Pomegranate (cv. Wonderful) at commercial maturity was obtained from Sonlia commercial packhouse, located in the area of Western Cape Province, South Africa (33°34′851″S, 19°00′360″E). To add variability, fruit obtained from the packhouse was identified to originate from two different orchards (Wellington and Porterville). A total of 200 fruit was transported to the Postharvest Technology Research Laboratory at Stellenbosch University. On arrival, fruit was equilibrated at ambient conditions before being sorted out for physical defects and uniformity. Fruit was then stored at 7 ± 0.3°C, 91 ± 3% RH for <5 days before processing. Three different extraction methods were employed as demonstrated in Table 1. For each extraction method, a total of 60 fruit was processed. The extracted juice from each method was then immediately stored at −80°C until further analysis.

Spectral acquisition of pomegranate juice was acquired in the mid-infrared region of 12,500–4,000 cm⁻¹ using the Alpha-P ATR-FT-MIR spectrometer (Bruker Optics, Ettlingen, Germany) that measured diffuse reflectance. The Alpha-P spectrometer is equipped with a high-throughput ZnSe ATR diamond crystal, with six internal reflectors, and an exceptionally high light throughput providing the highest ATR measurement sensitivity for the analysis of even low concentrated sample components. Prior to obtaining FT-MIR spectra, reference measurements were performed against air background and periodically at intervals of 30 min during sample spectra acquisition. After every subsequent measurement, the scanning stage was cleaned using distilled water and tissue paper. The Alpha-P instrument was operated under the following conditions: sample plate temperature was maintained at 40°C, 4 cm⁻¹ resolution scan, and 10 kHz scanner velocity with a total of 128 averaged background and sample scans per spectrum.

Crude pomegranate juice (1 ml) was extracted with 29 ml of 50% aqueous methanol solution (v/v). The resulting mixture was vortexed and then sonicated (DC400H, MRC Ltd., Israel) (frequency: high setting and power: 0.05 W) for 20 min in a cold water bath followed by centrifuging at 10,000 rpm for 5 min at 4°C (Merk, Eppendorf AG, Germany). The supernatant was subsequently collected into clean vials and stored at −80°C until the assay for phenolic components and antioxidant properties was performed.

Data processing and discriminant analysis were performed using SIMCA version 14 (Umetrics, Umeå, Sweden). In order to remove unwanted variation in the spectral data, several spectral filtering methods were evaluated. The baseline-corrected spectra were subjected to Savitzky–Golay transformation (first and second derivative), multiplicative scattering correction (MSC), and standard normal variate (SNV) correction. Explorative data analysis was performed using principal component analysis (PCA). PCA is an unsupervised method and is used to investigate the structure of a dataset and attempts to model the total variance of the original dataset via the uncorrelated principal components (Cozzolino et al., 2011). While orthogonal partial least squares discriminant analysis (OPLS-DA) is a supervised classification technique that isolates a predictive component and integrates an orthogonal correction filter, in order to differentiate the variation within the dataset (Wold et al., 1987; Bylesjö et al., 2006). Spectral data were subjected to PCA in order to investigate the separation of juice extraction methods and chemistry, while OPLS-DA was used for discrimination between different extraction methods.

Validation of the spectral dataset was accomplished by randomly splitting the dataset into test (50%) and train (50%) sets with the scale set as mean centering. OPLS-DA dataset was subjected to an S-Line plot. The S-line plot is especially suited for spectroscopy data. The plot displays the predictive loading in a form resembling the original spectra with the spectral peaks at the top end of the color scale influence the separation of the groups. Prediction datasets were subjected to variable importance projection (VIP), which summarizes the importance of the variables both to explain X and to correlate to Y. Terms with VIP values larger than 1 for variables show with large importance for model performance.

Statistical analyses were performed using statistical software (STATISTICA, version 16, StatSoft Inc., USA). Where appropriate, data were subjected to the one-way ANOVA according to Duncan's multiple range test and Pearson's correlation analysis. Results were presented as mean and standard errors for all studied variables.

Extraction method significantly (p < 0.05) affected the color properties of pomegranate juice (Table 2). Juice extracted by pressing half of the fruit using a commercial hand press (CH) showed the highest redness (a^* = 32.67), hue color (h° = 19.51), and color intensity (C^* = 34.77) compared with the rest of the extraction methods (Table 2). Our results are contrary to Miguel et al. (2004) who reported no significant differences in pomegranate juice color using two different extraction methods. Considering that each extraction method utilizes different fruit fractions, redness for pomegranate juice is in the order of half fruit > arils plus seed > arils. The higher red color readings obtained from CH extraction may be related to increased extractability of several anthocyanin and phenolic compounds found in the exposed rind, arils, and membrane of the halved pomegranate fruit.

For chemical properties, the method of extraction significantly (p < 0.05) influenced the TSS of pomegranate juice (Table 3). The juice extracted using the CH showed a higher TSS content (18.20%) compared with extraction using a juice blender (JB) (17.78%) and juice extractor (JE) (16.65%). A possible cause of low TSS content within JB and JE could be due to the crushing of arils which resulted in dilution due to the addition of seed content such as oil into the juice. In a similar study, Mphahlele et al. (2016b) reported higher individual soluble sugar and total sugar contents in juice extracted from half fruits compared with blending of the arils. However, the TA value (2.17%) was found to be significantly higher from juice extracted using the JB (arils plus seeds crushed) compared with the rest of the other extraction methods (Table 3). The TA value of 2.17% found in our study is higher than those reported by Mphahlele et al. (2016b) who found a lower TA value (1.56%) in juice from half pomegranates (cv. Wonderful). The ratio of TSS to TA value is an essential criterion for assessing the taste of pomegranate juice. The highest TSS/TA ratio was found in juice extracted from CH (Table 3). The TSS/TA ratio for pomegranate juice is in the order of CH (half fruit) > JB (arils plus seed) > JE (arils). In addition, we observed a strong negative correlation (−0.961) between TSS/TA ratio and TA value suggesting that TSS/TA ratio is highly influenced by TA content and is responsible for the bitter or sour taste in juice. This observation reported within our study is in agreement with those reported within literature (Mena et al., 2011; Arendse et al., 2015; Mphahlele et al., 2016b).

To further explore the relationship between TSS and TA as a potential maturity indicator, BrimA index was calculated according to the study by Jordan et al. (2001). Similar to the TSS/TA ratio, the highest BrimA index was observed in juice extracted from CH (15.83). A strong positive correlation (0.871) and negative correlation (−0.707) were observed between BrimA index, TSS, and TA, respectively. Strong correlations were found between TSS and BrimA, and this relationship suggests that BrimA is mainly influenced by TSS. The pH value of pomegranate juice characterizes its acidic taste, the former having an inverse correlation (−0.915) with the latter (Table 4). Pomegranate juice extracted from CH had a considerably higher pH value (3.02) than JE (2.98) and JB (2.69). Considering that pH values indicate the acidity of a solution, as bacterial growth is highly influenced by the pH of a medium, it could be suggested that juice extracted using a juice blender can be stored longer as low pH has the ability to inhibit bacterial growth. Similar range in pH values (2.50–2.66) using a blender (arils plus seed) has been reported by Rajasekar et al. (2012). Cloudiness of pomegranate juice was significantly affected by the extraction method. The JB method had the highest cloudiness value (0.43) compared with JE (0.22) and CH (0.18) (Table 3). The higher cloudiness observed in juice from the juice blender may be attributed to the breaking down of small particles from the seeds and arils, when they are crushed. Furthermore, a significantly negative correlation was observed with cloudiness and color attributes suggesting that an increase in juice cloudiness has an undesirable effect on the visual color of processed juice. For the density of pomegranate juice, no significant differences were observed between the three extraction methods (Table 3).

Methods of extraction significantly (p < 0.05) influence the total phenolic concentration of the pomegranate juice (Table 5). Juice extracted from the JB had the highest phenolic concentration (3.04 g/L) than the rest of the extraction methods (Table 5). Total phenolic concentration is in the order of JB (3.04 GAE g/L) > CH (2.47 GAE g/L) > JE (1.87 GAE g/L). We presumed that the high phenolic concentration in JB may be attributed to the extraction of phenolic compounds from the arils and seeds. Contrary to our results, Mphahlele et al. (2016a) reported a higher total phenolic concentration in juice extracted from half fruit compared with crushing of the arils. Anthocyanin compounds are responsible for the desirable red color of pomegranate juices as well as many other red-colored fruit juices (Li et al., 2010; Fischer et al., 2013). In this study, pomegranate juice from all extraction methods showed abundant of total anthocyanin content. We observed significantly (p < 0.05) higher levels of anthocyanin in juice extracted using JB (43.67 mg/L). Kalt et al. (2000) observed higher levels of anthocyanin at a low pH level, which is consistent with the presence of the intensely colored flavylium cations compared with the quinonoidal pseudobase, and chalcone forms, which are pale or colorless. Therefore, it could be argued that the high anthocyanin content observed from the JB juice resulted from a low pH (2.69) contained within the juice. Another interesting relationship was the high negative correlation between anthocyanin concentration and color components (a^* = −0.778, b^* = −0.807, C^* = −0.784, h° = −0.779) (Table 4). This negative correlation between color components and color pigment anthocyanin may suggest that anthocyanin could be used as an indicator of juice color. A strong negative correlation between anthocyanin and raspberry juice was observed by Palonena and Weber (2019).

Significantly higher (p < 0.05) antioxidant activity measured by radical scavenging activity was found from juice extracted using the JE (0.29 mM AAE/ml), whereas the JB (0.16 mM AAE/ml) had the least (Table 5). Non-significant but strong correlation was observed between DPPH assay and total phenolic concentration (0.997), suggesting that the concentration of phenolic compound plays a role in the antioxidant properties of pomegranate juice (Bunea et al., 2011) (Table 4). A similar comparative study of homemade and commercial grape juice by Burin et al. (2010) reports a correlation coefficient of 0.957 between total phenolic concentration and DPPH antioxidant activity. FRAP assay measures the antioxidant potential in juice by reducing ferric iron (Fe³⁺) to ferrous iron (Fe²⁺) by antioxidants. In this study, JB (0.57 ± 0.02 mM trolox/ml) and CH (0.58 ± 0.02 mM trolox/ml) extraction method did not vary significantly (p > 0.05) in antioxidant activity (Table 5). The higher FRAP antioxidant activity was observed in both JB and CH, and this may be attributed to the effective extraction of phenolic compounds from the seeds of the arils or peel of the fruit during the extraction process. Furthermore, a high correlation was observed between FRAP and total phenolic concentration (0.776), suggesting that phenolic compound concentration plays a role in the antioxidant potential of pomegranate juice. Previous studies observed high correlations between juice phenolics and antioxidant capacity (Gil et al., 2000; Shwartz et al., 2009), evidently suggesting that phenolic compounds are the main antioxidants in pomegranate juice.

Preliminary assessment of MIR spectra was investigated using PCA to examine the effects of the extraction method on pomegranate juice quality. The application of several spectral filters such as Savitzky–Golay first derivative, SNV, and MSC did not improve the discrimination of the PCA scores plot. The examination of score plots from baseline-corrected spectra showed the effects of extraction on the spectral and chemical properties of pomegranate juice (Figure 1). By plotting all 150 data points, scores from the first two PC explained 63.2% (PC₁ = 47.8%, PC₂ = 15.4%) of the total variation within the dataset (Figure 1A). Examination of the PCA scores plot generated from the three extraction methods showed well-defined sample clusters, whereas the CH and JE extraction methods co-clustered with one another (Figure 1A). Furthermore, the PCA plot may have revealed possible grouping (JB) of the spectral and chemical dataset (Figure 1B). For instance, a high level of juice cloudiness was observed for JB extraction (Table 3). This observation showed that IR spectroscopy could be used to differentiate between different extraction methods based on spectral and chemical data despite its simplified approach. Therefore, to examine the effects of the extraction method on pomegranate juice quality, the spectral and chemical datasets were further subjected to OPLS-DA.

The average baseline-corrected absorbance spectra for pomegranate juice extracted from three different methods are shown in Figure 2. The MIR absorbance spectra for all three extraction methods showed similar contours and wavelength bands. Assessment of wave bands was performed according to literature (Bureau et al., 2019). The spectral bands displayed contours having prominent absorbance peaks in the following regions: 1,062, 1,150, 1,350, 1,420, 1,726, 2,940, 3,070, and 3,526 cm⁻¹. The prominent bands observed at the regions of 1,726 cm⁻¹ and 3,526–3,070 cm⁻¹ have known to correspond to the O–H stretch in the water. The characteristic band observed at 2,940 cm⁻¹ (corresponding to C–H₃) is indicative of the presence of color pigments, such as chlorophyll and flavonoids, as interpreted by Favaro et al. (2018). The IR region between 1,420 and 950 cm⁻¹ is usually referred to as the “fingerprint” region, as previously reported by Nieuwoudt et al. (2004). This region corresponds to the vibration of the C–O, C–C, C–H, and C–N bonds and provides information regarding organic compounds such as sugars, alcohols, and organic acids that are probably present in a sample (Favaro et al., 2018). The bands observed at 1,420–1,150 cm⁻¹ have been ascribed to C–H, O–H, and C–O functional groups that are present in organic acids, while the most intense bands located in the region of 1,000–1,100 cm⁻¹ have been reported to be due to the contribution of the C–O stretch related to carbohydrates emanating from sugars (Moreira and Santos, 2004; Musingarabwi et al., 2016). The spectral profile for pomegranate juice is similar to those reported by several researchers who evaluated the quality characteristics of apple/cranberry juice (Samborska et al., 2018), orange juice (Clark, 2016) and peach nectar (Coelho et al., 2020).

OPLSA-DA was applied on two different extraction methods at a time (pairwise), and the discrimination of these different methods was based on the spectral wave numbers and the associated chemistry (Supplementary Figures). By comparing the JB and JE, OPLS-DA was capable of providing 100% discrimination between these two methods (Supplementary Figure 1A). From pairwise OPLS-DA, the S-line plot revealed that the wave numbers responsible for discrimination between the methods were within the following regions: 3,201, 1,750–1,091 cm⁻¹ (Supplementary Figure 1B). These wave numbers have been associated with O–H bonds of water molecules and C=O, C–O, C–C bonds of various compounds (carbohydrates, organic acids, flavonoids) as interpreted by several authors (Sinelli et al., 2008; Arendse et al., 2018; Favaro et al., 2018). The PCA score plot revealed that the extraction method JE is associated with pH and various color attributes (C^*, a^*, b^*, h°), while the extraction method JB has been associated with TSS, cloudiness, and total phenolic content (Supplementary Figure 1C). The VIP plot for the chemistry suggests that color attributes and cloudiness contribute more to the separation of JB and JE extraction methods (Supplementary Figure 1D).

Likewise, by comparing the extraction method JB and CH, OPLS-DA was capable of providing 100% discrimination (Supplementary Figure 2A). The S-line plot (Supplementary Figure 2B) showed wavelength bands within the region of 3,200, 1,750, 1,691, 1,250, and 1,090 cm⁻¹. The predominant band in the region of 1,090 cm⁻¹ can be assigned to the C–O stretching vibration of the sugars. Similar bands in the region of 1,000–1,090 cm⁻¹ were reported by He et al. (2007), for the evaluation of sugars in different commercial juices. The VIP plot suggests that color attributes, TSS/TA ratio, cloudiness, and anthocyanin contribute toward the separation of these two methods (Supplementary Figure 2D). The pairwise comparison between CH and JE resulted in 100% classification between the two methods using OPLS-DA (Supplementary Figure 3A). From the pairwise OPLS-DA, the S-line plot revealed that the MIR wave numbers responsible for discrimination between CH and JE were similar to those observed in Supplementary Figure 2B. The PCA score plot revealed that the extraction method CH is associated with TSS, BrimA, and TSS/TA ratio, while the extraction method JE has been associated with TA, total anthocyanin content, and total phenolic content (Supplementary Figure 3C). The main contributors to the separation of the two extraction methods were TSS/TA ratio, TA, TSS, total anthocyanin concentration, and color attributes (Supplementary Figure 3D). Noticeably, a high correlation was observed between various color attributes and anthocyanin concentration within this study, suggesting that FT-IR spectroscopy can distinguish juice processing methods based on flavonoid color pigments, such as anthocyanins.

To further investigate the discriminating power of FT-IR spectroscopy, OPLS-DA was applied to all three different extraction methods, including the CH, JB, and JE. OPLS-DA correctly classified all three extraction methods with 100% accuracy (Figure 3A). The score plot representing all three extraction methods suggests that CH has been associated with the chemistry of TSS/TA ratio and BrimA (Figure 3B). The extraction method JE has been associated with total anthocyanin content suggesting higher red color, while JB has been observed to be associated with cloudiness and total phenolic content (Figure 3B). The VIP plot summarizing the chemistry responsible for model performance is shown in Figure 3C. When comparing all three methods, the VIP plot suggests that TSS contributes slightly more to separate the three extraction methods. Similarly, to the pairwise comparison, the VIP plot also illustrates that TSS/TA ratio and color components (L^*, a^*, C^*, h°) all contribute toward model performance. The results of this study suggest that juice extraction methods can clearly be distinguished from one another based on TSS, acidity, anthocyanin concentration, color attributes, and flavor-associated compounds.