Host: The Japanese Society of Toxicology
Name : The 49th Annual Meeting of the Japanese Society of Toxicology
Date : June 30, 2022 - July 02, 2022
[Introduction] In recent years, in silico methods have been used to predict and evaluate the toxicity of chemical substances. However, such studies on toxicity with complex mechanisms, including developmental neurotoxicity (DNT), have been limited. In this study, to establish a new read-across method to evaluate DNT using molecular descriptors, we have investigated methods to select source substances. [Methods] From the previous report (Neurotoxicol Teratol 52: 25-35, 2015), 164 substances with reported evidence for DNT and 197 substances with no relevant reports were selected as DNT-positive and DNT-negative substances, respectively. As a basic set, 351 molecular descriptors were calculated using alvaDesc (Alvascience) and standardized. Relative Euclidean distances between substances were calculated for selecting neighbor (source) substances. [Results and Discussion] Overall, good evaluation accuracy was obtained 1) when descriptors showing significant differences (P < 0.05) between positive and negative substances were used, 2) when the threshold values of relative distances for selecting neighboring substances were set at 0.13 to 0.15, and 3) when the number of neighboring substances was set at 11, 13 or 15. These results suggest that DNT can be evaluated by a read-across method by selecting appropriate descriptors, setting an appropriate threshold value for judging neighbors, and by appropriately limiting the number of neighboring substances.
