Quantitative Comparison of Five Molecular Structure Spaces in Selecting Analogs of Chemicals

Five methods for characterizing intermolecular similarity have been used in the selection of analogs for a diverse set of seventy-six compounds. These methods include an atom pair (AP) based similarity measure, three principal component spaces derived from topostructural indices, topochemical indices, the combined set of all (topostructural and topochemical) indices, as well as one structure space consisting of principal components calculated from physicochemical properties. Each method has been used in the selection of sets analogs, ranging from five to forty in number in increments of five, for each of the seventy-six compounds. The degree of overlap of the sets of analogs selected by the five separate methods was analyzed.