Docking large, small molecule databases into target protein models has been used to support combinatorial library design by docking virtual libraries, to find novel lead structures by docking commercially available compounds, and to prioritize screening by docking corporate archives. Several algorithms are available for rapidly docking thousands of compounds and this talk will compare the performance of DockIt with that of FlexX and DgeomDock (a Chiron in-house developed distance geometry based program). In addition to examples of the screening of large databases, we will also show an example that involves the generation of binding mode hypotheses for selected molecules.
The fast but simplified general scoring functions currently available generate reasonable docking modes for a given molecule, but are less able to quantify the relative binding strength between molecules. We have developed tools to analyze the database docking output with respect to characteristics that we use when eyeballing the results: hydrogen bonds to the active site, occupancy of specific pockets in the binding site, and % of the ligand surface area buried by docking. The set of tools, called Magnet, generates descriptors for each docking which can be used to sort and select the more interesting docking results. An iterative process of defining important features, sorting the output, and reviewing the results, gradually pulls the most attractive dockings to the top of the list. Using the insights gained with respect to the binding site and the different binding modes that are observed, the Magnet columns can be used to derive a new, quantitative scoring function which is tailor-made with respect to the specific binding site and set of compounds under investigation. We have incorporated this magnetized scoring function into a genetic algorithm (GA) driven docking tool that is used in the generation of structure-biased combinatorial libraries. The fast general docking function finds the best few binding modes for each compound, and Magnet assigns the fitness score for comparison between molecules to drive the GA.