CLASSIFICATION OF LIMONOIDS AND PROTOLIMONOIDS USING NEURAL NETWORKS

A novel technique of classifying liminoids and protolimonoids using ar tificial neural networks is presented, The difficulties associated wit h natural product classification are discussed, as well as the relevan ce of artificial neural networks to the task of automated classificati on by computer, Data from the C-13 nuclear magnetic resonance spectra of the compounds is pre-processed using histogram binning, Neural netw orks are trained on this data correctly to classify the data as belong ing to a ''limonoid'', ''triterpenoid'' or ''other'' category and to d iscriminate the protolimonoids from the rest of the triterpenoids in t he ''triterpenoid'' data set, Finally, neural networks are trained to recognise each individual limonoid belonging to the ''limonoid'' data set, The accuracy of the neural networks is typically better than 90% on unseen or impure data. (C) 1997 John Wiley & Sons, Ltd.