Evaluation of the relationship between spectral and structural similarity in IR spectroscopy

Representation of a structural formula of a compound as a complete set of nonisomorphic k-vertex (2 ≤k≤7) connected fragments is used to evaluate the structural similarity of compounds. The spectra and structures of spectrally or structurally alike compounds are compared statistically using a database containing 32,000 IR spectra and structures. This study reveals some tendencies typical for both full IR spectra and their abbreviated versions represented by sets of most characteristic peaks. Applicability of IR databases to spectrum simulation for compounds with a specified structure is justified statistically. A method is proposed for evaluating the efficiency of a search algorithm selecting structural analogs of the compound from an IR database according to the query spectrum. Translated fromZhurnal Strukturnoi Khimii, Vol. 41, No. 2, pp. 379–390, March–April, 2000.