Predicting Potential Endocrine Disrupting Chemicals Binding to Estrogen Receptor α (ERα) Using a Pipeline Combining Structure-Based and Ligand-Based in Silico Methods

The estrogen receptors α (ERα) are transcription factors involved in several physiological processes belonging to the nuclear receptors (NRs) protein family. Besides the endogenous ligands, several other chemicals are able to bind to those receptors. Among them are endocrine disrupting chemicals (EDCs) that can trigger toxicological pathways. Many studies have focused on predicting EDCs based on their ability to bind NRs; mainly, estrogen receptors (ER), thyroid hormones receptors (TR), androgen receptors (AR), glucocorticoid receptors (GR), and peroxisome proliferator-activated receptors gamma (PPARγ). In this work, we suggest a pipeline designed for the prediction of ERα binding activity. The flagged compounds can be further explored using experimental techniques to assess their potential to be EDCs. The pipeline is a combination of structure based (docking and pharmacophore models) and ligand based (pharmacophore models) methods. The models have been constructed using the Environmental Protection Agency (EPA) data encompassing a large number of structurally diverse compounds. A validation step was then achieved using two external databases: the NR-DBIND (Nuclear Receptors DataBase Including Negative Data) and the EADB (Estrogenic Activity DataBase). Different combination protocols were explored. Results showed that the combination of models performed better than each model taken individually. The consensus protocol that reached values of 0.81 and 0.54 for sensitivity and specificity, respectively, was the best suited for our toxicological study. Insights and recommendations were drawn to alleviate the screening quality of other projects focusing on ERα binding predictions.