Harnessing the Full Power of Chemometric-Based Analysis of Total Reflection X-ray Fluorescence Spectral Data to Boost the Identification of Seafood Provenance and Fishing Areas

Abstract

Provenance and traceability are crucial aspects of seafood safety, supporting managers and regulators, and allowing consumers to have clear information about the origin of the seafood products they consume. In the present study, we developed an innovative spectral approach based on total reflection X-ray fluorescence (TXRF) spectroscopy to identify the provenance of seafood and present a case study for five economically relevant marine species harvested in different areas of the Atlantic Portuguese coast: three bony fish—Merluccius merluccius, Scomber colias, and Sparus aurata; one elasmobranch—Raja clavata; one cephalopod—Octopus vulgaris. Applying a first-order Savitzky–Golay transformation to the TXRF spectra reduced the potential matrix physical effects on the light scattering of the X-ray beam while maintaining the spectral differences inherent to the chemical composition of the samples. Furthermore, a variable importance in projection partial least-squares discriminant analysis (VIP-PLS-DA), with k − 1 components (where k is the number of geographical origins of each seafood species), produced robust high-quality models of classification of samples according to their geographical origin, with several clusters well-evidenced in the dispersion plots of all species. Four of the five species displayed models with an overall classification above 80.0%, whereas the lowest classification accuracy for S. aurata was 74.2%. Notably, about 10% of the spectral features that significantly contribute to class differentiation are shared among all species. The results obtained suggest that TXRF spectra can be used for traceability purposes in seafood species (from bony and cartilaginous fishes to cephalopods) and that the presented chemometric approach has an added value for coupling with classic TXRF spectral peak deconvolution and elemental quantification, allowing characterization of the geographical origin of samples, providing a highly accurate and informative dataset in terms of food safety.

Elemental and spectral chemometric analyses of Octopus vulgaris beaks as reliable markers of capture location

“…Although all the chemometrics approaches were evaluated using the spectral dataset, the most commonly used approach when dealing with spectral data is PLS‐DA (Duarte, Mamede, Carreiras, et al., 2022; Ghidini et al., 2019; Ren et al., 2013). Methods such as PLS‐DA are based on latent variables, which take into account the whole spectrum dataset without having to perform a previous feature selection and are, therefore, among the most accurate methods when dealing with this type of dataset (Leardi, 2000).…”

“…To reduce unwanted effects of light scatter caused by intrinsic physical structure features of the medium of the sample, spectral data transformations are commonly applied (Delwiche & Reeves, 2010; Duarte, Mamede, Carreiras, et al., 2022). For this purpose, the Savitzky–Golay transformation (Savitzky & Golay, 1964) was employed, using a localized linear regression of several neighboring points to determine the best‐fit polynomial differentiated and evaluated at the x values (in this case energy values).…”

“…According to this EU resolution, traceability is an instrument for tracing and tracking any food commodity, through all phases of the supply chain from collection/production to the market (European Commission, 2002). Different types of tags have been developed from elemental and isotopic (Bennion et al., 2021; Davis et al., 2021; Duarte, Duarte, et al., 2022; Mamede et al., 2022), biochemical (Fonseca et al., 2021; Portarena et al., 2015; Ricardo et al., 2015), genomic (e.g., Galimberti et al., 2014; Ogden, 2008), and spectral analyses (e.g., Duarte, Mamede, Carreiras, et al., 2022; Ghidini et al., 2019; Ren et al., 2013; Varrà, Ghidini, et al., 2021). Additionally, different methodologies have been implemented for the development of these tags, from simple statistical multivariate analysis (Duarte, Duarte, et al., 2022; Varrà et al., 2019) to advanced multi‐model chemometrics approaches (Duarte, Mamede, Duarte, et al., 2022; Varrà, Ghidini, et al., 2021) and machine learning algorithms (Milan et al., 2019).…”

“…Elemental tags of capture location have high efficiency for a wide range of food matrixes, including muscle, gonad, calcified structures, and secretions (Bennion et al., 2021; Duarte, Carreiras, et al., 2022; Mamede et al., 2022; Reis‐Santos et al., 2022; Ricardo et al., 2015; Varrà, Husáková, et al., 2021), highlighting the relevance of multi‐elemental analytical tools in the context of traceability studies. Total X‐ray fluorescence (TXRF) spectroscopy is a multi‐elemental high‐throughput untargeted analysis that can efficiently quantify the elemental concentrations in seafood products and, therefore, has been used to assess food safety and trace product geographical origin (Kelly et al., 2005; Lim et al., 2021; Rajapaksha et al., 2017) with the latter amplified by the discriminatory power of chemometrics analyses based on the X‐ray spectral signatures (Duarte, Mamede, Carreiras, et al., 2022).…”

Combination of Total-Reflection X-Ray Fluorescence Method and Chemometric Techniques for Provenance Study of Archaeological Ceramics

“…Recently, studies have explored the combination of TXRF with chemometric methods as a possible tool for the analysis of complex samples. This combination was already successfully applied in the discovery of obesity markers [ 29 ], identification of the geographical origin of bean seeds [ 30 ], classification of the origin and type of wine samples [ 31 ], discrimination of gunshot residues [ 32 ], and seafood provenance assessment [ 33 , 34 ]. This approach was also applied to the analysis of archaeological ceramics [ 35 ], but the data in this work were investigated using principal component analysis (PCA), which is an exploratory method and does not allow the classification of samples.…”