Algorithms for Computer-Aided Drug Design

problems of our time. Besides the identification of new,
appropriate drug targets, the design of a drug molecule is
a central step within the development process. Computer methods
play an increasing role in drug design, on the one hand because
more and more relevant data on molecules and their interactions
become available, on the other hand because computer predictions
become a reasonable alternative to experiments in early stages of
the drug design process.

In this talk, I will give a survey of models and algorithms for
selected prediction problems from the research area of drug design.
The 'Protein-Ligand Docking Problem' is the key problem of structure-
based drug design. Given the 3D-structure of a protein of interest and
a small potential drug molecule, the task is to predict the relative
orientation (molecule complex) as well as the binding affinity of
the complex.
Since 3D-structures of proteins are frequently not available, methods
predicting molecular properties based on the similarity of the molecule
to other molecules with known properties is a central concept in drug
design (the similarity principle). In order to follow this approach,
biologically relevant models and algorithms for comparing molecules have
to be developed.

After a brief introduction of algorithms developed for these problems,
their performance and limits will be demonstrated on real examples from
the pharmaceutical research area.