IBM and NeSC workshop on Protein Science
National e-Science Centre, Edinburgh, March
15-16 2002
Structure-Based and Computer-Aided Drug Design H. Kubinyi In the past decade, genomics, combinatorial chemistry, molecular modelling, and high-throughput screening developed as new strategies in drug research. Drug targets result from the human genome. The validation of a new target for therapy is performed in genetically modified animals. Ten thousands to hundred thousands compounds are synthesised, in parallel, and are tested in automated high-throughput screening systems. However, the success rate of these approaches is only modest. With the ongoing progress in protein crystallography, rational approaches in drug design become more and more important. Several examples of the structure-based design of marketed drugs and drug candidates being in clinical development illustrate the potential of this technique. Whereas structure-based design can be regarded as the predominant strategy of the last decade, computer-assisted drug research is a recent development for the de novo design of protein ligands. If thousands of candidates and even larger structural databases shall be tested whether certain members are suited to be ligands of a protein binding site, this cannot any longer be done by hand. The process has to be automated, i.e. performed with the help of the computer. Computer programs for this purpose are e.g. DOCK and LUDI. Compound collections of pharmaceutical companies and natural products are biased towards biologically active compounds. On the other hand, combinatorial libraries and compounds offered by commercial vendors need a proper selection of sublibraries and/or candidates that have the potential to be biologically active. "Drug-like~ properties, "lead-like" character, oral bioavailability, and sufficient metabolic stability are preconditions for valuable leads and are more important than the chemical accessibility of a library. An interesting approach to determine the ~drug-likeness~ of series of organic molecules, based on a neural net evaluation, has been developed at BASF. Other tools for filtering compound libraries, to enrich them with biologically active compounds, and the selection of "best" sublibraries from huge virtual libraries will be discussed. Computer methods, like the flexible docking program FlexX, and experimental methods, like the SAR by NMR method and the dynamic assembly of ligands within the binding site of a protein pave the way to a combinatorial design of protein ligands. Several research institutions and pharmaceutical companies are now going to develop dedicated computer programs for the structure-based design of ligands from selected building blocks, by combinatorial construction within their binding site. In this manner, out of a virtual library of billions of possible ligands only those shall be selected for synthesis and testing, which smoothly fit the binding site, in geometry as well as in their properties. Whereas all necessary tools for this purpose are already developed, a correct ranking of the intermediate solutions and the final results according to their estimated binding affinity is still very difficult. Reliable scoring functions for this purpose have to be developed. |
|
||||||||||
|
|||||||||||
