Leadlikeness and structural diversity of synthetic screening libraries

High program failure rates in the pharmaceutical industry have prompted the development of predictive software that can profile compound libraries as being 'druglike' (resembling existing drugs) and/or 'leadlike' (possessing the structural and physicochemical profile of a quality lead). In recent years, these two notions prompted pharmaceutical companies to clean up their corporate libraries of screening compounds. In order to maintain and expand the size and diversity of these corporate libraries, pharmaceutical companies still continue to add compounds to these, mainly by the acquisition of screening libraries. In this paper, we have analyzed 45 commercially available libraries, offered by suppliers of screening chemistry, for leadlikeness and diversity of the offered structures. To this end we have used a set of structural and physicochemical filters for leadlikeness that was developed in-house. These 45 supplier libraries contained a total of 5.3 million structures, of which 49% (2,592,778 structures) turned out to be unique, and only 12% (677,328 structures) were found to be both unique and leadlike. A diversity analysis revealed that big differences exist between the various offered libraries.