Next Article in Journal
Sorption Behaviour of Ibuprofen Using Activated Carbon Derived from Rose Geranium (Pelargonium graveolens L.) Leaves
Previous Article in Journal
Order-Preserving Multiple Pattern Matching in Parallel
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Applied Sciences
Volume 13
Issue 8
10.3390/app13085143
applsci-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Prediction of Phytochemical Constituents in Cayenne Pepper Using MIR and NIR Spectroscopy

by
Joel B. Johnson
1,*,
Aimen El Orche
2,
Janice S. Mani
1,
Abderrahmane Aït-Kaddour
3,
Kerry B. Walsh
1 and
Mani Naiker
1
1
School of Health, Medical & Applied Sciences, Central Queensland University, Bruce Hwy, North Rockhampton, QLD 4701, Australia
2
Team of Analytical and Computational Chemistry, Nanotechnology and Environment, Faculty of Sciences and Techniques, University of Sultan Moulay Slimane, Beni Mellal 23000, Morocco
3
Université Clermont Auvergne, INRAE, VetAgro Sup, UMR sur le Fromage, UMRF, 63370 Lempdes, France
*
Author to whom correspondence should be addressed.
Appl. Sci. 2023, 13(8), 5143; https://doi.org/10.3390/app13085143
Submission received: 22 March 2023 / Revised: 17 April 2023 / Accepted: 18 April 2023 / Published: 20 April 2023
(This article belongs to the Special Issue Advanced Methods and Technologies: Ensuring and Improving Food Safety and Quality)
Browse Figures
Versions Notes

Abstract

:
The aim of the present study was to evaluate the potential of handheld near-infrared (NIR) and benchtop mid-infrared (MIR) spectroscopy for the rapid prediction of antioxidant capacity, dry matter, and total phenolic contents in cayenne pepper (Capsicum annuum ‘Cayenne’). Using NIR spectroscopy, the best-performing model for dry matter had an R2pred = 0.74, RMSEP = 0.38%, and RPD of 2.02, exceeding the best results previously reported in the literature. This was also the first study to predict dry matter content from the mid-infrared spectra, although with lower accuracy (R2pred = 0.54; RMSEP = 0.51%, RPD 1.51). The models for antioxidant capacity and total phenolic content did not perform well using NIR or MIR spectroscopy (RPD values < 1.5), indicating that further optimization is required in this area. Application of support vector regression (SVR) generally gave poorer results compared to partial least squares regression (PLSR). NIR spectroscopy may be useful for in-field measurement of dry matter in the chili crop as a proxy measure for fruit maturity. However, the lower accuracy of MIR spectroscopy is likely to limit its use in this crop.

1. Introduction

Chilies (Capsicum spp.), also known as peppers, are well known for their pungent flavor and are used to season a wide variety of foods. In addition, chilies contain high levels of ascorbic acid (vitamin C), vitamin A and other micronutrients including zinc, iodine and iron [1]. More details are shown in Table 1. The principal pungent compound present—capsaicin—is an activator of the transient receptor potential ion channel of the vanilloid type (TRPV1) and hence has been investigated for treating pain and other neurological conditions [2]. In addition, the capsaicinoids, phenolics and flavonoids found in the fruit provide antioxidant activity with further potential health benefits [3,4,5].
In 2014, the worldwide production of chili pepper (Capsicum annuum) was estimated at 409,000 tons, with the majority grown in Asian countries [1]. In Australia, the chili industry is a relatively minor player, with around 2500 tons produced in the 2013–2014 season [7]. However, the industry is worth over $21 million p.a. (2013–2014 figures) and is rapidly expanding [7]. Farmgate prices average approximately $9.50 per kg, with a 244% increase in value over the preceding five years [7]. Research and development efforts continue in the Australian chili sector, with interest in improving yield, developing high-capsaicin varieties, and developing new processing methods.
In this study, we investigate the potential use of rapidly assessing quality parameters associated with chili fruit—specifically dry matter and total phenolic content—in the Australian Cayenne pepper crop. Dry matter (DM) content is often linked to maturity and eating quality across many crops [8,9] and has also been used as an index of fruit maturity in chilies [10]. Indeed, Niklis et al. [11] and Kasampalis et al. [12] found that the DM content of C. annuum consistently increased during maturation, although the exact trends were dependent on the variety. Hence if calibrated for a specific chili variety, the rapid estimation of DM content could be used to inform optimum harvest times.
As previously mentioned, the phenolic content and antioxidant capacity of chilies are also of interest due to the potential health benefits associated with these compounds [13,14]. Rapid estimation of phenolic content in chilies could be used for grading fruit quality or finding applications in breeding programs attempting to develop high phenolic content chili varieties.
Near-infrared (NIR) and mid-infrared (MIR) spectroscopy have been used for the assessment of a wide range of quality parameters in dried chili powder [15,16,17,18,19], principally the capsaicinoid content. The main difference between these two instrument types is the wavelength used: NIR operates between wavelengths of 750–2500 nm, while MIR spectroscopy uses wavelengths of 2500–25,000 nm (4000–400 cm−1). NIR spectroscopy has benefits, including cheaper cost, portability, and greater penetration depth into the sample—but at the expense of lower resolution and sensitivity for chemical moieties. A number of studies have used NIR spectroscopy for the prediction of dry matter and soluble solids [20,21], as well as vitamin C, total chlorophyll, and/or carotenoid contents in fresh fruit [10,12,20,22]. These studies have highlighted the benefits of rapid NIR spectroscopy as an in situ preliminary screening technique but noted that further work is required to develop robust models which can be applied to external populations.
The prediction of antioxidant capacity appears to have only been considered by one previous author [23], and there are very few studies outside of our laboratory attempting the prediction of total phenolic content [6,12]. In addition, many of these studies have been performed on bell peppers rather than cayenne peppers, which are more widely grown in Australia. Hence the aim of this study was to compare the use of NIR and MIR spectroscopy for the analysis of cayenne pepper (Capsicum annuum ‘Cayenne’) quality, including the prediction of DM content, total phenolic content, and antioxidant capacity.

2. Materials and Methods

2.1. Chilli Samples

One hundred samples (individual chilies) of green cayenne pepper (“Blade” variety) were provided by Austchili (Bundaberg, Queensland), the largest supplier of fresh chili in Australia. These were hand-selected from plants across two field sites to ensure that a wide range of environmental variability was incorporated into the samples chosen. The majority of samples selected were at the mature stage, and all were a uniform dark green color. Across all of the samples, the average length was 117 ± 7 mm (range 92–133 mm), while the average fresh weight was 12.4 ± 1.9 g (range 7.5–16.2 g). The dry matter ranged from 8.4–12.6% w/w (mean of 10.5 ± 0.9% w/w).
The samples were shipped to the laboratory overnight on ice before the NIR spectra were collected and the samples subsequently dried. The MIR spectra were collected after the drying process.

2.2. Collection of NIR Spectra

Near-infrared (NIR) spectra were collected from the samples using a MicroNIR OnSite handheld spectrometer (Viavi; Santa Rosa, CA, USA), operating between 908–1676 nm (6 nm resolution). This instrument uses a linear variable filter as the dispersing element coupled to a linear detector array (128-pixel uncooled InGaAs photodiode array) [24]. The spectra integration time of the instrument was set to 100 ms.
Two spectra were collected from two opposing sides of each individual chili, for a total of four spectra per chili sample (n = 400 spectra in total). In order to ensure consistency in the spectral acquisition location and minimize the prospect of background interference, chili samples were laid on a ceramic plate, and all spectra were collected approximately 3 cm down from the stem of the chili. In addition, reference black and white spectra were obtained to calibrate the instrument for every 10 samples. Spectra were exported in ASCII (*.csv) format.

2.3. Dry Matter Content and Total Phenolic Content

Following the collection of the NIR spectra, the dry matter content of the chili samples was determined gravimetrically by drying the whole chilies (approx. 5–10 g) in a forced air oven (Sunbean Food Lab Dehydrator) at 50 °C until reaching a constant mass. The DM content was expressed on a % fresh weight basis.
The dried chili samples were subsequently ground using a Retsch ZM1000 centrifugal grinding mill (Sydney, Australia) with a 1.0 mm mesh size. Polar phenolic compounds were extracted from this dried powder using 90% methanol, following previously described protocols [25]. The total phenolic content (TPC) and antioxidant activity—measured as ferric-reducing antioxidant power (FRAP) and cupric-reducing antioxidant capacity (CUPRAC)—of the extracts were determined as previously reported [26], with the results expressed in terms of gallic acid equivalents (GAE) per 100 g and Trolox equivalents (TE) per 100 g, respectively. All three (TPC, FRAP, and CUPRAC) were expressed on a dry weight basis.

2.4. Collection of MIR Spectra

Mid-infrared spectra were collected from the dried, ground chili powder using a Bruker Alpha Fourier transform infrared (FTIR) spectrophotometer (Ettlingen, Germany). A single-reflection diamond crystal attenuated total reflectance (ATR) module was used to ensure firm contact between the sample and the infrared beam. The spectra were recorded between 4000–400 cm−1 in triplicate for each sample (n = 300 spectra), with each spectra being the average of 24 scans at 4 cm−1 resolution. Spectra were exported in Opus (*.0) format.

2.5. Data Processing and Analysis

Chemometric analysis of the NIR and MIR spectra was conducted using the Unscrambler X software (Camo Analytics; Oslo, Norway). Figures were drawn in the Unscrambler X and in R Studio running R 4.0.2 [27].
For the development of the predictive chemometric models, the data set was split into a calibration set and a validation set, comprising 80% and 20% of the sample set, respectively. Two regression methods were trialed—partial least squares regression (PLS-R) and support vector regression (SVR). PLS-R is a linear quantitative modeling algorithm that has been extensively used in spectral analysis and particularly in the infrared wavelength range (NIR and MIR). It is always used to avoid multiple collinearities by reducing a large number of independent variables into a new set of unrelated variables called latent variables (LVs). In our case, the X matrix containing the MIR and NIR spectra was used to build the PLS2 models. The response matrix Y contains four variables to be predicted, namely DM, TPC, FRAP, and CUPRAC. The number of latent variables chosen in this study is based on cross-validation using the leave-one-out cross-validation procedure [28]. Support vector analysis (SVR) is oriented from SVM (support vector machine). This method is very interesting because it can solve practical problems such as a small sample, nonlinearity, and high dimension. In this study, the SVM method was applied using linear kernel algorithms applied using nu-SVM.
Before calculating regression models, the NIR and MIR spectra were subjected to different methods of pre-processing, including standard normal variate (SNV), multiplicate scatter correction (MSC), and the 1st and 2nd derivatives using a Savitzky-Golay algorithm (10 pts/side). Both SNV and MSC are generally applied to spectral data to decrease the influence of light scattering and the baseline shift. The purpose of the derivatives was to reduce the influence of random noise, highlight the subtle frequency band shape, maintain relative frequency band intensity information, and improve the resolution of overlapping frequency bands (Jie et al. [29]). The Savitzky-Golay derivative uses a smoothing procedure that is useful to decrease random noise and increase the signal-to-noise ratio.
The accuracy of the models was evaluated by root mean square error of prediction (RMSEP), cross-validation (RMSECV), calibration (RMSEC), determination coefficients calibration (R2cal), cross-validation (R2CV) and prediction (R2pred) and the ratio of prediction to deviation (RPD). RPD was calculated by dividing the standard deviation of the sample set by the RMSEP. Nicolaï et al. [30] considered that an RPD of 3 indicated an excellent model; between 2.5–3 indicated a good model; between 2–2.5 indicated that coarse quantitative predictions are possible; while an RPD of 1.5–2 suggested that the model could discriminate between low and high values of the analyte.

3. Results

3.1. Dry Matter, TPC, and FRAP Content

The mean, SD (standard deviation), minimum, maximum, and CV (coefficient of variation) of the DM, TPC, FRAP, and CUPRAC of the pepper fruit are presented in Table 2. As shown in Table 2, the DM content of the chili samples ranged from 8.39–12.64%, with a mean DM content of 10.52%. As shown by the CV of 4.8%, the range of DM contents was not as wide as that reported by previous researchers [20,21]; however, it represented the typical range of DM content found in pepper fruits around the point of harvest [11].
The TPC ranged from 898–2617 mg GAE/100 g, a slightly larger range than that obtained by Toledo-Martín et al. [6] (919–1803 mg GAE/100 g), with a much higher CV (20.5%). The FRAP and CUPRAC showed a similarly wide variation, with CVs of 23.7 and 19.8%, respectively. The mean FRAP (296 mg TE/100 g) was slightly higher than that previously found in green pepper pulp by Sora et al. [31] (107 mg TE/100 g), although lower than the value of ~425 mg TE/100 g found by Qiao et al. [32]. The mean CUPRAC (1917 mg TE/100 g) was also higher than the range of CUPRAC values (882–1280 mg TE/100 g) found by Parnea et al. [33] in a range of pepper varieties.

3.2. NIR and MIR Spectra

Figure 1 shows the unprocessed NIR spectra of the chili samples. Peaks were located at approximately 1450, 1190, and 970 nm, similar to that reported by Penchaiya et al. [20]. These absorption bands can be attributed primarily to the OH 1st overtone, CH 2nd overtone stretch, and OH 2nd overtone, respectively [34,35].
The MIR spectra of the dried chili powder showed a greater number of peaks (Figure 2), with major peaks corresponding to OH from moisture (approx. 3300 cm−1), CH3, and CH2 groups such as those found in fatty acids (3000–2800 cm−1), C=O (1750 cm−1), amide and cellulose groups (1000 cm−1) [36,37].

3.3. Prediction of DM, TPC, FRAP, and CUPRAC Using NIR Spectroscopy

Table 3 summarizes the results obtained for the prediction of the four parameters of interest using NIR spectroscopy combined with PLS-R or SVR. As can be seen in this table, the pre-processing methods used had a significant impact on the quality of the PLS-R and SVR models.
Using the PLSR algorithm, NIR spectroscopy showed good results for the prediction of dry matter content, with most of the R2 values ranging between 0.67–0.79 and 0.27–0.74 (0.60–0.74 excluding the SG1 pre-processing) respectively for the calibration and prediction datasets, respectively. The relative error obtained varied between 0.40–0.51% and 0.38–0.48% for calibration and validation, respectively. The pre-processing method had a strong influence on the model quality, with the best model constructed using no pre-processing. In contrast, quite poor results were found using PLS regression for the prediction of TPC, FRAP, and CUPRAC contents, with RPD values <1.5 for all of these analytes.
On the other hand, the application of the SVR algorithm showed satisfactory results for the prediction of the four analytes of interest. The best predictive results for TPC exhibited an R2pred of 0.65 and an RMSEP of 464 mg/100 g, using the first derivative Savitsky-Golay pre-processing. For FRAP, the best model presented an R2pred of 0.71 and an RMSEP of 70 mg/100 g. For CUPRAC, the best predictive model gave an R2pred of 0.73 and RMSEP of 512 mg/100 g, while the best model for DM showed an R2pred of 0.96 and RMSEP of 0.63% (Figure 3). Those models were calculated after applying the second derivative with the Savitsky-Golay algorithm to the spectral data (Table 3).
As has been noted for PLSR, the application of spectral pre-processing had a significant effect on the prediction results obtained by SVR. Furthermore, the optimal pre-processing method differed depending on the targeted analyte.

3.4. Prediction of DM, TPC, FRAP, and CUPRAC Using MIR Spectroscopy

The investigation of the MIR spectral data by PLS regression revealed good calibration results for the four parameters of interest but poor predictive ability when applied to the validation dataset (Figure 4; Table 4). The best-performing model was for DM content, with an R2pred of 0.54, RMSEP of 0.51%, and RPD of 1.51. The other models showed R2pred of ≤0.2 and RPD values below 1.5. Again, the pre-processing method had a moderate impact on model accuracy.
The use of the SVR method showed strong potential for the estimation of the four parameters of interest, as shown by the high R2 values ranging from 0.86 to 0.99 for both calibration and validation. The best prediction for CUPRAC and FRAP was obtained by applying the second derivative Savitzky-Golay pre-processing method, with an R2pred of 0.99 and 0.96 and RMSEP of 478 and 96 mg/100 g, respectively. For TPC and DM, the best prediction results were obtained using SNV and MSC pre-processing, respectively. The R2pred was >0.86 for both analytes, with an RMSEP of 433 mg TE/100 g and 0.66%, respectively.

3.5. Prediction of DM, TPC, FRAP, and CUPRAC Contents Using Data Fusion of NIR and MIR Spectroscopy

Finally, data fusion of MIR and NIR spectra was investigated for the prediction of the four parameters of interest (Figure 5). The application of PLS regression showed a good ability for the prediction of DM, with an R2pred of 0.66 and RMSEP of just 0.47% (Table 5). This was found using no pre-processing, similar to the results previously observed for both the NIR and MIR models for this analyte. For the other parameters, the application of PLS regression did not allow the prediction of TPC, FRAP, and CUPRAC with suitable accuracy (RPD < 1.5 for all).
Application of SVR showed good results for all calibration models and moderately high R2pred values for the validation models, although interestingly poorer results for the prediction of DM content. Again, SNV pre-processing showed optimal results for the prediction of FRAP and CUPRAC, with R2pred of 0.77 and 0.55 and an RMSEP of 67 & 467 mg/100 g, respectively. The best model for TPC prediction gave an R2pred of 0.69 and RMSEP of 425 mg/100 g. However, the RPD values remained low for the application of SVR to the prediction set.

4. Discussion

The best-performing results for the prediction of DM content found in this study (R2pred = 0.74; RMSEP = 0.38%; RPD = 2.02) were considerably better than the corresponding cross-validation statistics reported by Sánchez et al. [21] for the prediction of DM in fresh bell pepper fruit (R2CV = 0.62; RSECV = 0.66%; RPD = 1.64). As accurate prediction results are more difficult to obtain than accurate cross-validation results (because the samples being predicted do not comprise part of the test set, the model has not encountered these spectra before), this indicates considerably better accuracy and robustness of the model. A comparison with the results of other studies is provided in Table 6.
A lower accuracy was found for the prediction of DM content using MIR spectroscopy (R2pred = 0.54; RMSEP = 0.51%; RPD = 1.51). As the MIR spectra were collected from the dried chili powder, this method could not have been predicting the DM content directly from the water in the sample. Rather, this prediction is more likely to be an indirect correlation between the water content and other matrix constituents (e.g., cellulose) that is intrinsic to the chili samples.
The best model for total phenolic content had an R2pred of 0.69, RMSEP of 425 mg/100 g, and RPD of 1.07 (using SVR on the combined NIR and MIR dataset). This was much poorer than the cross-validation results reported by Toledo-Martín et al. [6], who found a cross-validation (DW basis) for the prediction of total phenolics in 14 different pepper types. However, Toledo-Martín et al. [6] used a high-resolution laboratory-based NIR instrument (FOSS NIRSystems; 400–2500 nm) with the spectra collected from freeze-dried chili powder. Furthermore, an external validation set was used in the present study rather than just cross-validation against the calibration samples. This likely contributes further to the lack of predictive power for TP content and highlights the challenges of detecting phenolics in fresh chilies. It is likely that high levels of moisture content could interfere with and obscure the absorption bands of this analyte. In addition, Toledo-Martín et al. [6] found minimal loading of peaks in the short-wave NIR region in their TP content model, indicating that phenolic compounds may not show strong absorption across the range of wavelengths included in this study (908–1676 nm).
Similarly, the best models for FRAP and CUPRAC had an RMSEP of 46 mg/100 g (RPD = 1.17) and an RMSEP of 379 mg/100 g (RPD = 1.07), respectively. This indicated that the infrared spectroscopy methods used in this study were unable to accurately discriminate between high or low levels of antioxidant capacity of chili samples. Prediction of antioxidant capacity (as measured by either of these methods) does not appear to have been previously attempted in this crop in the literature; however, the DPPH method has been used [23]. While these authors reported some success, the results are not directly comparable to either the FRAP or CUPRAC assays.

5. Conclusions

As demonstrated in this study, the use of rapid handheld NIR instrumentation shows promise for the prediction of dry matter content in cayenne peppers. This could potentially be useful for measuring the ripeness of fruit in the field while the fruit is still on the plant. In turn, this could guide decisions on harvesting times, staffing requirements, and irrigation requirements. The model for predicting total phenolic content did not perform well, suggesting that further research in this field is required. Furthermore, the optimal spectra pre-processing strategy differed between various analytes, suggesting that it is difficult to know beforehand which pre-processing would lead to the most accurate predictive models. On the other hand, MIR spectroscopy showed limited accuracy in this crop, which is likely to limit its use—particularly as it also is confined to benchtop operation at present.

Author Contributions

Conceptualization, J.B.J., K.B.W. and M.N.; methodology, J.B.J. and K.B.W.; software, A.E.O.; validation, J.B.J. and A.E.O.; formal analysis, J.B.J. and A.E.O.; investigation, J.B.J.; resources, J.B.J., K.B.W., M.N. and A.E.O.; data curation, J.B.J.; writing—original draft preparation, J.B.J. and A.E.O.; writing—review and editing, J.B.J., A.E.O., J.S.M., A.A.-K., K.B.W. and M.N.; visualization, A.E.O.; supervision, M.N.; project administration, J.B.J.; funding acquisition, M.N. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported in part by a New Staff Grant (RSH/5343) awarded by CQUniversity to one of the authors (M.N.). One of the authors (J.B.J.) acknowledges support from the Australian Government in the form of a Research Training Program stipend.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The datasets supporting the findings of this paper are available from the authors upon request.

Acknowledgments

The authors gratefully acknowledge the assistance of Phul Subedi in operating the NIR instrument. We also thank Austchilli for supplying the chili samples used in this study.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Olatunji, T.L.; Afolayan, A.J. The suitability of chili pepper (Capsicum annuum L.) for alleviating human micronutrient dietary deficiencies: A review. Food Sci. Nutr. 2018, 6, 2239–2251. [Google Scholar] [CrossRef]
  2. Arora, V.; Campbell, J.N.; Chung, M.-K. Fight fire with fire: Neurobiology of capsaicin-induced analgesia for chronic pain. Pharmacol. Ther. 2021, 220, 107743. [Google Scholar] [CrossRef] [PubMed]
  3. Scarano, A.; Chieppa, M.; Santino, A. Looking at Flavonoid Biodiversity in Horticultural Crops: A Colored Mine with Nutritional Benefits. Plants 2018, 7, 98. [Google Scholar] [CrossRef] [PubMed]
  4. Hernández-Pérez, T.; Gómez-García, M.d.R.; Valverde, M.E.; Paredes-López, O. Capsicum annuum (hot pepper): An ancient Latin-American crop with outstanding bioactive compounds and nutraceutical potential. A review. Compr. Rev. Food Sci. Food Saf. 2020, 19, 2972–2993. [Google Scholar] [CrossRef] [PubMed]
  5. Bal, S.; Sharangi, A.B.; Upadhyay, T.K.; Khan, F.; Pandey, P.; Siddiqui, S.; Saeed, M.; Lee, H.-J.; Yadav, D.K. Biomedical and Antioxidant Potentialities in Chilli: Perspectives and Way Forward. Molecules 2022, 27, 6380. [Google Scholar] [CrossRef]
  6. Toledo-Martín, E.M.; García-García, M.C.; Font, R.; Moreno-Rojas, J.M.; Gómez, P.; Salinas-Navarro, M.; Del Río-Celestino, M. Application of visible/near-infrared reflectance spectroscopy for predicting internal and external quality in pepper. J. Sci. Food Agric. 2016, 96, 3114–3125. [Google Scholar] [CrossRef]
  7. AusVeg. Veggie Stats: Chillies; AusVeg: Melbourne, Australia, 2016; pp. 19–21. [Google Scholar]
  8. Anderson, N.T.; Subedi, P.P.; Walsh, K.B. Manipulation of mango fruit dry matter content to improve eating quality. Sci. Hortic. 2017, 226, 316–321. [Google Scholar] [CrossRef]
  9. Mickelbart, M.V.; James, D. Development of a dry matter maturity index for olive (Olea europaea). N. Z. J. Crop Hortic. Sci. 2003, 31, 269–276. [Google Scholar] [CrossRef]
  10. Ignat, T.; Alchanatis, V.; Schmilovitch, Z.E. Maturity prediction of intact bell peppers by sensor fusion. Comput. Electron. Agric. 2014, 104, 9–17. [Google Scholar] [CrossRef]
  11. Niklis, N.D.; Siomos, A.S.; Sfakiotakis, E.M. Ascorbic Acid, Soluble Solids and Dry Matter Content in Sweet Pepper Fruit: Change during Ripening. J. Veg. Crop Prod. 2002, 8, 41–51. [Google Scholar] [CrossRef]
  12. Kasampalis, D.S.; Tsouvaltzis, P.; Ntouros, K.; Gertsis, A.; Gitas, I.; Moshou, D.; Siomos, A.S. Nutritional composition changes in bell pepper as affected by the ripening stage of fruits at harvest or postharvest storage and assessed non-destructively. J. Sci. Food Agric. 2021, 102, 445–454. [Google Scholar] [CrossRef] [PubMed]
  13. Materska, M.; Perucka, I. Antioxidant Activity of the Main Phenolic Compounds Isolated from Hot Pepper Fruit (Capsicum annuum L.). J. Agric. Food Chem. 2005, 53, 1750–1756. [Google Scholar] [CrossRef] [PubMed]
  14. Hervert-Hernández, D.; Sáyago-Ayerdi, S.G.; Goñi, I. Bioactive Compounds of Four Hot Pepper Varieties (Capsicum annuum L.), Antioxidant Capacity, and Intestinal Bioaccessibility. J. Agric. Food Chem. 2010, 58, 3399–3406. [Google Scholar] [CrossRef] [PubMed]
  15. Lim, J.; Kim, G.; Mo, C.; Kim, M.S. Design and Fabrication of a Real-Time Measurement System for the Capsaicinoid Content of Korean Red Pepper (Capsicum annuum L.) Powder by Visible and Near-Infrared Spectroscopy. Sensors 2015, 15, 27420–27435. [Google Scholar] [CrossRef]
  16. San Park, T.; Min Bae, Y.; Jin Sim, M.; Eun Kim, D.; In Cho, S. Analysis of Capsaicinoids from Hot Red Pepper Powder by Near-Infrared Spectroscopy. In Proceedings of the American Society of Agricultural and Biological Engineers Annual International Meeting 2008, Providence, Rhode Island, 29 June–2 July 2008. [Google Scholar]
  17. Bae, M.-J.; Han, E.-S.; Hong, S.-H. Use of near Infrared Spectroscopy in Quality Control of Red Pepper Powder. J. Near Infrared Spectrosc. 1998, 6, A333–A335. [Google Scholar] [CrossRef]
  18. Bonifazi, G.; Gasbarrone, R.; Serranti, S. Dried red chili peppers pungency assessment by visible and near infrared spectroscopy. In Proceedings of the SPIE Defense + Commercial Sensing, Baltimore, MD, USA, 14 May 2019. [Google Scholar]
  19. Domínguez-Martínez, I.; Meza-Márquez, O.G.; Osorio-Revilla, G.; Proal-Nájera, J.; Gallardo-Velázquez, T. Determination of capsaicin, ascorbic acid, total phenolic compounds and antioxidant activity of Capsicum annuum L. var. serrano by mid infrared spectroscopy (Mid-FTIR) and chemometric analysis. J. Korean Soc. Appl. Biol. Chem. 2014, 57, 133–142. [Google Scholar] [CrossRef]
  20. Penchaiya, P.; Bobelyn, E.; Verlinden, B.E.; Nicolaï, B.M.; Saeys, W. Non-destructive measurement of firmness and soluble solids content in bell pepper using NIR spectroscopy. J. Food Eng. 2009, 94, 267–273. [Google Scholar] [CrossRef]
  21. Sánchez, M.-T.; Torres, I.; de la Haba, M.-J.; Chamorro, A.; Garrido-Varo, A.; Pérez-Marín, D. Rapid, simultaneous, and in situ authentication and quality assessment of intact bell peppers using near-infrared spectroscopy technology. J. Sci. Food Agric. 2019, 99, 1613–1622. [Google Scholar] [CrossRef]
  22. Ignat, T.; Schmilovitch, Z.; Fefoldi, J.; Steiner, B.; Alkalai-Tuvia, S. Non-destructive measurement of ascorbic acid content in bell peppers by VIS-NIR and SWIR spectrometry. Postharvest Biol. Technol. 2012, 74, 91–99. [Google Scholar] [CrossRef]
  23. Kusumiyati, K.; Putri, I.E.; Munawar, A.A.; Suhandy, D. A Data Fusion Model to Merge the Spectra Data of Intact and Powdered Cayenne Pepper for the Fast Inspection of Antioxidant Properties. Sustainability 2022, 14, 201. [Google Scholar] [CrossRef]
  24. Power, A.C.; Chapman, J.; Chandra, S.; Roberts, J.J.; Cozzolino, D. Illuminating the flesh of bone identification—An application of near infrared spectroscopy. Vib. Spectrosc. 2018, 98, 64–68. [Google Scholar] [CrossRef]
  25. Johnson, J.; Collins, T.; Skylas, D.; Quail, K.; Blanchard, C.; Naiker, M. Profiling the varietal antioxidative content and macrochemical composition in Australian faba beans (Vicia faba L.). Legume Sci. 2020, 2, e28. [Google Scholar] [CrossRef]
  26. Johnson, J.; Collins, T.; Power, A.; Chandra, S.; Portman, D.; Blanchard, C.; Naiker, M. Antioxidative properties and macrochemical composition of five commercial mungbean varieties in Australia. Legume Sci. 2020, 2, e27. [Google Scholar] [CrossRef]
  27. R Core Team. R: A Language and Environment for Statistical Computing, Version 4.0.2; R Foundation for Statistical Computing: Vienna, Austria, 2020. [Google Scholar]
  28. Johnson, J.B.; El Orche, A.; Naiker, M. Prediction of anthocyanin content and variety in plum extracts using ATR-FTIR spectroscopy and chemometrics. Vib. Spectrosc. 2022, 121, 103406. [Google Scholar] [CrossRef]
  29. Jie, Z.; Pan, R.; Gao, W.; Xu, B.; Wei, L. Automatic detection of layout of color yarns of yarn-dyed fabric. Part 3: Double-system-Mélange color fabrics. Color. Res. Appl. 2016, 41, 626–635. [Google Scholar]
  30. Nicolaï, B.M.; Beullens, K.; Bobelyn, E.; Peirs, A.; Saeys, W.; Theron, K.I.; Lammertyn, J. Nondestructive measurement of fruit and vegetable quality by means of NIR spectroscopy: A review. Postharvest Biol. Technol. 2007, 46, 99–118. [Google Scholar] [CrossRef]
  31. Sora, G.T.S.; Haminiuk, C.W.I.; da Silva, M.V.; Zielinski, A.A.F.; Gonçalves, G.A.; Bracht, A.; Peralta, R.M. A comparative study of the capsaicinoid and phenolic contents and in vitro antioxidant activities of the peppers of the genus Capsicum: An application of chemometrics. J. Food Sci. Technol. 2015, 52, 8086–8094. [Google Scholar] [CrossRef] [PubMed]
  32. Qiao, G.; Wenxin, D.; Zhigang, X.; Sami, R.; Khojah, E.; Amanullah, S. Antioxidant and anti-inflammatory capacities of pepper tissues. Ital. J. Food Sci. 2020, 32, 265–274. [Google Scholar] [CrossRef]
  33. Parnea, A.; Tulpan, R.; Vasiliu, I.; Moldovan, C.; Dumbravă, D.G. Assessment of Total Polyphenol Content and Antioxidant Capacity of Some Pepper Varieties. J. Agroaliment. Process. Technol 2018, 24, 286–288. [Google Scholar]
  34. Workman, J.; Weyer, L. Practical Guide to Interpretive Near-Infrared Spectroscopy; CRC Press: Boca Raton, FL, USA, 2007. [Google Scholar]
  35. Lapcharoensuk, R.; Danupattanin, K.; Kanjanapornprapa, C.; Inkawee, T. Combination of NIR spectroscopy and machine learning for monitoring chili sauce adulterated with ripened papaya. E3S Web Conf. 2020, 187, 04001. [Google Scholar] [CrossRef]
  36. Sigma Aldrich. IR Spectrum Table & Chart. Available online: https://www.sigmaaldrich.com/technical-documents/articles/biology/ir-spectrum-table.html (accessed on 31 March 2023).
  37. Pavia, D.L.; Lampman, G.M.; Kriz, G.S. Introduction to Spectroscopy: A Guide for Student of Organic Chemistry, 3rd ed.; Brooks/Cole: London, UK, 2001. [Google Scholar]
  38. Johnson, J.B.; Mani, J.S.; Walsh, K.B.; Naiker, M. Infrared spectroscopy for the quality assessment of Habanero chilli: A proof-of-concept study. J. Food Meas. Charact. 2023, 17, 1764–1774. [Google Scholar] [CrossRef]
  39. Kusumiyati, K.; Putri, I.E.; Hamdani, J.S.; Suhandy, D. Real-Time Detection of the Nutritional Compounds in Green ‘Ratuni UNPAD’ Cayenne Pepper. Horticulturae 2022, 8, 554. [Google Scholar] [CrossRef]
Applsci 13 05143 g001 550
Figure 1. The NIR spectra of the chili samples, between 908–1676 nm were obtained using the handheld microNIR probe.
Figure 1. The NIR spectra of the chili samples, between 908–1676 nm were obtained using the handheld microNIR probe.
Applsci 13 05143 g001
Applsci 13 05143 g002 550
Figure 2. The MIR spectra of the chili samples are between 4000–400 cm−1.
Figure 2. The MIR spectra of the chili samples are between 4000–400 cm−1.
Applsci 13 05143 g002
Applsci 13 05143 g003 550
Figure 3. Predicted vs. reference results for the application of NIR spectroscopy for the prediction of dry matter, TPC, FRAP, and CUPRAC contents.
Figure 3. Predicted vs. reference results for the application of NIR spectroscopy for the prediction of dry matter, TPC, FRAP, and CUPRAC contents.
Applsci 13 05143 g003
Applsci 13 05143 g004 550
Figure 4. Predicted vs. reference results for the application of MIR spectroscopy for the prediction of dry matter, TPC, FRAP, and CUPRAC contents.
Figure 4. Predicted vs. reference results for the application of MIR spectroscopy for the prediction of dry matter, TPC, FRAP, and CUPRAC contents.
Applsci 13 05143 g004
Applsci 13 05143 g005 550
Figure 5. Predicted vs. reference results for the fusion of NIR and MIR spectroscopy datasets for the prediction of dry matter, TPC, FRAP, and CUPRAC contents.
Figure 5. Predicted vs. reference results for the fusion of NIR and MIR spectroscopy datasets for the prediction of dry matter, TPC, FRAP, and CUPRAC contents.
Applsci 13 05143 g005
Table
Table 1. The typical content of key beneficial nutrients found in chilies.
Table 1. The typical content of key beneficial nutrients found in chilies.
ParameterTypical Content% RDI in 1 Serve (75 g)Reference
Total phenolic content919–1803 mg GAE/100 gn/a[6]
Vitamin C80.6 mg/100 g67%[1]
Niacin (Vit B3)0.48 mg/100 g2%[1]
Vitamin E (α-tocopherol)0.37 mg/100 g2%[1]
Vitamin B60.224 mg/100 g13%[1]
Pantothenic acid (Vit B5)0.099 mg/100 g1%[1]
Thiamin (Vit B1)0.057 mg/100 g4%[1]
Riboflavin (Vit B2)0.028 mg/100 g2%[1]
Vitamin A18 µg RAE/100 g2%[1]
Folate10 µg/100 g2%[1]
Vitamin K7.4 µg/100 g5%[1]
GAE = gallic acid equivalents; n/a = not applicable; RAE = retinoic acid equivalents; RDI = recommended dietary intake (calculated for an adult male).
Table
Table 2. Range, mean standard deviation, and coefficient of variation for the DM, TP, FRAP, and CUPRAC datasets.
Table 2. Range, mean standard deviation, and coefficient of variation for the DM, TP, FRAP, and CUPRAC datasets.
DM Content (%)TPC
(mg GAE/100 g)		FRAP
(mg TE/100 g)		CUPRAC
(mg TE/100 g)
Minimum8.398981681256
Maximum12.6426176153138
Mean10.5213002961917
SD0.8626770379
CV (%)4.8220.5423.6519.77
SD = Standard deviation; CV = coefficient of variation.
Table
Table 3. Statistical parameters were obtained from the application of the PLSR and SVR algorithms on the NIR spectra dataset using different pre-processing methods.
Table 3. Statistical parameters were obtained from the application of the PLSR and SVR algorithms on the NIR spectra dataset using different pre-processing methods.
Pre-ProcessingAnalyteLVs (PLS)R2calRMSECR2predRMSEPRPD
PLSSVRPLSSVRPLSSVRPLSSVRPLSSVR
NoneDM160.790.260.4030.770.740.000.3840.7482.0201.037
TPC140.330.17134.704150.0120.000.00466.55478.4570.9740.950
FRAP90.470.4447.09548.2190.000.5865.66968.3650.8250.793
CUPRAC120.510.3227.701272.7660.110.00397.912408.7151.0881.059
SG1DM160.670.840.5060.3560.270.000.6460.4071.2001.905
TPC160.570.53107.029116.6740.000.65451.12463.8161.0080.980
FRAP70.480.5846.48541.5990.000.6266.39672.9110.8160.743
CUPRAC80.520.64225.74196.8140.10.00410.437421.3141.0551.028
SG2DM140.740.960.4510.1780.700.530.4110.7371.8871.052
TPC140.330.85133.99864.1450.000.00458.849498.7950.9910.911
FRAP140.540.7843.61230.3170.000.1869.897109.9850.7750.493
CUPRAC160.550.94217.481.9610.160.73387.086512.4561.1180.845
SNVDM140.740.570.450.5920.600.000.4770.6951.6261.116
TPC130.330.29134.351142.7890.000.00463.32483.7410.9810.940
FRAP120.520.4644.61947.3030.000.0066.70668.7681.0641.065
CUPRAC110.510.42227.34247.690.10.68406.903406.5410.8120.133
SNV Detrend 1DM130.730.650.4530.5480.680.860.4280.6181.8121.255
TPC120.310.27136.107141.1950.000.00467.973480.9620.9710.945
FRAP120.530.544.40137.40.000.1968.356137.1170.7930.395
CUPRAC110.510.45226.828241.3110.10.00403.157398.0941.0741.087
SNV Detrend 2DM150.770.560.420.6060.650.960.4490.6271.7271.237
TPC110.350.28131.832142.2850.000.00456.291478.6660.9960.950
FRAP100.520.4744.59146.7960.000.7166.03669.6740.8210.778
CUPRAC50.440.4148.316250.3910.10.00410.62401.7671.0541.078
MSCDM140.760.580.4330.5860.690.920.4210.6031.8421.286
TPC100.310.26136.506142.9630.000.00466.013478.4570.9750.950
FRAP110.510.4645.22947.4760.000.5966.64271.410.8130.759
CUPRAC110.520.4226.31251.5030.10.00403.342400.8651.0731.080
Table
Table 4. Statistical parameters were obtained from the application of the PLSR and SVR algorithms on the MIR spectra dataset using different pre-processing methods.
Table 4. Statistical parameters were obtained from the application of the PLSR and SVR algorithms on the MIR spectra dataset using different pre-processing methods.
Pre-ProcessingAnalyteLVs (PLS)R2calRMSECR2predRMSEPRPD
PLSSVRPLSSVRPLSSVRPLSSVRPLSSVR
NoneDM120.510.560.6130.8630.540.080.5131.2321.5120.629
TPC120.580.84113.28975.9560.000.77482.832499.780.9420.910
FRAP120.650.8942.96226.140.000.17128.45180.5530.4220.300
CUPRAC120.880.84120.36148.3910.000.37519.759618.7330.8330.700
SG1DM120.520.990.61316.0430.400.840.5850.8151.3260.952
TPC120.710.9993.6073.4310.000.80471.339533.9460.9640.851
FRAP120.590.9946.7069.6910.000.6865.72170.7120.8240.766
CUPRAC120.820.99116.0770.0060.000.79491.5671.9990.8810.644
SG2DM110.580.990.5690.0010.380.670.5951.0481.3030.740
TPC110.830.9972.5820.2040.000.79474.649483.1970.9580.941
FRAP100.660.9942.2480.0780.200.9646.32395.7371.1700.566
CUPRAC40.870.99126.1910.3720.100.99404.672477.7691.0700.906
SNVDM30.490.990.6310.1020.320.880.6240.6851.2431.132
TPC140.860.9866.21723.5420.010.86440.574433.1421.0321.050
FRAP90.430.9954.6374.1620.000.5756.47977.3430.9590.701
CUPRAC50.570.99232.5535.3940.100.52405.221505.1851.0680.857
SNV Detrend 1DM120.50.980.6230.1260.300.870.6330.7421.2251.045
TPC150.860.9865.91342.8870.020.85438.021458.2811.0380.992
FRAP30.170.9970.4968.2930.000.5954.03682.3961.0030.658
CUPRAC50.550.98241.93844.3340.100.50401.332503.2931.0790.860
SNV Detrend 2DM120.490.980.6290.1270.290.870.6350.7181.2211.080
TPC150.880.9961.82118.80.040.86435.235446.3481.0441.018
FRAP60.250.9962.8696.9420.000.6655.66570.1990.9730.772
CUPRAC50.570.98233.64148.3740.100.88406.149522.2351.0660.829
MSCDM120.520.990.6130.0620.370.880.6010.6641.2901.168
TPC140.850.9985.68513.0290.000.85449.565446.8841.0111.017
FRAP60.250.9963.0984.6710.000.5854.32876.2340.9970.711
CUPRAC50.570.97233.59663.0310.100.42407.824504.0071.0620.859
Table
Table 5. Statistical parameters were obtained from the application of the PLSR and SVR algorithms on the combined NIR and MIR spectra datasets, using different pre-processing methods.
Table 5. Statistical parameters were obtained from the application of the PLSR and SVR algorithms on the combined NIR and MIR spectra datasets, using different pre-processing methods.
Pre-ProcessingAnalyteLVs (PLS)R2calRMSECR2predRMSEPRPD
PLSSVRPLSSVRPLSSVRPLSSVRPLSSVR
NoneDM160.690.930.4920.2450.600.120.4670.7341.6611.057
TPC90.380.8138.71589.1220.000.00444.826504.431.0220.901
FRAP80.460.7253.61348.1730.000.1072.6691550.7460.350
CUPRAC40.440.94264.69293.5140.200.10377.131463.9641.1480.933
SG1DM120.630.990.540.0050.400.460.580.8061.3370.962
TPC100.380.99138.3020.6380.020.00442.631519.0861.0270.876
FRAP110.450.9953.8113.9660.000.0268.73871.2690.7880.760
CUPRAC20.490.99253.3939.410.100.35378.035576.2961.1450.751
SG2DM120.610.980.5490.1380.490.220.5390.961.4390.808
TPC110.460.98128.64123.9980.030.00437.273482.0281.0400.943
FRAP90.380.96216.22814.4080.000.1065.00579.0260.8340.686
CUPRAC50.630.940.57685.7390.100.51400.408517.8781.0810.836
SNVDM110.570.82131.9450.4670.480.230.5461.1731.4200.661
TPC100.440.5760.258138.8050.020.00437.747451.9481.0381.006
FRAP70.310.77202.68536.3180.000.7769.79167.4450.7760.803
CUPRAC100.670.840.521147.9520.100.55404.968466.7721.0690.927
SNV Detrend 1DM120.650.92132.3090.2670.550.270.5050.9491.5360.817
TPC100.430.7446.03395.4370.030.69436.035424.8681.0431.070
FRAP140.60.86204.90329.0650.000.3568.73773.3350.7880.739
CUPRAC100.670.840.514164.8650.100.14406.269379.481.1411.066
SNV Detrend 2DM120.660.94131.0380.2150.610.360.4710.8351.6460.929
TPC100.440.7460.76995.6290.040.52433.938433.2121.0481.049
FRAP70.440.81203.34535.0550.000.169.28766.5160.7820.815
CUPRAC100.670.860.574138.5880.100.22400.107435.6551.0820.994
MSCDM100.580.82132.5350.4520.480.210.5441.0981.4260.706
TPC100.430.8160.28879.7880.020.00438.065440.9131.0381.031
FRAP70.310.74203.17839.9670.000.5669.61365.960.7780.821
CUPRAC100.670.86203.178142.060.100.00406.983453.3911.0640.955
Table
Table 6. Summary of the results for the prediction of DM, TP, FRAP, and CUPRAC content in green cayenne pepper and their comparison to previous studies on other chili varieties. Note: R2pred and RMSEP values are provided where reported; R2CV and RMSECV values are provided where a test set was not analyzed.
Table 6. Summary of the results for the prediction of DM, TP, FRAP, and CUPRAC content in green cayenne pepper and their comparison to previous studies on other chili varieties. Note: R2pred and RMSEP values are provided where reported; R2CV and RMSECV values are provided where a test set was not analyzed.
StudyChili TypeDM ContentTPCFRAPCUPRAC
NIR spectroscopy
Present studyGreen cayenneR2pred = 0.74,
RMSEP = 0.38%		R2pred = 0.65,
RMSEP = 464 mg/100 g		R2pred = 0.71
RMSEP = 70 mg/100 g		R2pred = 0.73,
RMSEP = 512 mg/100 g
Sánchez, et al. [21]Bell pepper (various)R2CV = 0.62,
RMSECV = 0.66%
Toledo-Martín, et al. [6]Bell pepper (various) R2CV = 0.71,
RMSECV = 104 mg/100 g
Ignat, et al. [10]Bell pepper (various)R2CV = 0.93,
RMSECV = 0.4
Kasampalis, et al. [12]Bell pepper (various)R2CV = 0.53,
RMSECV = 0.63%		R2CV = 0.67,
RMSECV = 44 mg/100 g
Johnson, et al. [38]HabaneroR2CV = 0.65,
RMSECV = 0.50%		R2CV = 0.21,
RMSECV = 162 mg/100 g		R2CV = 0.27,
RMSECV = 38 mg/100 g
Kusumiyati, et al. [39]Green cayenneR2pred = 0.85,
RMSEP = 0.61%
MIR spectroscopy
Present studyGreen cayenneR2pred = 0.54,
RMSEP = 0.51%		R2pred = 0.86,
RMSEP = 433 mg TE/100 g		R2pred = 0.96,
RMSEP = 96 mg/100 g		R2pred = 0.99,
RMSEP = 478 mg/100 g
Johnson, et al. [38]HabaneroR2CV = 0.28,
RMSECV = 0.86%		R2CV = 0.37,
RMSECV = 145 mg/100 g		R2CV = 0.40,
RMSECV = 35 mg/100 g
Domínguez-Martínez, et al. [19] 1Serrano R2CV = 0.94,
SEP = 0.02 mg GAE/100 g
1 Study was performed on the ground, freeze-dried powder.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Johnson, J.B.; El Orche, A.; Mani, J.S.; Aït-Kaddour, A.; Walsh, K.B.; Naiker, M. Prediction of Phytochemical Constituents in Cayenne Pepper Using MIR and NIR Spectroscopy. Appl. Sci. 2023, 13, 5143. https://doi.org/10.3390/app13085143

AMA Style

Johnson JB, El Orche A, Mani JS, Aït-Kaddour A, Walsh KB, Naiker M. Prediction of Phytochemical Constituents in Cayenne Pepper Using MIR and NIR Spectroscopy. Applied Sciences. 2023; 13(8):5143. https://doi.org/10.3390/app13085143

Chicago/Turabian Style

Johnson, Joel B., Aimen El Orche, Janice S. Mani, Abderrahmane Aït-Kaddour, Kerry B. Walsh, and Mani Naiker. 2023. "Prediction of Phytochemical Constituents in Cayenne Pepper Using MIR and NIR Spectroscopy" Applied Sciences 13, no. 8: 5143. https://doi.org/10.3390/app13085143

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop