Open column, reversed-phase high-performance liquid chromatography with diode array detection and chemometric strategy for investigation of metabolic fingerprints of complex systems

Abstract

A new strategy for investigation of metabolic fingerprints in complex systems was developed based on a combination of open column chromatography, reversed-phase high-performance liquid chromatography and k-means clustering. A simplex centroid mixture design with four solvents, hexane, dichloromethane, ethyl acetate and methanol, was applied to a crude extract of the Bauhinia variegata L. to gradually change the mobile phase polarity of the open column. K-means clustering was used to screen the profiles of the 3033 spectra from high-performance liquid chromatography with diode array detection data. The seventeen k-means clusters were discriminated into three groups. The largest one containing 2617 (86.3%) spectra has a strong absorbance band around 210 nm that is commonly found in many naturally abundant compounds. The smallest group contains only 134 (4.4%) spectra and has strong absorbance bands at wavelengths well above 250 nm and are potentially interesting for further metabolic studies and fingerprinting. The other group judged to have spectral profiles intermediate between those of the other two contains 282 (9.3%) spectra. Ternary and quaternary mixtures of the statistical design provided potentially useful information for fingerprinting Bauhinia variegata L. extracts relative to what could be obtained using only pure solvents and binary mixtures that are commonly employed in open column chromatography.