A Peptidic Binding Site Model for PDE 4 Inhibitors

Abstract

The pseudoreceptor modelling program PrGen was used to construct a peptidic binding site model for phosphodiesterase 4 inhibitors. A training set of 21 diverse compounds (rolipram, nitraquazone and xanthine derivatives, imidazo pyrido pyrazinones and 5-oxyindoles) was used to construct the binding site surrogate consisting of five amino acid residues, a Zn+2 cofactor and an envelope of charged virtual particles. The model was validated by predicting the free energies of binding ΔG0pred of ten ligands (rolipram, imidazo pyrido pyrazinones and 5-oxyindoles). In seven cases the prediction was satisfactory. The rms deviation [4] in ΔG0 is 0.16 and 1.82 kcal/mol – resulting in an uncertainty in IC50 (or Ki) of 1.32 and 22.81 – for the training and the test set respectively, while the corresponding maximal prediction errors in ΔG0pred were 0.27 kcal/mol and 4.50 kcal/mol.