Results obtained suggested that the traceability 36system in the fishery sector may be supported by simplified machine learning techniques applied to a limited 37but effective number of inorganic predictors of origin.Īn effective and trustworthy traceability system contributes to improving food quality and safety and responds to consumers’ demand for food provenance information. The finished packaged product was better modelled 34by the QUEST algorithm which recognised the origin of anchovies with F-score of 97.7%, considering the 35information carried out by 5 elements (B, As, K. Classification rules 32generated by the trained CHAID model optimally identified unlabelled testing bulk anchovies (93.9% F-score) 33by using just 6 out 52 elements (As, K, P, Cd, Li, and Sr). To this purpose, four decision trees-based algorithms, corresponding 29to C5.0, classification and regression trees (CART), chi-square automatic interaction detection (CHAID), and 30quick unbiased efficient statistical tree (QUEST) were applied to the elemental datasets to find the most 31accurate data mining procedure to achieve the ultimate goal of fish origin prediction. Quadrupole inductively coupled plasma mass spectrometry (Q-ICP-MS) and direct mercury analysis were used 27to determine the elemental composition of 180 transformed (salt-ripened) anchovies from three different 28fishing areas before and after packaging. One application dealing with the use of multielemental analysis to authenticate seafood products is also available and it concerns the identification of caviar from different origins, which, as anchovy, is a salted product (Rodushkin et al., 2007). Nevertheless, the multiple identification of elements using techniques such as inductively coupled plasma-optical emission spectroscopy (ICP-OES) and inductively coupled plasma-mass spectrometry (ICP-MS) have been already successfully applied to identify the origin of transformed food products such processed tomato products (Fragni, Trifirò, & Nucci, 2015 Lo Feudo, Naccarato, Sindona, & Tagarelli, 2010), fruit juices (Turra et al., 2017), wines (da Costa, Ximenez, Rodrigues, Barbosa, & Barbosa, 2020), dried beef (Franke et al., 2008), hams (Epova et al., 2018), and different types of cheese (Magdas et al., 2019 Moreno-Rojas, Camara-Martos, Sanchez-Segarra, & Amaro-Lopez, 2012 Suhaj & Koreňovská, 2008). The use of salt during anchovy manufacturing may represent the most critical point since it can potentially mask the natural elemental content of fish. Considering the high efficiency, repeatability and stability of DART-QTOF acquisition, it is necessary to optimize its data acquisition and processing methods, in order to realize the faster assessment for lipidomics as well as food authenticity in the future.
Based on LC-ESI-QTOF (+) and LC-ESI-QTOF (-), 40 potential markers were found, and beef samples from Canada, Argentina and New Zealand shared higher similarity, while beef samples from Brazil, Australia and Uruguay had more similarities.
Then, support vector machine (SVM) was constructed based on them and LC-ESI-QTOF (+) was superior to DART-QTOF (+) in SVM models, achieving a highest training accuracy of 100% and a validation accuracy of 89.91%. The data were screened based on detection rate, RSD value, fold change and P value, and library matching to obtain 852 and 879 raw peaks, 62 and 165 differential peaks and 17 and 25 potential markers in DART-QTOF (+) and LC-ESI-QTOF (+), respectively. Compared to LC-ESI-QTOF (+), DART-QTOF (+) is particularly suitable for smaller compounds analysis, and has better repeatability and stability. The comparison of DART-QTOF and LC-ESI-QTOF was studied on the discrimination of beef from different origins based on lipidomics. For a higherĭiscriminative power, this method should be combined with other ways of authentication.
Validation allowed identifying some, but not all, origins. The LDA gave mean correct classification rates of 77 andħ9% for poultry meat and dried beef, respectively. Sr, Te, Tl, U, and V for dried beef out of about 50 elements each. Elements significantly discriminatingĪmong countries were As, Na, Rb, Se, Sr, and Tl for poultry meat and As, B, Ba, Ca, Cd, Cu, Dy, Er, Fe, Li, Mn, Pd, Rb, Se, Validation was performedīy estimating the origin of the first samples based on the data of the second, larger, dataset. In order to validate the applicability of this technique, the results were additionallyĬombined with data from an earlier assessment including 25 poultry meat and 23 dried beef samples. Of variance and linear discriminant analysis (LDA) to identify the single or combination of elements with the highest potential Element concentrations of 56 poultry meat and 53 dried beef samples were determined and statistically analyzed using analysis
0 kommentar(er)
