Similarity-searching and clustering algorithms for processing databases of two-dimensional and three-dimensional chemical structures

Abstract

Databases of chemical structures play an increasingly important role in the fine-chemicals industry, e.g. for the development of novel pharamaceuticals and agrochemicals (Ash et al., 1991). These databases contain tens or hundreds of thousands of chemical substances, either in two-dimensional (2D) or in three-dimensional (3D) form, and several different searching mechanisms have been developed to provide access to the data that is stored in them. The most common mechanisms are structure searching, which involves the retrieval of a single specific molecule, and substructure searching, which involves the retrieval of all of those molecules that contain a user-defined partial structure, e.g. a putative pharmacophore pattern. An extended programme of research in the University of Sheffield has sought to develop a complementary means of access, called similarity searching, and this chapter provides an overview of some of the algorithms that have been developed for this purpose since the programme commenced in the early 1980s. Specifically, we are interested in techniques that will allow a user of a chemical database to input a target structure of interest, and then to retrieve those molecules in the database that are structurally most similar to the target molecule. Our programme of research has also considered how cluster-analysis methods can be used for the processing of chemical databases.