University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Graduation Talks


Diversity Selection for Virtual High Throughput Screening


Thorsten Meinl, University Konstanz
Konstanz, Germany

date & place

Wednesday, 17.12.2008, 16:15 h
Room C 252


High Throughput Screening (HTS) is widely used in pharmaceutical companies in the early drug discovery process. The activity of several hundreds of thousands molecules against a certain disease is tested in automated process. Nevertheless this process is time consuming and quite expensive. Virtual High Throughput Screening tries to support this process with computer-based approaches. One particular problem is the selection of molecules from a huge virtual compound library that shall be bought and tested in the next HTS. For this, models need to be built that try to predict the activity of the molecules, so that only the most active molecules can be bought. Another constraint, however, is to also select diverse molecules so that chemists have many potential starting points for further optimization of the molecules. These two objectives - selecting highly active molecules but still a very diverse set of compounds - are contradictory. In my thesis I am trying to solve this multiobjective optimization problem. This includes a proper theoretical formulation and several approaches ranging from empirical approaches, over classical genetic algorithm, branch-and-bound searches to highly problem specific algorithms.