Development of QSAR-based two-stage prediction model for estimating mixture toxicity

Abstract

Conventionally, concentration addition (CA) and independent action (IA) models based on additive toxicity are often used to estimate the mixture toxicity of similarly- and dissimilarly-acting chemicals, respectively. A two-stage prediction (TSP) model has been developed as an integrated addition model that can perform the CA and IA calculations stage by stage. However, the use of the conventional TSP model is limited if the mode of toxic action (MoA) for every mixture component is not readily known. The aim of this study was to develop and evaluate a quantitative structure–activity relationship-based TSP (QSAR-TSP) model for estimating mixture toxicity in the absence of knowledge on the MoAs of the constituents. For this purpose, different clustering methods of mixture constituents using computerized analysis based on the structural similarity between chemicals were applied as a part of the predictions of mixture toxicity. The relative importance of molecular descriptors was additionally determined by Random Forest analysis. This study highlights the prediction power of the QSAR-TSP model and its potential to overcome the limitations of the conventional TSP model, and how clustering methods of mixture components that employ chemical structural information to categorize might be applied to predict mixture toxicity effectively.