Intrinsic Dimensionality of Chemical Structure Space

摘要

The paper is an attempt to define chemical structure space for a large universe of chemicals. The dimensionality and scaling of chemical structure space has been sought using thermodynamic properties and the biological activity of molecules. A set of 19,972 chemicals from registries of chemical production was selected, for which graph theoretic indices have been computed. The dimensionality of the data set was reduced to a set of 8 principal components. Computer programs for minicomputers were developed to display the 'universe' of chemical structures through user-selected windows.