New Molecular Screening Approach for Discovery of Novel CycloOxygenase-2 Inhibitors through Multi Linear Regression Models

Background: Inflammation is basically caused with conversion of Arachidonic acid into Prostaglandin H2 by CycloOxygenase. In this study a new algorithmic procedure is applied in order to screen molecules not only with high affinity to COX-2, but also different from their ancestor compounds. Methods: NSAIDs, COX-1 and COX-2 molecules were downloaded from Drug Bank and Protein Data Bank. Drugs were docked with both proteins by FlexX software. Top 10 molecules with lowest COX-2 interaction energies and highest differences between COX-2 and COX-1 IEs were selected for structural similarity searches in PUBCHEM and ENCANCED NCI databases. Second generation molecules were docked with proteins once again. Compounds with lower IEs than parents, were collected. Bioactivities and bioavailabilities of compounds were analyzed by PASS software and Lipinski rules. A best multi linear regression model was developed based on some physicochemical descriptors for further studies. Results: 50 NSAIDs were selected and 2000 similar molecules gathered. Screening the molecules based on Lipinski rules, bioactivities and drug likenesses, a trustable BMLR model was developed with more than 80% accuracy including following descriptors: Log P, Log D, Molar Refractivity, Polarity Number, and Aromaticity Ratio. Finally, 6 compounds were selected as best structurally new compounds for further in vitro analysis. Conclusions: Final molecules having high druglikeness and affinity and structurally different from their ancestors, can be used in order to develop new lead compounds with higher selectivity.