A robust technique for the group classification of the C-13NMR spectra of natural products from Meliaceae

Neural networks have been found to be good general classifiers of the binne d C-13 NMR spectra of limonoids and associated extractives from the Meliace ae family. 5 ppm binning of the C-13 NMR spectra was found to be superior t o 10 ppm binning. The excellent results obtained by developing the system o n real extractives using the 5 ppm binning, revealed that the system develo ped was robust and general enough to tolerate the inherent problems associa ted with natural product extraction. Experiments concerning the effect of n oise and missing bins were also conducted and documented. A literature-sour ced sample set was also added in order to ensure the system was not specifi cally designed to cope only with the unique data set on which it had been t rained.