Host: The Japanese Society of Toxicology
Name : The 49th Annual Meeting of the Japanese Society of Toxicology
Date : June 30, 2022 - July 02, 2022
The evaluation of skin sensitization is important to confirm the safety of cosmetic products. In recent years, several alternatives to animal experiments based on the Adverse outcome pathway (AOP) have already been reported, and the Integrated Approach on Testing and Assessment (IATA) has been developed. We have reported a model for sensitization risk assessment using Artificial Neural Network (ANN). The ANN model reported in this paper was constructed using the paid software QwikNet. In this study, we developed a new sensitization risk prediction model for substances with known structures using R, a free software for statistical analysis.
Using the results of DPRA, KeratinoSens™, h-CLAT, and toxicity alerts by ToxTree as descriptors, a new sensitization prediction model using ANNs was constructed using R, and compared with the previous ANN model. In this report, we also conducted a case study using several cosmetic materials such as fragrances.
The constructed new model was found to be as accurate as the existing model. In the case study using samples not included in the training set, the model showed generally the same predictability, but deviations from the actual measured values were confirmed for some materials. For these substances, additional evaluation by Read Across suggested the possibility of complementing the predictions of the ANN model. The present results demonstrate the reproducibility of the ANN model among software and may contribute to improving the versatility of the prediction model. In this report, an evaluation flow combining the above findings is also proposed.
