In order to elucidate the essential structural features for Epidermal Growth Factor Receptor (EGFR) inhibitors, a ligand-based pharmacophore hypothesis was built on the basis of a set of twelve known EGFR inhibitors belonging to three different classes using Molecular Operating Environment (MOE) software. In a first step, three alignments, one for each group of compounds were generated. All of them were then submitted to MOE pharmacophore search in order to obtain a final pharmacophore model representative of the whole dataset. A pharmacophore model including three features was developed comprising one hydrogen-donor (F1) and two aromatic/ hydrophobic/ acceptor features (F2 and F3). The developed model was used to predict the activities of test set compounds by applying linear regression variable selection analysis. The model exhibited excellent linearity with correlation coefficient (r) value, i.e., 0.943, and squared predictive correlation coefficient (r2) of 0.889 between experimental and predicted activity values of test set compounds. Our model demonstrated good performance in a separate test set of 25 compounds: it accurately identified 67.7% of the compounds of medium and high inhibitory activities and misclassified only 28.5% of the compounds with low inhibitory activities. The results proved our pharmacophore model to be a filter of great sensitivity and specificity.
