Predicting Nanomaterials toxicity pathways based on genome-wide transcriptomics studies using Bayesian networks

Abstract

The potential toxicity of Nanomaterials (NMs) is widely documented but risk assessment continues to pose a challenge. In this study, data derived from toxicogenomic studies are used to build a Bayesian Network (BN) model. This approach integrates transcriptomics data to successfully predict a number of biological effects. The model uses experimental conditions such as dose, duration and cell type along with NM physicochemical properties, and is developed to predict the effects of NM exposure on in vitro biological systems. The model version proposed in this study is shown to successfully predict a number of biological processes with a success rate >80% for most outcomes. The model validation shows a robust and promising methodology for incorporating transcriptomics studies in a hazard and, extendedly, risk assessment modelling framework.