The application of NMR-pattern-recognition methods to the classification of reduced, peracetylated oligosaccharide residues.

In the present paper homo- and hetero-nuclear correlation spectroscopies have been used to assign proton and carbonyl carbon resonances of a number of reduced, peracetylated mono- and oligo-saccharide derivatives. Each of the native structures for which assignments were made represent residues or substructures typically found in N- or O-linked glycans. Using the assigned NMR parameters as a basis, residues contained in parent structures were classified according to their residue type and glycosidic substitution sites using a relatively simple K-Nearest Neighbor pattern recognition approach. The method was able to correctly assign 99% of 77 "test residues" to their correct structural class using the full set of 19 assigned parameters as a basis. Similar correlations made between data and structure were less successful when reduced variable sets selected on the basis of SIMCA optimization were used.