The global popularity of plant food supplements is widely recognised. As their consumption
; rises, these products become more susceptible to unethical practices such as adulteration and
; fraud. Chemical adulteration is a well-studied area, but herbal adulteration is often overlooked.
; This is concerning because herbal products can also encompass toxic properties. However,
; limited research has been conducted to investigate the side effects or toxic potential of plants,
; which becomes even more complicated when considering their presence in plant food
; supplements. The complexity arises from the nature of such supplements, which consist of
; plant mixtures, making their authentication a problem. This research addresses the issue of
; herbal adulteration in plant food supplements. The title of this thesis represents the 3-
; dimensional data obtained by fingerprinting techniques, to which chemometrical treatments are
; applied for the treatment of the data.
; To tackle thechallenge of herbal adulteration, this project explores the domain of
; chromatographic fingerprinting through two detection techniques: Diode Array Detection
; (DAD) and Mass Spectrometry (MS). Given the complexity of botanical mixtures, traditional
; fingerprinting methods, which rely on recognition of active compounds, may not be sufficient
; to distinguish between authentic and adulterated products. Therefore, to investigate the
; analytical data, a strategy was devised integrating chemometrics, a powerful mathematical tool
; used to extract meaningful information from complex chemical data. The aims governing this
; thesis comprise the application of chromatographic fingerprinting and chemometric strategies
; for DAD (Chapters 3 and 4) and LC-MS based fingerprints (Chapter 5). Further, the best
; approach was considered in a limited market study for weight loss supplements (Chapter 6).
; For DAD, a multiwavelength approach was explored, where instead of a single wavelength,
; orthogonal adjacent wavelengths were selected after correlation analysis. This was to add as
; much information as possible to the fingerprints. A comparison was made between results from
; a single-wavelength approach and our multi-wavelength approach to determine which method
; is superior. Moreover, an effort was made to derive a single common fingerprinting method for
; all the selected regulated plants within each indication group. To test and validate the process
; and generate data for chemometric processing, triturations in varying concentrations were
; prepared. This was achieved by selecting 10 blank botanical matrices that belonged to
; IV
; categories other than weight loss. Lactose was added to check the suitability of the method.
; The reference plants and these botanical matrices were mixed in varying concentrations ½,
; ⅕, 1/10, 1/15, ½0 and injected into the system along with the reference plants to obtain the
; data set. This raw data was then exported from Empower to Excel. A fingerprint region that
; represented the maximum information in the chromatogram, and the selected wavelengths were
; considered to construct the data cube. The dimensions were time points × wavelengths ×
; absorbance, resulting in a three-dimensional data set. Chemometric processing was then
; applied. Correlation-Optimised Warping (COW) was used to align the data, followed by
; different pre-processing techniques to improve the quality of the spectral data. For
; chemometric modeling, unsupervised and supervised techniques were considered.
; Unsupervised techniques are exploratory methods that help provide insights into the data, such
; as principal component analysis (PCA) and hierarchical clustering analysis (HCA). Supervised
; techniques, on the other hand, can be referred to as classification or discrimination techniques
; that group unknown data based on similarities/ dissimilarities. These techniques include partial
; least squares discriminant analysis (PLS-DA) and soft independent modeling of class analogy
; (SIMCA). The best techniques were selected from this approach, and classification was carried
; out. A short proof of concept was performed to validate the models in real life (chapters 3 and
; 4). For MS fingerprinting, the dimensions of the dataset were time points × intensity × mass
; data. The MS data were then also treated similarly to the DAD approach. The best modeling
; results were evaluated, and a complementary data analysis, as well as a comparison between
; the DAD and MS approaches, was carried out. The results indicated that the MS approach was
; better, but comparable to DAD (chapter 5). Therefore, a market study was conducted for both
; the illegal and legal markets in Belgium, with a focus on weight loss supplements using the
; MS approach. The results revealed instances of herbal adulteration and fraudulent products,
; predominantly in the illegal market, though with also the presence of a banned plant also in the
; legal market. This becomes a matter of grave concern and a direct threat to public health.
; Although the techniques proved to be feasible, a second-step validation is necessary to confirm
; the presence or absence of regulated plants in plant food supplements, especially in the context
; of juridical procedures or decisions to take products out of the market. These can include nextgeneration
; sequencing techniques (NGS) or DNA metabarcoding, which are being explored
; and researched to distinguish plants from mixtures, even at the species level.