Leveraging New Approach Methodologies: Ecotoxicological Modelling of Endocrine Disrupting Chemicals to Danio rerio through Machine Learning and Toxicity Studies

Emerging endocrine-disrupting chemicals (EDCs) are a diverse group of toxic substances that disrupt the endocrine system. These substances can only be approved after obtaining concrete evidence of their toxicity. New approach methodologies (NAMs) offer information tailored to the intended application while reducing the use of animals. NAMs aim to develop quantitative structure-activity relationship (QSAR) and quantitive-Read-Across structure-activity relationship (q-RASAR) models to predict and categorize the acute toxicity of known and unknown EDCs. The q-RASAR model was constructed and verified using validation metrics (R2 = 0.886 and Q2 = 0.814). The substructure fingerprint was well-fitted for the classification model and the model was validated using 10-fold average accuracy (Q = 86.88%), specificity (Sp = 88.89%), Matthew's correlation curve (MCC = 0.621) and receiver operating characteristics (ROC = 0.828). The dataset of unknown substances revealed that phenolphthalein (Php) exhibited a significant level of toxicity. The docking and simulation study results indicated that the computationally derived important features successfully bound to the target zebrafish sex hormone binding globulin (zfSHBG). The experimental LC50 value of 0.790 mg L−1 was very close to the predicted value of 0.763 mg L−1, which provides high confidence to the developed model.