In silico models to predict dermal absorption from complex agrochemical formulations

Abstract

Dermal absorption is a critical part in the risk assessment of complex mixtures such as agrochemical formulations. To reduce the number of in vivo or in vitro absorption experiments, the present study aimed to develop an in silico prediction model that considers mixture-related effects. Therefore, an experimental 'real-world' dataset derived from regulatory in vitro studies with human and rat skin was processed. Overall, 56 test substances applied in more than 150 mixtures were used. Descriptors for the substances as well as the mixtures were generated and used for multiple linear regression analysis. Considering the heterogeneity of the underlying data set, the final model provides a good fit (r² = 0.75) and is able to estimate the influence of a newly composed formulation on dermal absorption of a well-known substance (predictivity Q²Ext = 0.73). Application of this model would reduce animal and non-animal testings when used for the optimization of formulations in early developmental stages, or would simplify the registration process, if accepted for read-across.