Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Cefic LRI AIMT-8: Prediction of STOTRE classification by new approach methodologies

: Escher, Sylvia E.; Cronin, Mark T.D.; Firman, J.; Magdziarz, T.; Rathman, J.; Yang, C.

The Toxicologist 168 (2019), No.1, Abstract PS 1861
ISSN: 0731-9193
Society of Toxicology (Annual Meeting) <58, 2019, Baltimore/Md.>
Fraunhofer ITEM ()

The AIMT-8 project aims to assess the ability of in vitro data from the Tox21 program to predict STOT-RE categories of a range of chemicals. The application of New Approach Methodologies (NAMs) in risk assessment is an area of intensive research. AIMT-8 aims to advance the understanding of the use of NAMs by analysing a different way of predicting systemic toxicity, especially the STOT-RE classification. STOT-RE classification is based on the NOAEL of the in vivo study and does not consider the type of effect or the organ affected. If prediction of STOT-RE classification by NAMs is possible, this will contribute to a paradigm shift in risk assessment and will motivate the use of NAMs in prioritisation and labelling, and eventually in safety assessment as well. STOT-RE classifications were gathered and derived from different sources e.g. from a set of 90 day studies with repeated oral exposure (extracted from the RepDose/ToxRef/Hess and Cosmos databases) and the inventory of harmonised classifications provided by ECHA. In parallel, we analysed the AC50 values from Tox21 and considered all values occurring at sub-cytotoxic concentrations. The spare data matrix of the 43 individual assays was aggregated to seven categories representing six toxicity pathways and cytotoxicity values. The intersection of both the in vivo and in vitro data resulted in a data set of 749 compounds. For later in vitro to in vivo extrapolation (IVIVE) relevant data such as plasma protein binding, renal and hepatic clearance were identified from existing data sets and the literature. Prior to in silico profiling, the structural information was quality controlled and corrected. From a range of statistical methods, k-nearest neighbours (kNN) and random forest (RF) approaches were selected for the development of the classification model. In addition, seventeen read-across groups were defined. The grouped compounds share structural characteristics and specific/unspecific in vivo apical findings/target organs. Groups were distinguished for which a shared toxicological effect pattern might be indicative of a shared mode of action from those with unspecific toxicological effects e.g. weight changes or no toxicological effects. In these read-across groups, we analysed the mechanistic links between the in vitro results and the in vivo apical endpoint leading to STOT-RE classification. Financial support for this work was provided by the CEFIC LongRange Research Initiative (CEFIC LRI AIMT-8).