Sample collection

Four artisanal cheese dairies producing waxed dry cheese were selected after obtaining the owner's consent to participate in the study. The dairies are located in the towns of Tapanatepec (16° 22′ 0″ N, 94° 12′ 0″ W) and Zanatepec (16° 28′ 26.4″ N, 94° 21′ 5.29ʹ W), Oaxaca, Mexico. Using non-probability sampling, samples from each cheese producer were taken randomly in triplicate (approximately 100 g) from different batches of milk, and analysed at two ripening times, 7 and 30 days, resulting in a total of 24 samples. Later, cheeses were ground to reduce particle size and mixed to homogenize.

FT-NIR and physical-chemical analysis

In the physicochemical analysis, the partial proximate chemical composition of the cheese was evaluated, to determine protein, fat, moisture, total solids, and salt contents, using the dairy FT-NIR analyser FoodScan^TM Lab (FOSS, Inc, Hilleroe, Denmark). Thus, 100 g of each waxed dry cheese sample were placed in the glass Petri dish on the rotating plate of the equipment. The FT-NIR spectra were recorded using a transmittance wavelength range of 850–1048 nm. Proximate parameters were estimated using the FossManager™ software (FOSS, Inc., Hillerød, Denmark). The pH was determined using a potentiometer (HI98107, Hanna Instruments, Inc, Woonsocket, RI, USA). Water activity (Aw) was determined using an Aqualab water activity meter (3TE Series, Decagon Devices, Inc, Washington, USA). All assays were performed in triplicate.

Statistical analysis of FT-NIR data

Physicochemical data were analysed using a completely randomised design with an asymmetric factorial (A x B) arrangement, where A is the time factor with two levels (7 and 30 days), and B is the treatment factor (cheese dairies) with four levels. An analysis of variance (ANOVA) was applied with α = 0.05, where a statistically significant difference was identified with a (p-value ≤ 0.05). Finally, a comparison of effects between treatment means was performed using Tukey's method. Statistical univariate analysis was carried out using SAS statistical software version 9.4 (SAS, Institute Inc., Cary, NC, USA).

FT-IR-ATR spectroscopic analysis

FT-MIR spectra were obtained by FT-IR-ATR. Approximately 2 mg of waxed dry cheese sample (n = 24) were placed on the ATR crystal and pressed until the signal intensity was recorded (each sample was processed in triplicate). FT-IR-ATR spectra were acquired between 4000 and 500 cm⁻¹ with a resolution of 4 cm⁻¹ and 64 scans in a Thermo Scientific™ Nicolet iS5 Thermo Scientific™ (Thermo Electron Scientific Instruments LLC, Madison, WI, USA). This Fourier transform mid-infrared spectrometer is equipped with an iD7 attenuated total reflectance accessory with a monolithic diamond crystal. The spectrometer was operated using the OMNIC™ software (Thermo Scientific, Madison, WI, USA). FT-IR-ATR collected spectra were exported as CSV files to further analysis.

FT-IR-ATR data preprocessing

FT-IR-ATR data initial preprocessing was performed using Excel (version 2502, Microsoft Office® 365) as follows. From data exported as CSV files, an initial matrix was generated, consisting of 7261 rows (recorded as wave numbers) and 72 columns (representing 24 cheese samples, each with three replicates, grouped by cheese producer). Data at the extremes of the spectra (400–899 and 3701–4000 cm⁻¹) and in the middle region (from 1801 to 2399 cm⁻¹) were removed due to the lack of relevant bands.Two different spectral regions were retained in the ranges of 900–1800 cm⁻¹ and 2400–3700 cm⁻¹, as these regions exhibit the highest number of bands with discriminant potential (Das et al., Reference Das, Chowdhury, Chakraborty, Sikdar, Saha, Mukherjee, Karmakar and Chattopadhyay2021). The average spectra for the replicates of each sample were calculated to reduce sample variation. This action was conducted to obtain a new matrix consisting of 5082 rows (wave numbers recorded) and 24 columns (cheese samples, grouped by cheese factory). To facilitate data reduction, a binning procedure was performed in intervals of 21 data points along the wavenumber axis. Consequently, the final FT-IR-ATR data matrix had 204 rows (wavenumber averages) and 24 columns. Additional data preprocessing for the chemometric analysis using PCA (as exploratory technique) and PLS-DA (as discriminative technique) were conducted utilizing the Metaboanalyst 6.0 platform with the “Statistical Analysis One Factor” module (Canadian Institutes for Health Research, 2024). This is a well-stablished online server dedicated to metabolomics data statistical analysis, interpretation and integration. Data preprocessing was started by selecting spectral bins with unpaired columns. Subsequently, data filtering included applying a variance filter using the median absolute deviation, and an abundance filter using mean intensity value. Finally, the data were scaled using Pareto scaling (Sysi-Aho et al., Reference Sysi-Aho, Katajamaa, Yetukuri and Orešič2007). In addition, cross-validation (considering three components) and permutation tests (using B/W separation distance, and set permutation numbers of 100 and 1000, respectively) were performed to assess the PLA-DA model reliability.

FT-NIR and physicochemical analysis of waxed dry cheese

FT-NIR spectroscopy is a well-established technique to determine the proximate composition of cheese based on chemometric methods (Stocco et al., Reference Stocco, Cipolat-Gotet, Ferragina, Berzaghi and Bittante2019). Despite its limitations, this alternative tool provides fast, accurate, and reliable results The most frequently predicted parameters include moisture, fat, protein, and NaCl (Bittante et al., Reference Bittante, Patel, Cecchinato and Berzaghi2022). In the study of waxed dry cheese, total solids (TS) and non-fatty solids (NFS) were also evaluated using FT-NIR. In addition, pH was measured by potentiometry, and Aw was determined through moisture analysis (Kaczyński, Reference Kaczyński2024). In Table 1 (where the four artisanal cheese dairies were named A, B, C, and D), the changes in the proximate composition and physicochemical variables monitored during the ripening period of waxed dry cheese are shown. A statistically significant increase (p < 0.05) was observed in the percentage content of protein, fat, non-fat solids, salt, and total solids. Conversely, moisture, pH, and Aw declined during the ripening process.

Table 1.

FT-NIR physicochemical analysis of waxed dry cheese produced in four different artisanal cheese dairies (A-D) in oaxaca, Mexico

TS: Total solids; NFS: Non fatty solids

* Statistical difference between ripening time (p < 0.05);

ᵠ Statistical difference between artisanal cheese dairies (p < 0.05).

The increase in protein content during cheese ripening has been previously reported in dairy products, such as Van Herby cheese and white cheeses from Turkey (Özcan Yardım and Durak, Reference Özcan Yardım and Durak2023); also, for the Zacazonapan aged cheese from México (Torres-Salas et al., Reference Torres-Salas, Hernández-Montes and Hernández-Rodríguez2023). Thus, for waxed dry cheese an increasing protein concentration as a function of ripening time was expected. Simultaneously, the reductions in moisture and Aw are correlated, because both indicate water loss. In conjunction, these processes are common during cheese ripening and have been previously reported in feta cheese (Thodis et al., Reference Thodis, Kosma, Nesseris, Badeka and Kontominas2023), cheddar cheese (Mamo, Reference Mamo2017), and Mexican Manchego-style cheese (Sardiñas-Valdés et al., Reference Sardiñas-Valdés, García-Galindo, Chay-Canul, Velázquez-Martínez, Hernández-Becerra and Ochoa-Flores2021). Additionally, it has been reported that a high salt content in the curd stimulates syneresis in cheeses, leading to moisture loss and an increase in protein concentration in the curd (Guinee and Fox, Reference Guinee and Fox2017). Moreover, ripening time plays a crucial role, since during this period the protein matrix undergoes changes such as the binding of carboxyl groups and the release of amino acids (Sousa et al., Reference Sousa, Ard and Mcsweeney2001). In the case of waxed dry cheese, the moisture loss is associated with the characteristics of the process, mainly due to the high salt addition during dry salting and throughout ripening, where various factors stimulate water loss. Additionally, the increase in fat content during cheese ripening is associated with the moisture reduction. However, the increase in total fat content is also related to lipolysis, a biochemical process involving the hydrolysis of fat and the release of fatty acids (Thierry et al., Reference Thierry, Collins, Abeijón Mukdsi, McSweeney, Wilkinson and Spinnler2017). Indeed, cheese lipolysis is a crucial process that significantly contributes to the development of the cheese aroma and flavour, contributing with these characteristics in waxed dry cheese.

Concomitantly, the decrease in pH is related to the catabolism of lactose and the production of organic acids, primarily by lactic acid microorganisms (De Jesus et al., Reference De Jesus, Silva, Gonçalves, De Souza, Santos and Ferrão2020). In the case of waxed dry cheese, the reduction in pH may be related to the metabolism of lactose through the metabolic pathways of lactic acid bacteria, to produce lactate, citrate and acetate, all of them acidic metabolites (Jo et al., Reference Jo, Benoist, Ameerally and Drake2018). It has been reported that changes in pH have a direct relationship with the time and temperature of ripening (> 30 °C), resulting in various sensory changes, mainly related to flavour and aroma (Califano and Bevilacqua, Reference Califano and Bevilacqua2000). According to the analysis conducted by (Beresford and Williams, Reference Beresford, Williams, Fox, McSweeney, Cogan and Guinee2004), raw cow's milk contains a higher number of lactic acid microorganisms, leading to a wide diversity of ripening microorganisms, which results in specific biochemical and sensory properties that characterize this type of cheese. Waxed dry cheese is produced from raw milk and rennet from Zebu cattle, and undergoes a ripening process for a period of 30 days at environmental temperatures above 30 °C. It is expected that the present microorganisms contribute to the accumulation of organic acids, which would explain the significant changes in pH during cheese ripening.

In addition, during the transformation of raw milk into cheese, a series of fermentations takes place, resulting in the hydrolysis of proteins, or proteolysis. This process is driven by the metabolism of different proteolytic enzymes, which may be native to raw cow's milk, added through rennet, or released by lactic acid bacteria (Fox et al., Reference Fox, Guinee, Cogan and McSweeney2017). Proteolytic processes are responsible for the release of low molecular weight proteins, water-soluble peptides, and free amino acids, leading to an increase in protein concentration (Mamo, Reference Mamo2017). Since waxed dry cheese is produced from raw cow's milk, a higher amount of native proteolytic enzymes is presumed. Similarly, the rennet is natural, derived from bovine abomasum. In conjunction, these factors, along with the ripening process, might enhance proteolytic activity. However, further studies are necessary to describe the biochemical process occurring during waxed dry cheese ripening and explain the changes in proximate composition, pH and Aw.

FT-IR-ATR spectroscopic analysis of waxed dry cheese

Cheese samples were analysed by means of FT-IR-ATR to detect chemical changes related to ripening process, with the advantage of direct analysis without an extensive sample preparation. In a second step, a chemometric model based in FT-IR-ATR spectra was developed. Figure 1 shows representative FT-IR-ATR spectra obtained from the waxed dry cheese samples, along with the assignment of the main bands associated with the spectral vibrations of fatty acids, proteins, and carbohydrates molecules, mostly. In this figure, distinguishable characteristics related to the two different ripening periods of cheese are observed. These vibrational changes are correlated to variations in the cheese physicochemical composition. A more detailed description of the band assignment is provided below.

Figure 1.

FT-IR-ATR selected spectra of waxed dry cheese at two ripening times: 7 days () and 30 days (). Most relevant bands were assigned according to different functional groups.

In the spectra shown in Figure 1, two important regions are easily noticeable within the spectral range of 3700–2400 cm⁻¹ and 1800–900 cm⁻¹. Particularly, in the range of 3500–3100 cm⁻¹, intense bands are clearly observed. These bands are related to the stretching vibrations of – OH groups from water molecules, as well as hydroxyl groups in carbohydrates and proteins, and the stretching vibration of the – NH bond associated with the peptide bond in proteins. The low absorption bands at 2960 and 2879 cm⁻¹ were assigned to the asymmetric and symmetric stretching of – CH₃, respectively. Meanwhile, the intense bands identified at 2916 and 2850 cm⁻¹ were attributed to the asymmetric and symmetric stretching of – CH₂ – groups. These stretching vibrations of methyl and methylene groups mainly correspond to the aliphatic chains of fatty acids. This assignment is consistent with the marked difference in absorption for the bands of – CH₂ – compared to – CH₃ groups, which is expected for this type of molecule. In the second relevant region of the waxed dry cheese spectrum, which ranges from 1800 to 900 cm⁻¹, the characteristic vibrations of distinctive functional groups were assigned as can be read below. The low-intensity bands identified between 1728 and 1715 cm⁻¹ were associated with the C = O symmetric stretching of the carbonyl group, from esters and carboxylic acids (Andrade et al., Reference Andrade, Pereira, Ranquine, Azarias, Bell and De Carvalho Dos Anjos2018). In samples such as waxed dry cheese, this can be primarily attributed to the carbonyl groups of the lipid carboxylate group (Sakkas et al., Reference Sakkas, Pappas and Moatsou2021), with a lesser contribution from acidic amino acids such as glutamate and aspartate.

In addition, in the spectra of waxed dry cheese, the bands at 1625 and 1534 cm⁻¹ stand out due to their intensity. These bands were assigned to the amide I band (1625 cm⁻¹), associated with the C = O symmetric stretching of carbonyl groups, and the amide II band (1534 cm⁻¹), attributed to the N–H bending vibration of the functional groups. Amide I and II bands in the spectra of waxed dry cheese indicates the presence of the peptide bonds in milk proteins, as expected. Particularly, for cheddar cheese it has been reported that the intensity of the amide I band increase and amide II band decreases as a function of ripening time (Dewantier et al., Reference Dewantier, Torley and Blanch2023), as this change is caused by protein hydrolysis. Additionally, changes in the intensity of the amide I peak could be associated with modifications in the secondary structure of casein, protein aggregation, and protein-water interactions (Martín-del-Campo et al., Reference Martín-del-Campo, Picque, Cosío-Ramírez and Corrieu2007). Indeed, during the ripening process of waxed dry cheese, a slight reduction in the intensity of amide I and II bands was observed (see Figure 1), despite the increase in the protein content (as indicated in Table 1). Consequently, further analysis of the amide bands (such as peak deconvolution and area under curve statistics) is needed to establish the biochemical process related to these changes in waxed dry cheese.

The region between 1480–1355 cm⁻¹ exhibits the characteristic scissoring band of methylene groups (1465 cm⁻¹), asymmetrical bending vibration of methyl groups (1450 cm⁻¹), and symmetrical bending vibration of methyl groups (1375 cm⁻¹), all with moderate absorption. For cheeses, these bands are associated with – CH₃ and C–H functional groups (Sakkas et al., Reference Sakkas, Pappas and Moatsou2021), mainly from lipid molecules and aliphatic chains of amino acids. In this region, the O–H group bending vibration is also observed, generally between 1420-1330 cm⁻¹. Usually, in primary and secondary alcohols the O–H bending is coupled to C–H wagging vibrations, to give to two bands near to 1420 and 1330 cm⁻¹. In Waxed Dry Cheese FT-IR-ATR spectra, only one band is observed at 1417 cm⁻¹. This vibration could be generated by contributions from both the in-plane vibrations of the O–H groups of carbohydrates and the wagging vibrations of the – CH₂ – groups of lipids and amino acids. It is noteworthy that in an FT-IR study of 81 artisanal cheeses from Turkey made from sheep and goat milk, these same bands were observed (Ayvaz et al., Reference Ayvaz, Mortas, Dogan, Atan, Yildiz Tiryaki and Karagul Yuceer2021).

Finally, in the lower energy region of the FT-IR-ATR spectrum, ranging from 1330 to 950 cm⁻¹, several bands (1230, 1177, 1087, 1059, and 981 cm⁻¹) of waxed dry cheese were identified. These bands were associated with the C–O stretching vibrations and the C–C–O bonds asymmetric vibrations. Both functional groups are associated with the presence of carbohydrates, such as lactose (Andrade et al., Reference Andrade, Pereira, Ranquine, Azarias, Bell and De Carvalho Dos Anjos2018), as well as other functional groups such as ethers. The O–H out-of-plane bending vibration of carboxylic acid is observed at 920 cm⁻¹, as can be seen in the waxed dry cheese spectra.

It is important highlight that the FT-IR-ATR profile or waxed dry cheese throughout ripening, is closely similar to the maturation process of other artisanal cheeses, such as the Brazilian Minas cheese (De Jesus et al., Reference De Jesus, Silva, Gonçalves, De Souza, Santos and Ferrão2020) and the French Maroilles cheese (Karamoko and Karoui, Reference Karamoko and Karoui2024).

Partial least squares – discriminant analysis (PLS-DA)

Several authors have highlighted the utility of applying multivariate statistical methods to the spectroscopic analysis of chesses for purposes such as identification, authentication, and quality control, ripening monitoring, among others (Maestrello et al., Reference Maestrello, Solovyev, Stroppa, Bontempo and Franceschi2024; Visconti et al., Reference Visconti, Diniz, de Sousa Fernandes, Rodriguez and Di Anibal2024). In the study herein described, the evaluation of the waxed dry cheese maturation process was conducted in two stages: first through principal component analysis (PCA), followed by partial least squares – discriminant analysis (PLS-DA), using FT-IR-ATR spectral bins as input variables. To achieve this, FT-IR-ATR data preprocessing steps were applied, including binning, filtering, and scaling (Mokari et al., Reference Mokari, Guo and Bocklitz2023). Binning facilitated data reduction, while variance and abundance filtering, based on the median absolute deviation and mean intensity value, respectively, were employed to minimize data dispersion. Finally, Pareto scaling (mean-centred divided by the square root of the standard deviation of each variable) was applied to reduce the influence of large values resulting from the strongest vibrations also minimizing noise impact (Khalighi et al., Reference Khalighi, Ma, Ren and Varveri2024).

The goal of PCA is to reduce the dimensionality of the input variables to create a smaller dataset by selecting variables with significant variance (Fattahi et al., Reference Fattahi, Kazemi, Khojastehnazhand, Roostaei and Mahmoudi2024). In addition, PCA has been proposed as a practical indicator of the reliability of further discriminant, analysis such as PLS-DA and orthogonal PLS-DA (Worley and Powers, Reference Worley and Powers2016). In this study, the PCA model was used as an exploratory tool to verify similarities among samples. Visual inspection of scores and loading plots was used to identify potential clustering distribution. To perform the PCA of waxed dry cheese from FT-IR-ATR data, the MetaboAnalyst 6.0 platform was used. As shown in the PCA plots included in the Supplementary Material file (Figure SM1), the model reached a 93.3% of variance explained with three components. However, the 3D score plot (PC1 vs PC2 vs PC3) revealed a slight overlap between Hotelling's ellipses.

Subsequently, PLS-DA was applied to the spectra to address the clusters overlap, identify the most important variables, and select those that contributed significantly to sample discrimination. The resulting PLS-DA model (also generated using MetaboAnalyst 6.0) was calculated with eight components, which accounted for 99.7% of the cumulative variance (Figure 2a). However, to determine the optimal number of components that adequately describe the model, the first derivative method of cumulative variance was used to identify the inflection point. The inflection point, where slope changes, served as the selection criterion for the optimal number of components (Figure 2b). Thus, three principal components were established for the PLS-DA model, representing 90% of the cumulative variance (PC1: 64.8%, PC2: 13.2%, PC3: 12%). According to Lei et al., (Reference Lei, Lin and Sun2019), correct classification rates for the validation set in a PLS-DA model are achieved when the selected number of components explains more than 80% of the data in relation to the model. In the PLS-DA model derived from waxed dry cheese this criterion was achieved.

Figure 2.

PLA-DA of FT-IR-ATR data of waxed dry cheese. a) accumulated variance contributions plot for each component in the PLS-DA model: number of model components () and model accumulated variance days (); b) first derivative of variance cumulative contributions showing the inflection point (slope change) at principal component number 3; c) 3D score plot of the PLS-DA model at 7 days () and 30 days () of ripening; and, d) 3D loading plot of the PLS-DA model, the marker became reddish as its contribution to differentiate groups increase.

Figure 2c presents the three-dimensional plot of the first three components of the PLS-DA model, showing distinguishable clustering related to the ripening time of the waxed dry cheese. This cluster distribution indicates that despite cheese spectra have a very similar band profile, they also exhibit differential characteristics depending on ripening times. This finding aligns with results reported by Vásquez et al. (Reference Vásquez, Magán, Oblitas, Chuquizuta, Avila-George and Castro2018), who observed a correlation between spectral profile changes and ripening times in Swiss-type cheeses. Finally, Figure 2d displays the PLS-DA loading plot based on three components, representing 90% of the model's variance explanation. In this plot, the distribution of loadings allowed to distinguishing variables related with samples clustering. Specifically, for waxed dry cheese, this plot enabled visualization of the differential variables associated with ripening time.

In discriminant methods, a high number of variables combined with a small sample size tends to lead to data overfitting. To ensure the reliability of the waxed dry cheese PLS-DA model, cross-validation was performed based on accuracy, R², and Q². Accuracy is particularly relevant because it measures the performance of the model and indicates how well different groups (or classes) are distinguished. R² indicates model's data variation, while Q² evaluates its predictive capability for new data. For the three-component PLS-DA model, the accuracy, R², and Q² values obtained were 0.900, 0.831, and 0.616, respectively. Although the ideal value for these parameters is 1, achieving such values is challenging due to the inherent variation among samples of the same class (Westerhuis et al., Reference Westerhuis, Hoefsloot, Smit, Vis, Smilde, Velzen, Duijnhoven and Dorsten2008). In fact, a cut-off value above 0.5 is considered as acceptable (Xu et al., Reference Xu, Bai, Chen, Yu, Wu and Hu2024). Furthermore, permutation tests provide an additional approach to determining whether the classification into groups is statistically significant, also useful to identifying potential misclassifications. The resulting permutation test plots obtained in MetaboAnalyst (Figure SM2 in Supplementary Material) exhibits a rightward-extended distribution, indicating that the separation between the two groups becomes increasingly significant as the number of permutations increases. In this way, cross-validation and permutation tests demonstrated the reliability of the PLS-DA model derived from FT-IR-ATR data of waxed dry cheese in discriminating between ripening times.

As previously noted, in the waxed dry cheese FT-IR-ATR spectra, signal intensity changes characterize the different maturation stages, similar results were also reported for cheddar cheese (Dewantier et al., Reference Dewantier, Torley and Blanch2023). Fig. 3 presents the main VIP (variable importance in projection) values obtained in the first three components of the PLS-DA analysis, where changes were observed across various spectral regions. In the 3482–3336 cm⁻¹ region (Figure 3a), a decrease in spectral vibrations associated with water was identified, indicating water loss during cheese maturation. This process is likely influenced by syneresis and leads to an increase in cheese firmness (Fox et al., Reference Fox, Guinee, Cogan and McSweeney2017). Additionally, an increase in peaks intensity was observed in the spectral region of 2912 and 2849 cm⁻¹ (Figure 3b), a region associated with fatty acids. This behaviour has been related to the hydrolysis of triglycerides by milk lipases, resulting in the release of free fatty acids (De Jesus et al., Reference De Jesus, Silva, Gonçalves, De Souza, Santos and Ferrão2020). In addition, Andrade et al. (Reference Andrade, Pereira, Ranquine, Azarias, Bell and De Carvalho Dos Anjos2018), reported that lipolysis occurring in most ripened cheeses causes shifts in absorption bands linked to CH₂ and CH₃ functional groups. Both reports are in close correspondence with our findings.

Figure 3.

VIP score plot (high ; low ) from the first three principal components used in the PLS-DA analysis, T1 (7 days) and T2 (30 days); a) VIP scores for the first principal component; b) VIP scores for the second principal component; and c) VIP scores for the third principal component.

In Figure 3c, changes in band profile are observed in the region of 1633, 1623, and 1612 cm⁻¹ during the ripening process of waxed dry cheese. According to Yaman et al. (Reference Yaman, Aykas, Jiménez-Flores and Rodriguez-Saona2022) the spectral range of 1680 a 1533 cm⁻¹, is associated to amide I and amide II bands. The amide I vibration primarily results from the C = O stretching of proteins, along with contribution of C–N and – NH stretching. The out-of-plane stretching of – NH, C–N, and stretching of C–O, C–C, N–C, generate the amide II band. Additionally, Barth (Reference Barth2007) reported that in the wavenumber range of 1633 to 1612 cm⁻¹, bands are mainly related to certain amino acids, including arginine (1633–1614 cm⁻¹), histidine (1631 cm⁻¹), lysine (1626–1629 cm⁻¹¹), tyrosine (1614–1621 cm⁻¹), and asparagine (1612 cm⁻¹). This molecular fingerprint related to proteins content could be related to the changes observed in the waxed dry cheese infrared spectra. Finally, the dendrogram and heatmap from PLS-DA model showed the clustering related to wave number variables, confirming the differences associated with the ripening process (shown in Figure SM3 in Supplementary Material).

FT-IR-ATR spectroscopy provides an accessible analytical tool for waxed dry cheese production. FT-IR-ATR has been proven to assess the yield and composition of the products (Margolies and Barbano, Reference Margolies and Barbano2018); provide a strategy to establish quality parameters (Ayvaz et al., Reference Ayvaz, Mortas, Dogan, Atan, Yildiz Tiryaki and Karagul Yuceer2021), and in the case of waxed dry cheese, monitoring changes during maturation. FT-IR-ATR was conducted to characterize the unique fingerprint of this artisanal cheese made from raw bovine milk, which supports authenticity and provides added value. However, despite the results described herein, complementing the study of waxed dry cheese with chemometric models based on NMR and FT-NIR spectral data would be desirable. These highly robust techniques for cheese analysis could provide new insights into the characterization of this emblematic product from Southern Mexico.