COMP 163 |
| Molecular libraries steadily increase in size. Thus, molecular docking methods have to cope with rapidly increasing amounts of compounds. We investigated three approaches to speed up our docking tool FlexX: i) eliminating undesired molecules by effective filters; ii) reducing computational costs by using a sparse interaction model; iii) avoiding re-dockings by caching partial solutions. All three approaches can speed up virtual screening by a factor of 3-5 or more depending on the dataset and the applied filters. The combination of all three methods can easily speed up docking by a factor of 10 without any loss in accuracy. Thus the performance approaches the region of docking 10 Mio compounds per day or nearly 500,000 molecules per hour on a 300 node cluster. Evaluation results will be presented on various data sets. |
|
Drug Discovery
1:00 PM-4:15 PM, Monday, August 20, 2007 BCEC -- 161, Oral
Division of Computers in Chemistry |