Development of a Genetic Algorithm Method Especially Designed for the Comparison of Molecular Models: Application to the Elucidation of the Benzodiazepine Receptor Pharmacophore

Since the three-dimensional molecular structure of the benzodiazepine receptors is not yet univoquely known, the direct elucidation of the interaction mode between their active binding sites and their potent ligands is rather difficult. The comparison of selected ligands is thus an indirect approach which could help to determine the pharmacophore elements. In the present work, ligands for the benzodiazepine receptors are characterized using electron density maps at medium resolution, reconstructed from calculated structure factors using crystallography simulation programs. As the obtained threedimensional maps are rather complex, they then can be simplified by a topological analysis in order to represent the ligands as connected graphs. An original genetic algorithm method is finally elaborated and implemented to carry out graph comparison. The design of the algorithm implies appropriate and efficient coding and evaluation of the generated graph superimpositions. The major aim of this study consists in determining the nature and arrangement of the molecular fragments taking part in the binding of ligands to their benzodiazepine receptor sites.