Receptor and ligand-based 3D-QSAR study on a series of nonsteroidal anti-inflammatory drugs

Self-organizing molecular field analysis (SOMFA) has been used to study the correlation between the molecular properties of a series of 33 stilbene analogs and the cyclooxygenase-2 (COX-2) inhibitory activities. Thus, two different alignments and four charge-assigning methods using grid spacing of 1 Å were investigated to produce eight SOMFA models. The best one derived from the superposition of docked conformation with PM3 charge showed satisfied statistical results. It is characterized by a good noncross-validated r ² (0.806), cross-validated $ r_{text{cv}}^{ 2} $ (0.799) and F test value (128.673). In addition, the resulted SOMFA model was validated by an external set of ten compounds. The statistical results, predicted correlation coefficient $ r_{text{Test}}^{ 2} $  = 0.651, metric $ r_{{m({text{Test}})}}^{ 2} $  = 0.514, and the squared correlation coefficient between observed and predicted values r ² = 0.762, show a satisfied predictive ability. The analysis of SOMFA model, through electrostatic and shape grids, helped in understanding the possible structural modifications of molecules to improve the inhibitory potency.