Abstract

QSAR paradigm depends on the application of the similarity principle that states that structurally similar compounds have similar physicochemical and biological properties. Similarity assessment is, therefore, critical to define pertinence of application of a QSAR model, designing adequate datasets for modeling purposes and in virtual screening of new potential active compounds at the beginning of a drug research and development campaign, when few substrates of a certain molecular target are known. This chapter will review recent studies related to similarity coefficients and different fingerprinting systems and feature weighting schemes used in similarity assessment. Applications of these advances in chemoinformatics are also briefly discussed, including similarity-based virtual screening, assessment of chemical libraries diversity and applicability domain estimation based on similarity measures.