2.3. Identification of Pharmacophore Hypotheses

Pharmacophore models were built using Pharmacophore Alignment and Scoring Engine (PHASE) running on Maestro 10.4 (Schrödinger). The best docking pose was considered to be a bioactive conformer of each conformer and used for the development of the pharmacophore model. The activities of compounds were scaled from a minimum value of 1.01 to a maximum value of 3.93 (to be consistent with ones in the table), with an activity threshold of 1.35; this meant that compounds with a Ki of <44 nM were considered to be antagonists (actives). The pharmacophore features used for hypothesis generation were hydrogen bond acceptor (A), hydrogen bond donor (D), hydrophobic group (H), positively ionisable (P), negatively ionisable (N), and aromatic rings (R) defined by a set of chemical structure patterns. For the current dataset of 125 conformers, five or six features were chosen for model construction. The pharmacophore feature of active ligands that contain identical sets of features with very similar spatial arrangements were grouped together to give rise to a common pharmacophore hypothesis. In the present study, pharmacophore-based QSAR modeling was conducted by dividing the dataset into a 35-member training set (70%) and a 15-member test set (30%) in a random manner. The chemical features associated with ligand-protein binding and the training set data are used to create a model, and the test set is used to evaluate this model [19].