Improved identification of metabolites in complex mixtures using HSQC NMR spectroscopy

The automated and robust identification of metabolites in a complex biological sample remains one of the greatest challenges in metabolomics. In our experiments, HSQC carbon–proton correlation NMR data with a model that takes intensity information into account improves upon the identification of metabolites that was achieved using COSY proton–proton correlation NMR data with the binary model of [Y. Xi, J.S. de Ropp, M.R. Viant, D.L. Woodruff, P. Yu, Metabolomics, 2 (2006) 221–233]. In addition, using intensity information results in easier-to-interpret “grey areas” for cases where it is not clear if the compound might be present. We report on highly successful experiments that identify compounds in chemically defined mixtures as well as in biological samples, and compare our two-dimensional HSQC analyses against quantification of metabolites in the corresponding one-dimensional proton NMR spectra. We show that our approach successfully employs a fully automated algorithm for identifying the presence or absence of predefined compounds (held within a library) in biological HSQC spectra, and in addition calculates upper bounds on the compound intensities.