APPLICATION OF THE MAXIMAL COMMON SUBSTRUCTURE ALGORITHM TO AUTOMATIC INTERPRETATION OF C-13-NMR SPECTRA

A novel program has been developed for deducing automatically the most common structural fragments from a set of similar structures by means of the maximal common substructure (MCSS) algorithm. These reference structures are obtained by using spectral search techniques on our NMR spectral database containing about 90 000 C-13-NMR spectra, with the C-13-NMR spectrum of the compound under investigation as the query spectrum. The molecular formula of the target structure can be further used to reduce the number of structures within the final hit list to about 100 so that they can be processed in a few minutes of CPU time on a graphics workstation. In order to improve the results, the structures are preprocessed by removing those carbon atoms which are outside a user-defined range (usually 2-4 ppm) of each query shift value. The investigations shows that the MCSSs thus obtained usually contain main structural features of the target compound and can be directly used as structural constraints in the structure generation process.