We developed models for predicting fish early-life stage (ELS) toxicities oriented to industrial chemicals. The training set was constructed without data from the Office of Pesticide Programs Pesticide Ecotoxicity Database, the main source for the pesticide-biased training set used in our previous work (SAR QSAR Environ. Res. 29:9, 725–742). In addition to the descriptors from the previous study, we also used water solubility to develop the new models, which were evaluated against the test set used in our previous study so that we could focus on the effects of the different training set and the additional descriptor. The statistics for the new models were hardly better than those for the previous models, which suggests, contrary to our expectations, that pesticide-biased data can successfully be used to develop models for predicting the fish ELS toxicities oriented to industrial chemicals. Acute Daphnia magna toxicity was important for the predictive QSAARs in both studies. A distance-based method for defining the applicability domains indicated that water solubility was a key indicator for detecting underestimated chemicals. The comparison of fish ELS toxicities for chemicals presented in different literatures revealed the uncertainty of the experimental data, which may lead to the low predictivity.

We developed models for predicting fish early-life stage (ELS) toxicities oriented to industrial chemicals. The training set was constructed without data from the Office of Pesticide Programs Pesticide Ecotoxicity Database, the main source for the pesticide-biased training set used in our previous work (SAR QSAR Environ. Res. 29:9, 725–742). In addition to the descriptors from the previous study, we also used water solubility to develop the new models, which were evaluated against the test set used in our previous study so that we could focus on the effects of the different training set and the additional descriptor. The statistics for the new models were hardly better than those for the previous models, which suggests, contrary to our expectations, that pesticide-biased data can successfully be used to develop models for predicting the fish ELS toxicities oriented to industrial chemicals. Acute Daphnia magna toxicity was important for the predictive QSAARs in both studies. A distance-based method for defining the applicability domains indicated that water solubility was a key indicator for detecting underestimated chemicals. The comparison of fish ELS toxicities for chemicals presented in different literatures revealed the uncertainty of the experimental data, which may lead to the low predictivity.

We constructed models for acute to chronic estimation of the Daphnia magna reproductive toxicities of chemical substances from their Daphnia magna acute immobilization toxicities. The models combined the acute toxicities with structural and physicochemical descriptors. We used multiregression analysis and selected the descriptors for the models by means of a genetic algorithm. Of the best 100 models (as indicated by the lack of fit score), 90% included the following descriptors: acute toxicity (i.e. an activity parameter), distribution coefficient (log D) and structural indicator variables that indicate the presence of –NH₂ attached to aromatic carbon and the presence of a chlorine atom. We compared the predictive abilities of five of these quantitative structure–activity–activity relationship (QSAAR) acute to chronic estimation models with the predictive ability of a simple linear regression model. The comparison revealed that inclusion of structural and physicochemical descriptors such as those in QSAAR models can improve models for extrapolation from acute to chronic toxicity. Our results also provide a QSAAR framework that is expected to be useful for the further development of chronic toxicity estimation models.

We constructed models for acute to chronic estimation of the Daphnia magna reproductive toxicities of chemical substances from their Daphnia magna acute immobilization toxicities. The models combined the acute toxicities with structural and physicochemical descriptors. We used multiregression analysis and selected the descriptors for the models by means of a genetic algorithm. Of the best 100 models (as indicated by the lack of fit score), 90% included the following descriptors: acute toxicity (i.e. an activity parameter), distribution coefficient (log D) and structural indicator variables that indicate the presence of –NH₂ attached to aromatic carbon and the presence of a chlorine atom. We compared the predictive abilities of five of these quantitative structure–activity–activity relationship (QSAAR) acute to chronic estimation models with the predictive ability of a simple linear regression model. The comparison revealed that inclusion of structural and physicochemical descriptors such as those in QSAAR models can improve models for extrapolation from acute to chronic toxicity. Our results also provide a QSAAR framework that is expected to be useful for the further development of chronic toxicity estimation models.

We propose interspecies quantitative structure–activity–activity relationships (QSAARs), that is, QSARs with descriptors, to estimate species-specific acute aquatic toxicity. Using training datasets consisting of more than 100 aromatic amines and phenols, we found that the descriptors that predicted acute toxicities to fish (Oryzias latipes) and algae were daphnia toxicity, molecular weight (an indicator of molecular size and uptake) and selected indicator variables that discriminated between the absence or presence of various substructures. Molecular weight and the selected indicator variables improved the goodness-of-fit of the fish and algae toxicity prediction models. External validations of the QSAARs proved that algae toxicity could be predicted within 1.0 log unit and revealed structural profiles of outlier chemicals with respect to fish toxicity. In addition, applicability domains based on leverage values provided structural alerts for the predicted fish toxicity of chemicals with more than one hydroxyl or amino group attached to an aromatic ring, but not for fluoroanilines, which were not included in the training dataset. Although these simple QSAARs have limitations, their applicability is defined so clearly that they may be practical for screening chemicals with molecular weights of ≤364.9.

We propose interspecies quantitative structure–activity–activity relationships (QSAARs), that is, QSARs with descriptors, to estimate species-specific acute aquatic toxicity. Using training datasets consisting of more than 100 aromatic amines and phenols, we found that the descriptors that predicted acute toxicities to fish (Oryzias latipes) and algae were daphnia toxicity, molecular weight (an indicator of molecular size and uptake) and selected indicator variables that discriminated between the absence or presence of various substructures. Molecular weight and the selected indicator variables improved the goodness-of-fit of the fish and algae toxicity prediction models. External validations of the QSAARs proved that algae toxicity could be predicted within 1.0 log unit and revealed structural profiles of outlier chemicals with respect to fish toxicity. In addition, applicability domains based on leverage values provided structural alerts for the predicted fish toxicity of chemicals with more than one hydroxyl or amino group attached to an aromatic ring, but not for fluoroanilines, which were not included in the training dataset. Although these simple QSAARs have limitations, their applicability is defined so clearly that they may be practical for screening chemicals with molecular weights of ≤364.9.

We propose interspecies quantitative structure–activity–activity relationships (QSAARs), that is, QSARs with descriptors, to estimate species-specific acute aquatic toxicity. Using training datasets consisting of more than 100 aromatic amines and phenols, we found that the descriptors that predicted acute toxicities to fish (Oryzias latipes) and algae were daphnia toxicity, molecular weight (an indicator of molecular size and uptake) and selected indicator variables that discriminated between the absence or presence of various substructures. Molecular weight and the selected indicator variables improved the goodness-of-fit of the fish and algae toxicity prediction models. External validations of the QSAARs proved that algae toxicity could be predicted within 1.0 log unit and revealed structural profiles of outlier chemicals with respect to fish toxicity. In addition, applicability domains based on leverage values provided structural alerts for the predicted fish toxicity of chemicals with more than one hydroxyl or amino group attached to an aromatic ring, but not for fluoroanilines, which were not included in the training dataset. Although these simple QSAARs have limitations, their applicability is defined so clearly that they may be practical for screening chemicals with molecular weights of ≤364.9.