HR-1H NMR spectroscopy and multivariate statistical analysis to determine the composition of herbal mixtures for infusions

Introduction The ever-growing diffusion and consumption of herbal teas, due to their sensory attributes and well-known health benefits exposes them to the real risk of adulteration, especially in the case of commercial mixtures already minced for infusion. Therefore, novel and suitable tools for the control of these valuable products are increasingly required. Objectives This work provides new insights for the authenticity study of infusions. The main objective was verifying the potential of proton nuclear magnetic resonance (H-1-NMR) combined with partial least square (PLS) regression to build highly predictive models, useful for the determination of the real amounts of herbs in mixtures, by the simple analysis of the related infusion. Materials and methods Peppermint, fennel, lemon balm, and passiflora were chosen to set-up an experimental plan according to a central composite design (CCD). One-dimensional nuclear Overhauser effect spectroscopy (1D-NOESY) spectra were properly pretreated and then analysed by chemometrics to extract significant information from the raw data. Results Venetian-blind cross-validation and different chemometric indicators (RMSEC, RMSECV, RMSEP,R-CAL(2),(RCV,RPRED2)-R-2) were used to establish the best model, which include four factors explaining 88.70 and 83.77% of the total variance inXandY, respectively. Conclusions These promising results have laid the basis for further development of the method, to extend its applicability and make it more scalable. This tool could replace expensive separative techniques and protect the rights of consumers with particular attention to safety issues and quality assurance.

Automated summary

This study utilized proton nuclear magnetic resonance combined with partial least square regression to establish highly predictive models for determining the actual amounts of herbs in herbal tea mixtures. By analyzing 1D-NOESY spectra using chemometrics, the best model was developed, laying the foundation for further development to enhance applicability and scalability.