PRGEN - PSEUDORECEPTOR MODELING USING RECEPTOR-MEDIATED LIGAND ALIGNMENT AND PHARMACOPHORE EQUILIBRATION

Based on the structures of known ligand molecules, a pseudoreceptor mo deling concept developed at our laboratory allows the construction of a peptidic binding-site model for a structurally uncharacterized biore gulator. Such a three-dimensional receptor surrogate - validated using an external set of test compounds - should be able to semi-quantitati vely predict the binding affinities of related molecules. To reduce pr oblems resulting from the mutual obscuring of functional groups within a pharmacophore hypothesis, we have devised a procedure referred to a s receptor-mediated ligand alignment. It permits to identify an altern ate position, orientation and conformation for each ligand molecule by means of conformational search within a primordial receptor model, co nstructed about the most potent ligands of a series. To derive an ener getically relaxed model with a high correlation between calculated and experimental Gee energies of ligand binding, we have developed a liga nd equilibration protocol. During this iterative procedure, the recept or surrogate and the pharmacophore are allowed to relax individually, with and without correlation-coupled energy minimization, respectively , until a high correlation is achieved in a relaxed state. In our appr oach (software PrGen), free energies of ligand binding are estimated b ased on ligand-pseudoreceptor interactions using a directional force f ield, ligand desolvation energy and the change of both ligand-internal energy and ligand entropy upon receptor binding. The concept was test ed by generating and evaluating a pseudoreceptor for the cannabinoid r eceptor. The binding-site surrogate for this system was constructed ab out a pharmacophore comprising 18 cannabinoid antagonists. It consists of 26 amino-acid residues and is capable of predicting free energies of ligand binding, Delta G degrees, for an external set of 10 test mol ecules to within 0.8 kcal/mol (RMS) of their experimental value, corre sponding to an uncertainty factor of 4 in the binding affinity. Maxima l individual errors of predicted binding affinities do not exceed a fa ctor of 12.