2.3. Validation of the Pharmacophore

Validation of the pharmacophore model is an essential step for its selection and evaluation. In the present study, two commonly used validation approaches, mainly, the receiver operating characteristic (ROC) curve and the Güner–Henry (GH) approach, were used [41,42]. The ROC curve analysis was performed during hypothesis generation in both ligand- and structure-based procedures. First, a small dataset was prepared with known active and inactive compounds. The four compounds used for pharmacophore generation were considered as known actives, and the other eight were taken as inactive. The top three hypotheses from each approach were selected and further validated with a second validation technique, the GH or decoy set method. A decoy set of 110 compounds was generated with 6 already known active inhibitors of CDK7 (IC₅₀ < 100 nm) [30,31] and 104 inactive compounds. The Ligand Pharmacophore Mapping module in DS was used to screen the decoy dataset. The resulting mapping data were used for assessment of the pharmacophore quality by evaluating the following equation:

The selected and validated hypotheses from the ligand- and structure-based pharmacophore procedures were exploited as 3D queries to screen four natural compound databases in DS.