Similarity metrics and descriptor spaces - Which combinations to choose?

Molecular similarity is widely used in virtual screening. There are a very large number of possible combinations of molecular descriptors and analysis methods that can be combined to pre-select compounds. The objectives strongly influence the methods chosen, in particular whether the desired outcome is to design a diverse library for initial screening; to follow up with additional similar hits (to perhaps help in establishing SAR) or to discover novel scaffolds (lead hopping) with the objective of obtaining novel patentable series (perhaps with different pharmacokinetics). Some of the factors that influence these decisions are discussed along with applications that compare and contrast methods and their performance in different situations.