Abstract
Developing new medical drugs is expensive. Among the first steps is a screening process, in which molecules in existing chemical libraries are tested for activity against a given target. This requires a lot of resources and manpower. Therefore it has become common to perform a virtual screening, where computers are used for predicting the activity of very large libraries of molecules, to identify the most promising leads for further laboratory experiments. Since computer simulations generally require fewer resources than physical experimentation this can lower the cost of medical and biological research significantly. In this paper we review practically fast algorithms for screening databases of molecules in order to find molecules that are sufficiently similar to a query molecule.