Design and analysis of large chemical databases for drug discovery

The drug discovery paradigm has changed in two important ways. The human genome project is giving us many more new biological targets for drug discovery. Hundreds of unknown disease genes are expected to turn up in the next few years. Combinatorial chemistry and the availability of commercial compounds have made millions of compounds available for drug screening. It is no longer possible to test all available compounds for every new target of potential biological importance.; In this thesis novel statistical methods for design and analysis of large chemical databases are described. The design problem is to choose a representative set of thousands of chemical compounds from a library that may have hundreds of thousands to millions of compounds, for assay against a biological target (screening). The analysis problem is to find regions of a high dimensional space where active compounds reside. These methods improve the efficiency and effectiveness of the drug discovery process for reducing drug screening costs and time.