In the present study we have investigated whether pharmacophore models may account for the activity and selectivity of the known cyclooxygenase-2 (COX-2) selective inhibitors of the phenylsulfonyl tricyclic series, i.e., Celecoxib (1) and Rofecoxib (3), and whether transferring this structural information onto the frame of a nonsteroidal antiinflammatory drug (NSAID), known to tightly bind the enzyme active site, may be useful for designing novel COX-2 selective inhibitors. With this aim we have developed a pharmacophore based on the geometric disposition of chemical features in the most favorable conformation of the COX-2 selective inhibitors SC-558 (2; analogue of Celecoxib (1)) and Rofecoxib (3) and the more restrained compounds 4 (DFU) and 5. The pharmacophore model contains a sulfonyl S atom, an aromatic ring (ring plane A) with a fixed position of the normal to the plane, and an additional aromatic ring (ring plane B), both rings forming a dihedral angle of 290 degrees +/- 10 degrees. The final disposition of the pharmacophoric groups parallels the geometry of the ligand SC-558 (2) in the known crystal structure of the COX-2 complex. Moreover, the nonconserved residue 523 is known to be important for COX-2 selective inhibition; thus, the crystallographic information was used to position an excluded volume in the pharmacophore, accounting for the space limits imposed by this nonconserved residue. The geometry of the final five-feature pharmacophore was found to be consistent with the crystal structure of the nonselective NSAID indomethacin (6) in the COX-2 complex. This result was used to design indomethacin analogues 8 and 9 that exhibited consistent structure-activity relationships leading to the potent and selective COX-2 inhibitor 8a. Compound 8a (LM-1685) was selected as a promising candidate for further pharmacological evaluation.