THE INTEGRATION OF STRUCTURE-BASED DRUG DESIGN & COMBINATORIAL CHEMISTRY FOR EFFICIENT DRUG DISCOVERY. F. Raymond Salemme, 3-Dimensional Pharmaceuticals, Inc., Eagleview Corporate Center, 665 Stockton Drive, Suite 104 Exton, PA 19341

Structure-Based drug design allows the atom by atom modification of drugs leads whose binding to a receptor target can be directly visualized using x-ray crystallography. Although structure-based design has become increasingly widely used in pharmaceutical discovery owing to improved technology for rapid 3D structure determination of drug-ligand complexes, key technical issues have emerged which limit process efficiency. Specifically, the picture provided by the protein-ligand 3D structure tells the "how" but not the energetic "why" of ligand binding. This situation basically reflects the physical complexity that underlies the change in free energy accompanying ligand binding, which involves a multiplicity of factors including changes in ligand bonding [with both solvent water and the target protein], changes in ligand conformation or flexibility, changes in ligand polarization, as well as corresponding changes in the target protein. The accurate estimation of how these changes will occur and manifest themselves in the binding energetics for a given ligand is beyond the capabilities of currently available computational methods, although some success is possible in estimating the effects of small structural perturbations. Studies of complexes formed between streptavidin and a variety of ligands, where high resolution x-ray crystallographic results were complemented with detailed thermodynamics measurements of ?H and ?S, provide illuminating examples in this context. Still, from a practical standpoint, investigators have typically found it necessary to carry out the iterative synthesis and structural analysis of an extensive series of compounds to empirically define the important aspects of ligand binding energetics and refine the properties of the target ligand.

An alternative to the serial approach to structure-property refinement involves the development of methods for parallel synthesis of compounds that meet specific geometric requirements of a target receptor binding site. Custom chemical scaffolds can be designed that are directed to fit a particular receptor binding site and that can be synthetically elaborated through combinatorial reaction with commercially available reagents. Although such targeted combinatorial schemes can produce libraries of drug-like compounds with thousands to millions of members, parallel automated synthesis methods are presently capable of synthesizing libraries containing on the order of a hundred discrete compounds. This practical limitation on synthetic throughput motivates the development of effective computer search strategies that can iteratively define and refine the selection of sub-libraries that will best investigate the structure-property relationships for a given library-target combination. This has been achieved through the computer generation of large "virtual" libraries of synthetically accessible compounds which are designed to explore specific features suggested from a 3D structural model and/or other SAR features thought to be important in the ultimate development of a successful drug [e.g. features relating to bioavailability, toxicology, etc.]. Computer codes are then used to select library sub-sets for rounds of automated chemical synthesis and bioassaytesting. Test data resulting after each round are used both as selection criteria for additional 3D target-ligand structure determinations, and to iteratively refine molecular properties using more traditional SAR methods. In many cases, extensive 3D structural data generated from a chemically "easy" combinatorial library can be combined with elements of pure structure-based designs to rapidly develop novel compounds. This integrated approach will be described in the context of developing and refining libraries of potent and specific protease inhibitors.

Ref: Crystallographic and Thermodynamic Comparison of Natural and Synthetic Ligands Bound to Streptavidin, P.C. Weber, J.J. Wendoloski, M.W. Pantoliano, and F.R. Salemme (1992) J. Amer. Chem. Soc., 114:3197-3200