be used to model the QSTR final classification function, instead of the ANN used (Concu et al 2017).…”

Practices and Trends of Machine Learning Application in Nanotoxicology

“…Neural Networks were conceived based on functions of the central nervous system and became very popular in discovering relationships between parameters [88]. Different architectures and topologies were noted in the reviewed studies such as RBF, MLP, and GRNN [122]. In MLP, each network is built from several layers connected by weights.…”

“…Different models that use measured p-chem properties and experimental data, including biological data, exploit all the features since those properties are nano-specific [60,77]. QSAR-perturbation models, in addition to classical QSARs, make use of all available descriptors by generating several pairs of variables using the moving average approach [122,125]. Contrary to the feature reduction problem of theoretical generated descriptors, using nano-specific properties comes with data lacunas and the need for more descriptors.…”