RESUMO
BACKGROUND: A variety of chemotypes have been studied as estrogen receptor (ER) ß-selective ligands for potential drugs against various indications, including neurodegenerative diseases. Their structure--activity relationship data and the x-ray structures of the ERß ligand-binding domain bound with different ligands have become available. Thus, it is vitally important for future development of ERß-selective ligands that robust quantitative structure-activity relationship (QSAR) models be built. METHODS/RESULTS: We employed a newly developed structure--based QSAR method (structure-based pharmacophore keys QSAR) that utilizes both the structure--activity relationship data and the 3D structural information of ERß, as well as a robust QSAR workflow to analyze 37 ligands. Four sets of QSAR models were obtained, among which approximately 30 models afforded high (>0.60) training-r(2) and test set-R(2) statistics. CONCLUSION: We have obtained an ensemble of predictive models of ERß ligands that will be useful in the future discovery of novel ERß-selective molecules.