Privilegierte Strukturen zur Inhibitionvon Serinproteasen – Docking und direktes Design in der Proteinbindetasche

Goal of this thesis was the development of tools for structure-based design of combinatorial libraries of putative inhibitors. Successful automated computational ligand design requires, however, rigorous validation with respect to the predicted binding geometries. As a starting point, different computational as well as experimental techniques were applied to assess the predictive power of docking tools in generating binding geometries for structure-based library design. Combining experimental data and computer-generated geometries provides solid ground for the design of an expanded series of inhibitors for serine proteases. The serine proteases Thrombin, Trypsin and FactorXa were used as well established model systems. In the second part of the thesis different methods for lead generation were developed and two standard techniques of virtual screening protocols were compared. The first method, KNOBLE (KNOwledge Based Ligand Enumeration), is a new strategy to assemble a focused combinatorial library in silico using a structure-based approach. Starting point for the design of such a focused combinatorial library is the selection of a central fragment equipped with appropriate linker groups allowing the introduction of various substituents, well suited to address the key interactions of the conserved recognition pattern of the respective target enzyme via routine synthetic chemistry. Putative building blocks for synthesis are selected and virtually assembled in the computer. The actual retrieval of suitable substituents and the selection of appropriate decoration patterns are guided by the exposed physicochemical properties of the target protein. As a case study to probe this new strategy, we selected the family of serine proteases. Subsequent to computational selection and optimization, some selected library members were synthesized and subjected to biological testing showing promising activity in the micromolar range. Recent improvements in both, software and hardware, principally allow the usage of docking methods in a high-throughput mode as a tool for virtual screening without prior application of sophisticated pharmacophore filters. Therefore, a direct comparison of the results either based on a UNITY pharmacophore search or with docking by means of FlexX performed within the environment of the Docking DataBase (DDB) was drawn. Finally, an approach for the generation of protein-derived pharmacophores was developed using the programs FTrees and DrugScoreCSD. Via the translation of computationally generated contour maps favorable binding regions are detected for multiple chemical probes and transformed into “HotSpot”-spheres. These were exploited to derive protein-based FTrees that served for subsequent searches of putative ligand candidates.