Application of evolving factor analysis to on-flow LC–NMR data using spectral windows

Abstract

A new method has been introduced for curve resolution of on-flow liquid chromatography coupled with nuclear magnetic resonance spectroscopy (LC–NMR). The method takes advantage of clustering of NMR resonances, which is a special feature of LC–NMR. It is implemented in four steps; the first step is determining the regions of NMR peak clusters, the second is the rank analysis of each peak cluster, the third is the assignment of peak clusters to different compounds and the fourth is the application of evolving factor analysis (EFA). The transformation matrix in the proposed method is calculated using compound-related spectral windows. Canonical correlation analysis (CCA) and the orthogonal projection approach (OPA) have been employed determining relationships between spectral peaks and compounds in the third step. It is shown that CCA does not work when chromatographic resolution is poor. The proposed method has been compared with ordinary least squares using all pure concentration profiles and with conventional EFA. The method is tested on two simulated and two real datasets and results are found promising for both types of datasets.