Chapter 2. Analysis of Chemical Space The concept o f chemical space is discussed in this Chapter. The selection o f relevant
chemical space is arguably the most critical task in computer-aided molecular design. Its im-
pact on virtual screening, chemical compound library design, structure-activity relationship, and molecular similarity calculation are discussed. The Chapter covers algorithms for classical
unsupervised projection methods such as principal component analysis, Sammon mapping; and further nonlinear methodologies such as encoder networks, and the self-organizing map. Numerous example applications are given, including the discrimination o f drug and non-drug
molecules, clustering o f lead molecules with a specific type o f biological activity, e.g., antidepressants, in combinatorial library design, and the characterization o f ligand binding sites in protein structures.