Toxicity prediction and ecological risk assessment of new contaminants to rare and endangered species using machine learning-QSAR: a case study of conserving <i>Gobiocypris rarus</i> in the Yangtze River Basin

Figures (4) Tables (1)

Figure 1.
Overview of the modeling data set. (a) Pollutant classifications in acute and chronic toxicity data sets. (b) Boxplot of acute toxicity for different pollutants. (c) Boxplot of chronic toxicity for different pollutants.
Figure 2.
Acute ML-QSAR model performance of different algorithms: (a) RF, (b) GLM, (c) SVM, (d) KNN, (e) Neural Network,and (f) GLMNET. The data points for the training set are plotted as blue dots. The data points for the test set are plotted as red dots. Train R²represents R², and test R²represents R_ex². R² and R_ex² values are rounded to two decimal places.
Figure 3.
Chronic ML-QSAR model performance of different algorithms: (a) RF, (b) GLM, (c) SVM, (d) KNN, (e) Neural Network, and (f) GLMNET. The data points for the training set are plotted as blue dots. The data points for the test set are plotted as red dots. Train R²represents R², and test R²represents R_ex². R² and R_ex² values are rounded to two decimal places.
Figure 4.
Feature importance analysis of the model. Importance rank of different descriptors for ML-QSAR models: (a) acute, (b) chronic, and (c) descriptor explanations for CIC3, ATSC1P, ATSC2v, and ATSC2i.

Table 1.

Acute and chronic ecological risk entropy associated with 12 pollutants impacting G. rarus in different life stages in its habitat