Multidimensional Chromatographic Fingerprinting Combined with Chemometrics for the Identification of Regulated Plants in Suspicious Plant Food Supplements

The popularity of plant food supplements has seen explosive growth all over the world, making them susceptible to adulteration and fraud. This necessitates a screening approach for the detection of regulated plants in plant food supplements, which are usually composed of complex plant mixtures, thus making the approach not so straightforward. This paper aims to tackle this problem by developing a multidimensional chromatographic fingerprinting method aided by chemometrics. To render more specificity to the chromatogram, a multidimensional fingerprint (absorbance x wavelength x retention time) was considered. This was achieved by selecting several wavelengths through a correlation analysis. The data were recorded using ultra-high-performance liquid chromatography (UHPLC) coupled with diode array detection (DAD). Chemometric modelling was performed by partial least squares-discriminant analysis (PLS-DA) through (a) binary modelling and (b) multiclass modelling. The correct classification rates (ccr%) by cross-validation, modelling, and external test set validation were satisfactory for both approaches, but upon further comparison, binary models were preferred. As a proof of concept, the models were applied to twelve samples for the detection of four regulated plants. Overall, it was revealed that the combination of multidimensional fingerprinting data with chemometrics was feasible for the identification of regulated plants in complex botanical matrices.

Automated summary

The popularity of plant food supplements has led to increased adulteration and fraud, making it necessary to develop a screening approach for the detection of regulated plants in these supplements. This paper proposes a multidimensional chromatographic fingerprinting method aided by chemometrics to address this problem. The method combines ultra-high-performance liquid chromatography (UHPLC) with diode array detection (DAD) and utilizes partial least squares-discriminant analysis (PLS-DA) for chemometric modelling. The results show that this approach is feasible for identifying regulated plants in complex botanical matrices.