Resolution of on-flow liquid chromatography proton nuclear magnetic resonance using canonical correlation and constrained linear regression

A method is described for the resolution of Liquid Chromatography Nuclear Magnetic Resonance (LC-NMR) data. Two practical case studies are presented, of 3 and 4 isomers with overlapping resonances and embedded chromatographic peaks, all resonances analysed being in the aromatic region. The first steps consist of preprocessing. Evolving Principal Component Analysis (EPCA) is used in an exploratory manner to indicate how many analytes are present in the mixture. Canonical Correlation Analysis (CCA) is used to determine the number and relationship of analytes between spectroscopic peak clusters, followed by Constrained (non-negative) Least Squares. It is shown that this approach offers considerable improvements over Evolving Factor Analysis (EFA). Good reconstructions are obtained from the chromatograms.