Detection of the Inoculated Fermentation Process of Apo Pickle Based on a Colorimetric Sensor Array Method
Abstract
:1. Introduction
2. Materials and Methods
2.1. Materials
2.2. Screening of Starter Cultures
2.3. Apo Pickle Sample Preparation
2.4. Biochemistry of Apo Pickle
2.5. Organic Acid Detection of Apo Pickle
2.6. Volatile Organic Compounds Detection of Apo Pickle
2.7. Fabrication of the CSA
2.8. CSA Data Acquisition
2.9. Statistical Analysis
3. Results and Discussion
3.1. Screening of Starter Cultures
3.2. Biochemical Change during Fermentation
3.3. Change of Organic Acid during Fermentation
3.4. Analysis of Volatile Organic Compounds by the GC–MS
3.5. Analysis of Volatile Organic Compounds by the CSA
3.5.1. Classification Result of the PCA
3.5.2. Classification Result of the LDA
3.5.3. PLSR Analysis
4. Conclusions
Author Contributions
Funding
Data Availability Statement
Acknowledgments
Conflicts of Interest
References
- Zhang, F.; Tang, Y.; Ren, Y.; Yao, K.; He, Q.; Wan, Y.; Chi, Y. Microbial composition of spoiled industrial-scale Sichuan paocai and characteristics of the microorganisms responsible for paocai spoilage. Int. J. Food Microbiol. 2018, 275, 32–38. [Google Scholar] [CrossRef] [PubMed]
- Xiong, T.; Guan, Q.; Song, S.; Hao, M.; Xie, M. Dynamic changes of lactic acid bacteria flora during Chinese sauerkraut fermentation. Food Control 2012, 26, 178–181. [Google Scholar] [CrossRef]
- Lavefve, L.; Marasini, D.; Carbonero, F. Microbial Ecology of Fermented Vegetables and Non-Alcoholic Drinks and Current Knowledge on Their Impact on Human Health. Adv. Food Nutr. Res. 2019, 87, 147–185. [Google Scholar] [CrossRef]
- Lorn, D.; Nguyen, T.K.; Ho, P.H.; Tan, R.; Licandro, H.; Wache, Y. Screening of lactic acid bacteria for their potential use as aromatic starters in fermented vegetables. Int. J. Food Microbiol. 2021, 350, 109242. [Google Scholar] [CrossRef]
- Moon, S.H.; Kim, C.R.; Chang, H.C. Heterofermentative lactic acid bacteria as a starter culture to control kimchi fermentation. LWT 2018, 88, 181–188. [Google Scholar] [CrossRef]
- Zhao, D.; Ding, X. Studies on the low-salt Chinese potherb mustard (Brassica juncea, Coss.) pickle. I—The effect of a homofermentative l(+)-lactic acid producer Bacillus coagulans on starter culture in the low-salt Chinese potherb mustard pickle fermentation. LWT-Food Sci. Technol. 2008, 41, 474–482. [Google Scholar] [CrossRef]
- Erkmen, O.; Bozoglu, T.F. Fermented vegetables and fruits. In Food Microbiology Principles into Practice. Vol.2: Microorganisms in Food Preservation and Processing; Erkmen, O., Bozoglu, T.F., Eds.; John Wiley and Sons, Ltd.: Chichester, UK, 2016. [Google Scholar]
- Liang, H.; He, Z.; Wang, X.; Song, G.; Chen, H.; Lin, X.; Ji, C.; Li, S. Effects of salt concentration on microbial diversity and volatile compounds during suancai fermentation. Food Microbiol. 2020, 91, 103537. [Google Scholar] [CrossRef] [PubMed]
- Cui, Y.; Chen, T.; Li, B.; Liu, X.; Xia, J.; Han, J.; Wu, Y.; Yang, M. Colorimetric sensor array–smartphone–remote server coupling system for rapid detection of saccharides in beverages. J. Iran. Chem. Soc. 2018, 15, 1085–1095. [Google Scholar] [CrossRef]
- Jo, Y.; Chung, N.; Park, S.W.; Noh, B.S.; Jeong, Y.J.; Kwon, J.H. Application of E-tongue, E-nose, and MS-E-nose for discriminating aged vinegars based on taste and aroma profiles. Food Sci. Biotechnol. 2016, 25, 1313–1318. [Google Scholar] [CrossRef]
- Erkmen, O.; Erten, H.; Sağlam, H. Fermented Products Technology and Microbiology; Nobel Publication: Ankara, Turkey, 2020. [Google Scholar]
- Jia, M.; Pan, Y.; Zhou, J.; Zhang, M. Identification of Chinese teas by a colorimetric sensor array based on tea polyphenol induced indicator displacement assay. Food Chem. 2021, 335, 127566. [Google Scholar] [CrossRef]
- Sun, S.; Qian, S.; Zheng, J.; Li, Z.; Lin, H. A colorimetric sensor array for the discrimination of Chinese liquors. Analyst 2020, 145, 6968–6973. [Google Scholar] [CrossRef] [PubMed]
- Wei, Z.; Xiao, X.; Wang, J.; Wang, H. Identification of the Rice Wines with Different Marked Ages by Electronic Nose Coupled with Smartphone and Cloud Storage Platform. Sensors 2017, 17, 2500. [Google Scholar] [CrossRef] [Green Version]
- Johnson, B.J.; Malanoski, A.P.; Erickson, J.S. Development of a Colorimetric Sensor for Autonomous, Networked, Real-Time Application. Sensors 2020, 20, 5857. [Google Scholar] [CrossRef]
- Mazzone, P.J.; Hammel, J.; Dweik, R.; Na, J.; Czich, C.; Laskowski, D.; Mekhail, T. Diagnosis of lung cancer by the analysis of exhaled breath with a colorimetric sensor array. Thorax 2007, 62, 565–568. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Mazzone, P.J.; Wang, X.-F.; Xu, Y.; Tarek Mekhail, M.C.; Beukemann, J.N.; Kemling, J.W.; Suslick, K.S.; Sasidhar, M. Exhaled Breath Analysis with a Colorimetric Sensor Array for the Identification and Characterization of Lung Cancer. Thorac. Oncol. 2012, 7, 137–142. [Google Scholar] [CrossRef] [Green Version]
- Huang, L.; Wang, M.; Liu, H. Identification of Adulterated Extra Virgin Olive Oil by Colorimetric Sensor Array. Food Anal. Methods 2021, 15, 647–657. [Google Scholar] [CrossRef]
- Kaeppler, K. Crossmodal Associations between Olfaction and Vision: Color and Shape Visualizations of Odors. Chemosens. Percept. 2018, 11, 95–111. [Google Scholar] [CrossRef]
- Li, H.; Zhang, B.; Hu, W.; Liu, Y.; Dong, C.; Chen, Q. Monitoring black tea fermentation using a colorimetric sensor array-based artificial olfaction system. J. Food Process. Preserv. 2018, 42, e13348. [Google Scholar] [CrossRef]
- Chen, Q.; Liu, A.; Zhao, J.; Ouyang, Q.; Sun, Z.; Huang, L. Monitoring vinegar acetic fermentation using a colorimetric sensor array. Sens. Actuators B Chem. 2013, 183, 608–616. [Google Scholar] [CrossRef]
- Dai, C.; Huang, X.; Huang, D.; Lv, R.; Sun, J.; Zhang, Z.; Aheto, J.H. Real-time detection of saponin content during the fermentation process of Tremella aurantialba using a homemade artificial olfaction system. J. Food Process Eng. 2019, 42, e13101. [Google Scholar] [CrossRef]
- Song, G.; He, Z.; Wang, X.; Zhao, M.; Cao, X.; Lin, X.; Ji, C.; Zhang, S.; Liang, H. Improving the quality of Suancai by inoculating with Lactobacillus plantarum and Pediococcus pentosaceus. Food Res. Int. 2021, 148, 110581. [Google Scholar] [CrossRef] [PubMed]
- Choi, Y.-J.; Yong, S.; Lee, M.J.; Park, S.J.; Yun, Y.-R.; Park, S.-H.; Lee, M.-A. Changes in volatile and non-volatile compounds of model kimchi through fermentation by lactic acid bacteria. LWT 2019, 105, 118–126. [Google Scholar] [CrossRef]
- Liu, H.; Zhang, Y.; Huang, L.; Wang, M. A colorimetric gas-sensitive array sensor using filter paper for the analysis of fish freshness. Food Chem. 2022, 377, 132029. [Google Scholar] [CrossRef] [PubMed]
- Kim, S.Y.; Ko, J.A.; Kang, B.S.; Park, H.J. Prediction of key aroma development in coffees roasted to different degrees by colorimetric sensor array. Food Chem. 2018, 240, 808–816. [Google Scholar] [CrossRef] [PubMed]
- Wang, D.; Chen, G.; Tang, Y.; Li, H.; Shen, W.; Wang, M.; Liu, S.; Qin, W.; Zhang, Q. Effects of temperature on paocai bacterial succession revealed by culture-dependent and culture-independent methods. Int. J. Food Microbiol. 2020, 317, 108463. [Google Scholar] [CrossRef]
- Lee, M.E.; Jang, J.Y.; Lee, J.H.; Park, H.W.; Choi, H.J.; Kim, T.W. Starter cultures for kimchi fermentation. J. Microbiol. Biotechnol. 2015, 25, 559–568. [Google Scholar] [CrossRef]
- Xu, W.; Jiang, H.; Liu, T.; He, Y.; Chen, Q. Qualitative discrimination of yeast fermentation stages based on an olfactory visualization sensor system integrated with a pattern recognition algorithm. Anal. Methods 2019, 11, 3294–3300. [Google Scholar] [CrossRef]
- Chen, Y.; Wu, C.; Xu, W.; Lu, Z.; Fu, R.; He, X.; Ma, Z.; Zhang, H. Evaluation of degradation capability of nitrite and biogenic amines of lactic acid bacteria isolated from pickles and potential in sausage fermentation. J. Food Process. Preserv. 2021, 46, e16141. [Google Scholar] [CrossRef]
- Kim, S.-H.; Kang, K.H.; Kim, S.H.; Lee, S.; Lee, S.-H.; Ha, E.-S.; Sung, N.-J.; Kim, J.G.; Chung, M.J. Lactic acid bacteria directly degrade N-nitrosodimethylamine and increase the nitrite-scavenging ability in kimchi. Food Control 2017, 71, 101–109. [Google Scholar] [CrossRef]
- Liu, A.; Li, X.; Pu, B.; Ao, X.; Zhou, K.; He, L.; Chen, S.; Liu, S. Use of Psychrotolerant Lactic Acid Bacteria (Lactobacillus spp. and Leuconostoc spp.) Isolated from Chinese Traditional Paocai for the Quality Improvement of Paocai Products. J. Agric. Food Chem. 2017, 65, 2580–2587. [Google Scholar] [CrossRef]
- Zhao, Y.; Wu, Z.; Miyao, S.; Zhang, W. Unraveling the flavor profile and microbial roles during industrial Sichuan radish paocai fermentation by molecular sensory science and metatranscriptomics. Food Biosci. 2022, 48, 101815. [Google Scholar] [CrossRef]
- Moon, E.W.; Yang, J.S.; Yoon, S.R.; Ha, J.H. Application of colorimetric indicators to predict the fermentation stage of kimchi. J. Food Sci. 2020, 85, 4170–4179. [Google Scholar] [CrossRef] [PubMed]
- Baek, S.; Maruthupandy, M.; Lee, K.; Kim, D.; Seo, J. Freshness indicator for monitoring changes in quality of packaged kimchi during storage. Food Packag. Shelf Life 2020, 25, 100528. [Google Scholar] [CrossRef]
- Liu, Z.; Xiao, M.; Xu, Y.; Li, D.; Zhu, W.; Huang, T.; Peng, F.; Guan, Q.; Peng, Z.; Xie, M.; et al. Effect of homo- and hetero-fermentative lactic acid bacteria on physicochemical properties, amino acid, and volatile flavor compounds during paocai fermentation by pure culture. J. Food Process. Preserv. 2022, e17052. [Google Scholar] [CrossRef]
- Li, Z.; Xie, S.; Sun, B.; Zhang, Y.; Liu, K.; Liu, L. Effect of KCl replacement of NaCl on fermentation kinetics, organic acids and sensory quality of sauerkraut from Northeast China. J. Food Process. Preserv. 2022, 46, e16622. [Google Scholar] [CrossRef]
- Gerardi, C.; Tristezza, M.; Giordano, L.; Rampino, P.; Perrotta, C.; Baruzzi, F.; Capozzi, V.; Mita, G.; Grieco, F. Exploitation of Prunus mahaleb fruit by fermentation with selected strains of Lactobacillus plantarum and Saccharomyces cerevisiae. Food Microbiol. 2019, 84, 103262. [Google Scholar] [CrossRef]
- Paweł, S.; Magdalena, S.; Szymon, S.; Piechowicz, W. Chemical composition and sensory quality of sauerkraut produced from different cabbage varieties. LWT 2021, 136 Pt 1, 110325. [Google Scholar] [CrossRef]
- Palani, K.; Harbaum-Piayda, B.; Meske, D.; Keppler, J.K.; Bockelmann, W.; Heller, K.J.; Schwarz, K. Influence of fermentation on glucosinolates and glucobrassicin degradation products in sauerkraut. Food Chem. 2016, 190, 755–762. [Google Scholar] [CrossRef]
- Peng, Q.; Zheng, H.; Meng, K.; Zhu, Y.; Zhu, W.; Zhu, H.; Shen, C.; Fu, J.; Elsheery, N.L.; Xie, G.; et al. The way of Qu-making significantly affected the volatile flavor compounds in Huangjiu (Chinese rice wine) during different brewing stages. Food Sci. Nutr. 2022, 10, 2255–2270. [Google Scholar] [CrossRef] [PubMed]
- Sheng, J.; Shan, C.; Liu, Y.; Zhang, P.; Li, J.; Cai, W.; Tang, F. Comparative evaluation of the quality of red globe grape juice fermented by Lactobacillus acidophilus and Lactobacillus plantarum. Int. J. Food Sci. Technol. 2022, 57, 2235–2248. [Google Scholar] [CrossRef]
- Wieczorek, M.N.; Drabińska, N. Flavour Generation during Lactic Acid Fermentation of Brassica Vegetables—Literature Review. Appl. Sci. 2022, 12, 5598. [Google Scholar] [CrossRef]
- Hong, S.P.; Lee, E.J.; Kim, Y.H.; Ahn, D.U. Effect of Fermentation Temperature on the Volatile Composition of Kimchi. J. Food Sci. 2016, 81, C2623–C2629. [Google Scholar] [CrossRef]
- Zhu, C.; Deng, J.; Jiang, H. Parameter Optimization of Support Vector Machine to Improve the Predictive Performance for Determination of Aflatoxin B1 in Peanuts by Olfactory Visualization Technique. Molecules 2022, 27, 6730. [Google Scholar] [CrossRef]
- Ouyang, Q.; Zhao, J.; Chen, Q.; Lin, H. Classification of rice wine according to different marked ages using a novel artificial olfactory technique based on colorimetric sensor array. Food Chem. 2013, 138, 1320–1324. [Google Scholar] [CrossRef] [PubMed]
- Salinas, Y.; Ros-Lis, J.V.; Vivancos, J.L.; Martinez-Manez, R.; Marcos, M.D.; Aucejo, S.; Herranz, N.; Lorente, I. Monitoring of chicken meat freshness by means of a colorimetric sensor array. Analyst 2012, 137, 3635–3643. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Wang, Y.; Huang, X.; Aheto, J.; Ren, Y.; Zhang, X.; Wang, L. Novel colorimetric sensor array for Chinese rice wine evaluation based on color reactions of flavor compounds. Food Process Eng. 2021, 44, e13889. [Google Scholar] [CrossRef]
- Zaragoza, P.; Fuentes, A.; Fernandez-Segovia, I.; Vivancos, J.L.; Rizo, A.; Ros-Lis, J.V.; Barat, J.M.; Martinez-Manez, R. Evaluation of sea bream (Sparus aurata) shelf life using an optoelectronic nose. Food Chem. 2013, 138, 1374–1380. [Google Scholar] [CrossRef] [PubMed]
Strain Code | Species |
---|---|
LP-046 | Lactobacillus plantarum |
LP-567 | Lactobacillus plantarum |
LP-165 | Lactobacillus plantarum |
LS-620 | Lactobacillus sakei |
LM-216 | Leuconostoc mesenteroides |
LB-373 | Lactobacillus brevis |
LZ-395 | Levilactobacillus zymae |
Number | Gas-Sensitive Material |
---|---|
1 | Bromocresol green b |
2 | Methyl Red sodium salt b |
3 | Bromophenol blue b |
4 | Chlorphenol red b |
5 | Bromothymol blue sodium b |
6 | Bromcresol purple a |
7 | Bromophenol blue c |
8 | Chlorphenol red c |
9 | Bromcresol purple c |
Time (d) | Lactic Acid (mg/mL) | Acetic Acid (mg/mL) | Citric Acid (mg/mL) | Succinic Acid (mg/mL) | Fumaric Acid (mg/mL) |
---|---|---|---|---|---|
0 | 2.17 ± 0.20 f | 2.26 ± 0.31 d | 0.43 ± 0.06 b | 0.06 ± 0.02 ab | 0.01 ± 0.00 a |
1 | 2.56 ± 0.26 d | 3.78 ± 0.04 a | 0.55 ± 0.03 a | 0.05 ± 0.01 b | 0.01 ± 0.00 a |
2 | 3.64 ± 0.07 c | 3.08 ± 0.11 b | 0.53 ± 0.01 a | 0.07 ± 0.00 a | - |
3 | 3.87 ± 0.00 bc | 3.30 ± 0.10 bc | 0.57 ± 0.02 a | 0.07 ± 0.01 ab | - |
5 | 3.84 ± 0.11 bc | 3.20 ± 0.04 bc | - | 0.07 ± 0.01 a | - |
7 | 4.17 ± 0.18 a | 3.04 ± 0.15 bc | - | 0.06 ± 0.00 ab | - |
11 | 4.01 ± 0.05 ab | 2.96 ± 0.20 c | - | 0.06 ± 0.20 ab | - |
15 | 3.55 ± 0.14 c | 2.93 ± 0.07 c | - | 0.06 ± 0.00 ab | - |
Time (d) | 0 | 1 | 2 | 3 | 5 | 7 | 11 | 15 | Total Accuracy (%) |
---|---|---|---|---|---|---|---|---|---|
Backtracking validation (%) | 100 | 100 | 100 | 100 | 91.7 | 100 | 100 | 100 | 99.0 |
Cross validation (%) | 100 | 100 | 83.3 | 100 | 75.0 | 83.3 | 100 | 100 | 92.7 |
Function | Eigenvalues | Percentage of Variance (%) | Cumulative Percentage (%) |
---|---|---|---|
1 | 16.992 | 43.3 | 43.3 |
2 | 8.211 | 20.9 | 64.2 |
3 | 6.197 | 15.8 | 80.0 |
4 | 3.969 | 10.1 | 90.1 |
5 | 1.638 | 4.2 | 94.3 |
6 | 1.316 | 3.4 | 97.7 |
7 | 0.919 | 2.3 | 100.0 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Wang, M.; Liu, J.; Huang, L.; Liu, H. Detection of the Inoculated Fermentation Process of Apo Pickle Based on a Colorimetric Sensor Array Method. Foods 2022, 11, 3577. https://doi.org/10.3390/foods11223577
Wang M, Liu J, Huang L, Liu H. Detection of the Inoculated Fermentation Process of Apo Pickle Based on a Colorimetric Sensor Array Method. Foods. 2022; 11(22):3577. https://doi.org/10.3390/foods11223577
Chicago/Turabian StyleWang, Mengyao, Jiawei Liu, Lu Huang, and Haiying Liu. 2022. "Detection of the Inoculated Fermentation Process of Apo Pickle Based on a Colorimetric Sensor Array Method" Foods 11, no. 22: 3577. https://doi.org/10.3390/foods11223577
APA StyleWang, M., Liu, J., Huang, L., & Liu, H. (2022). Detection of the Inoculated Fermentation Process of Apo Pickle Based on a Colorimetric Sensor Array Method. Foods, 11(22), 3577. https://doi.org/10.3390/foods11223577