de Andrade, Barbara M., Margalho, Larissa P., Batista, Diego B., Lucena, Izylla O., Kamimura, Bruna A., Balthazar, Celso F., Brexó, Ramon Peres, Pia, Arthur K.R., Costa, Ramon A.S., Cruz, Adriano G., Granato, Daniel, Sant'Ana, Anderson S., Luna, Aderval S., and de Gois, Jefferson S.
This work aims to determine the mineral composition of Brazilian artisanal cheese (BAC) and classify the cheese types using chemometric techniques. Samples of BAC were analyzed and divided according to their region: Northeast cheeses (Coalho and Manteiga); Midwest cheeses (Caipira); Southeast cheeses (Araxá, Campo das Vertentes , Cerrado , Serra da Canastra and Serro) and South cheeses (Colonial and Serrano). Major (Ca, K, Mg, Na) and trace elements (Cu, Mn, Zn) content were determined by inductively coupled plasma optical emission spectrometer (ICP-OES). Artificial neural network (ANN), K-nearest neighbor (KNN), Random Forest (RF), Support vector machines with the radial kernel (SVM), and Learning Vector Quantization (LVQ) were used as supervised statistical methods to differentiate and classify the cheeses according to the type, and producing region. Coalho cheese showed the highest Ca, Cu, Mn, and Zn content with 91.09 mg g−1, 63.51 μg g−1, 3.29 mg g−1, and 137.51 μg g−1, respectively. Canastra cheese showed the highest K and Na content with 11.68 mg g−1 and 96.98 mg g−1, respectively. Cerrado cheese showed the highest Mg content (7.56 mg g−1). These differences in mineral content are explained by the type of cheese (fresh, ripened), producing region (climate, animal feeding, raw milk), and the technological process of cheese making (salts, metallic instruments). BAC is an essential source of minerals beneficial to health, and its authenticity is paramount for the consumer. RF and SVM classifiers were able to classify the cheese type dataset using the accuracy and Kappa coefficient as the figures of merit (Accuracy= 0.8347 and Kappa= 0.8105, Accuracy=0.8323 and Kappa=0.8078, respectively). For the classification of the production region, all algorithms presented excellent figures of merit. The classification performance can be characterized as outstanding for both cases. Such techniques guarantee the authenticity of Brazilian artisanal cheeses and the Protected Designation of Origin of the cheeses from each region, enhancing the added value of the cheese. • The mineral composition of Brazilian artisanal cheese was determined. • Statistical and chemometrics tools were applied to classify cheese. • Random forest and Support vector machine models with satisfactory results. • Techniques that guarantee the authenticity of artisanal cheeses were developed. [ABSTRACT FROM AUTHOR]