Latent semantic structure indexing (LaSSI) for defining chemical similarity

A novel method for computing chemical similarity from chemical substructure descriptors is described. This new method, called LaSSI, uses the singular value decomposition (SVD) of a chemical descriptor-molecule matrix to crea te a low-dimensional representation of the original descriptor space. Ranki ng molecules by similarity to a probe molecule in the reduced-dimensional s pace has several advantages over analogous ranking in the original descript or space: matching latent structures is more robust than matching discrete descriptors, choosing the number of singular values provides a rational way to vary the "fuzziness" of the search, and the reduction in the dimensiona lity of the chemical space increases-searching speed. LaSSI also allows the calculation of the similarity between two descriptors and between a descri ptor and a molecule.