2021 Volume 24 Pages 33-42
We examined two groups of quantitative structure–activity–activity relationship (QSAAR) models for predicting Japanese medaka (Oryzias latipes) early-life stage (ELS) toxicities of chemicals for the purpose of chronic hazard and environmental risk assessments. The models included not only typical molecular descriptors but also acute Daphnia magna toxicity data, ELS test conditions, and information about chemical categories (e.g., pesticides). We found that acute Daphnia magna toxicity was an important descriptor for predicting fish ELS toxicity, along with molecular descriptors. The group II models, which were based on 119 training data for three warm freshwater species (fathead minnow, Japanese medaka, and zebrafish) had higher predictivity than the group I models, which were based on a 172 training data for four freshwater species (fathead minnow, Japanese medaka, zebrafish, and rainbow trout). In addition, the group II models had higher predictivity than the QSAAR models we reported previously (SAR QSAR Environ. Res. 29 : 9, 725–742 and 30 : 11, 825–846). Models developed by means of the strategy used to develop the group II models would be usable for estimating chronic fish toxicity in screening assessments such as those required under the Japanese Chemical Substances Control Law.
