Skip to main content
NCBI home page
Scientific Reports logoLink to Scientific Reports
. 2024 Nov 7;14:27080. doi: 10.1038/s41598-024-78426-y

Regression prediction of tobacco chemical components during curing based on color quantification and machine learning

Yang Meng 1,#, Qiang Xu 1,#, Guangqing Chen 2, Jianjun Liu 2, Shuoye Zhou 2, Yanling Zhang 1, Aiguo Wang 1, Jianwei Wang 1, Ding Yan 3, Xianjie Cai 3, Junying Li 4, Xuchu Chen 4, Qiuying Li 5, Qiang Zeng 5, Weimin Guo 1,7,8,✉, Yuanhui Wang 6,7,8,✉
PMCID: PMC11543802  PMID: 39511398

Abstract

Color is one of the most important indicators to characteristic the quality of tobacco, which is strongly related to the variations of chemical components. In order to clarify the relationship between the changes of tobacco color and chemical components, here we established several prediction models of chemical components with the color values of tobacco based on machine learning algorithms. The results of correlation analysis showed that tobacco moisture content was highly significantly correlated with the parameters such as a*, H* and H°, the reducing sugar and total sugar content of tobacco was significantly correlated with the color values, and the starch content was highly significantly correlated with the color values except for b* and C*. The random forest models performed best in predicting tobacco moisture, reducing sugar, total sugar and starch constructed with the R2 of the model validation set was higher than 0.90, and the RPD value was greater than 2.0. The consistent between the predictions and measurements verified the availability and feasibility using color values to predict some chemical components of the tobacco leaves with high accuracy, and which has distinct advantages and potential application to realize the real-time monitoring of some chemical components in the tobacco curing process.

Keywords: Tobacco curing, Color values, Chemical components, Machine learning algorithm, Prediction model

Subject terms: Biochemistry, Biotechnology

Introduction

The change of color in plants, particularly in tobacco leaves, holds significant importance as it serves as a visual indicator of their growth, development, and maturity stages1. This phenomenon is underpinned by intricate biological processes involving protein regulation, material metabolism, and interactions among various molecules within the plant tissues2. Understanding these color changes is crucial for assessing the optimal harvest time, predicting the quality of tobacco products, and ensuring efficient curing processes3.

Tobacco curing is a crucial step in tobacco production. It is not only related to the physical drying of tobacco leaves, but also involves complex biochemical reactions and internal material transformation, which directly affects the final quality and commercial value of tobacco leaves4. The color change is the most obvious and the most easily observed characteristic during the process of tobacco curing. Three main categories of chemical components affect the color of flue-cured tobacco leaves. The first category is the pigments such as carotene and lutein in tobacco. The second category is the polyphenol conversion of enzymatic browning reaction. The appearance of tobacco color from green to yellow indicates that the enzymatic process has been completed5. The third category is the complex produced by non-enzymatic browning reaction6.

The physiological and biochemical reactions during tobacco curing cover multiple levels, including moisture evaporation, carbohydrate metabolism, pigment conversion, protein degradation, aroma production, and redox reactions. These processes together shape the final quality of tobacco leaves7. Pigments play a crucial role in determining the color and appearance of tobacco leaves during the curing process8. Throughout the curing process, the degradation of chlorophyll and the concurrent increase in carotenoids represent a pivotal shift in pigment components that significantly impacts the color of the leaves9. In addition to pigments such as chlorophylls, carotenoids, and anthocyanins that directly affect visible color, polyphenols, particularly those involved in oxidation reactions, also indirectly contribute to the color development and overall appearance of cured tobacco leaves10. Non-enzymatic browning reaction is also known as Maillard reaction. It gives special aroma to tobacco leaves, optimizes the aroma quality of tobacco leaves, and changes the color of tobacco leaves11. Some conventional compounds, such as reducing sugars and starch, are important substrates for biochemical reactions and have a significant effect on color. Meanwhile, the temperature and time of curing also affect the color of tobacco leaves12,13.

With the development of precision and intelligence in tobacco curing, some tobacco workers use spectroscopy to predict the chemical components content of tobacco14,15. Spectrometers are costly compared with spectrophotometers and difficult to adapt to production needs16. Machine learning algorithms can better analyze and summarize all kinds of complex information scientifically and efficiently, and are widely used to deal with linear and nonlinear problems17,18. Machine learning algorithms have shown significant applications in aspects of food quality control, food safety, and predictive analytics19,20. However, few studies reported the use of machine learning algorithms to predict the chemical components of tobacco based on color quantification.

The primary objectives of this study are to (i) clarify the color values closed related to the chemical components of tobacco leaves during curing process, (ii) establish and validate the predicted models of chemical components of tobacco leaves based on machine learning algorithms, (iii) provide theoretical basis and method reference for intelligent monitoring of tobacco curing status and improving the quality of tobacco curing.

Materials and methods

Plant materials

The tobacco variety used in this study is K326, which was cultivated in Nanping City, located within the southeastern Chinese province of Fujian. To ensure the reliability and representativeness of the samples, a meticulous selection process was carried out on the tobacco plants. Only those exhibiting uniform growth patterns, closely matching leaf colors, similar leaf sizes, and consistent field quality were chosen as experimental subjects. The middle and upper leaves were taken and conventionally cured in the tobacco baking room. The curing process was meticulously managed and divided into three distinct stages: yellowing stage, color fixing stage and dry tendon stage21,22. Samples were taken at the time points of 0 h, 24 h, 48 h, 60 h, 72 h, 84 h, 96 h, 108 h, 120 h and 132 h in the curing process. All time points were taken 10 times from the upper, middle and lower layers of the baking room. Nine representative intact tobacco leaves were selected at the sampling port of the same layer each time, and each three leaves were taken as a sample. In summary, three samples were taken at each sampling time point in the same layers. A total of 180 samples were taken.

The samples were first used to determine the color values and the moisture content of tobacco leaves., and then the main veins of tobacco leaves were removed. The leaves samples were frozen at -20 °C and lyophilized using freeze dryer (FreeZone2.5Plus, Labconco, USA), and the milled sample powder was used to determine the chemical components. The stationary phase and mobile phase used for the determination of pigments and polyphenols were of chromatographic grade, and the remaining reagents were of analytical grade. All reagents were purchased from Kemiou Chemical Reagent Co., Ltd, Tianjin, China.

Analysis of samples

Determination of color values

During the curing process, using a portable spectrophotometer (Ci64, X-rite, USA) to determine the lightness value (L*), greenness/redness value (a*), blueness/yellowness value (b*), color ratio (H*), hue (H°) and saturation (C*) of tobacco leaves. Each tobacco leaf was determined at six detection points according to the previously reported method23.

Determination of pigment content

Sample (2 g) in a conical flask (50 mL), added 25 mL 90% acetone, and ultrasound for 20 min. About 2 mL of the mixture was taken from the flask and filtered through a 0.45 μm organic membrane, and the filtrate was collected in a high performance liquid chromatography (HPLC) vial. The contents of lutein and β-carotene in the filtrate were determined by HPLC24.

Determination conditions of HPLC: The separation of pigments was achieved using a reversed-phase C18 column with a particle size of 4 μm and dimensions of 3.9 mm internal diameter by 150 mm in length. The mobile phase consisted of (A) isopropanol and (B) acetonitrile at an 80% concentration, applied with a gradient elution technique at a flow rate of 1.5 mL/min. The optimal gradient program was set as follows: 100% B from 0 to 40 min, followed by 100% A from 40 to 46 min. The column was maintained at a temperature of 30 °C, and the sample injection volume was 10 µL. Detection was performed at a wavelength of 448 nm.

Determination of polyphenol content

Sample (0.1 g) in a conical flask (50 mL), added 20 mL 50% methanol, and ultrasound for 20 min. About 2 mL of the mixture is then removed from the flask and filtered through a 0.45 μm hydrophilic membrane. The filtrate is collected in an HPLC vial. The content of polyphenol content in the filtrate were determined by HPLC25,26.

Determination conditions of HPLC: The separation of polyphenols was conducted on a reversed-phase C18 column featuring a particle size of 5 μm, an internal diameter of 4.6 mm, and a length of 250 mm. The mobile phase used was composed of two solutions: (A) a mixture of water, methanol, and acetic acid in a ratio of 10:88:2 (v/v/v), and (B) a mixture of water, methanol, and acetic acid in a ratio of 88:10:2 (v/v/v). Gradient elution was employed at a flow rate of 1 mL/min. The optimal gradient program was as follows: 100% A from 0 to 16.5 min, a transition to 80% A and 20% B from 16.5 to 30 min, and finally 20% A and 80% B from 30 to 40 min. The column was kept at a constant temperature of 30 °C, and the injection volume was 10 µL. Detection was carried out at a wavelength of 340 nm.

Determination of moisture content and conventional chemical components

The moisture content of wet basis was determined by drying method. The tobacco leaves were placed in an oven (DHG-9140 A, China), and 10 tobacco leaves were dried at 105 °C ± 3 °C for at least 6 h to determine the moisture content of the tobacco leaves27.

The conventional chemical components are determined by a continuous flow meter (AA3, SEAL Analytical, Germany)28. Conventional chemical components include nicotine, total nitrogen, reducing sugar, total sugar and starch content.

Determination of nicotine: A total of 0.25 g tobacco samples was extracted with 25 mL water in a 50 mL Erlenmeyer flask. The nicotine in the extract reacted with p-aminobenzenesulfonic acid and cyanogen chloride, which was produced by the on-line reaction of potassium cyanide and chloramine T. The reaction products were determined at 460 nm by a colorimeter29.

Determination of total nitrogen: 0.1 g sample was weighed in the digestive tube, and 0.1 g of mercuric oxide, 1.0 g of potassium sulfate and 5 mL of concentrated sulfuric acid were added. After digestion and decomposition, the nitrogen was converted into ammonia. Under alkaline conditions, ammonia was oxidized to chloride by sodium hypochlorite, and then reacted with sodium salicylate to produce an indigo dye, which was determined at 660 nm30.

Determination of water-soluble sugars (reducing sugar and total sugar): 0.25 g of sample was extracted in a 50 mL flask with 25 mL of 5% acetic acid aqueous solution. The sugars in the extract reacted with p-hydroxymethyl hydrazide to produce a yellow azo compound in an alkaline medium at 85 C. The colorimetric determination was performed at 410 nm31.

Determination of starch: 0.25 g sample was weighed and ultrasonically extracted with 25 mL of 80% ethanol-saturated sodium chloride solution for 30 min to remove the interfering substances in tobacco products. The extract was discarded and ultrasonically extracted with 40% perchloric acid for 10 min. The starch reacted with iodine under acidic conditions and was determined at 570 nm32.

Determination of free amino acids

An automatic amino acid analyzer (Hitachi 8900, Japan) was used to determine the content of free amino acids in tobacco samples33.

Construction of prediction model

Sample set selection

In order to ensure the uniformity of the distribution of the components to be predicted in the train set and the validation set, the gradient quality method was used to divide the tobacco leaves sample set. All samples were sorted in ascending order according to the content of the components to be predicted, and then one sample was taken as the validation set sample at an equal interval. All Fujian tobacco leaves samples were divided into a modeling set and a validation set at a ratio of 3 : 1. The test set for the model is the Yunnan tobacco sample set. Samples treatment and assay methods were consistent with those of the Fujian samples (Fig. 1).

Fig. 1.

Fig. 1

Open in a new tab

The specific division of tobacco data.

Modeling method

Four algorithms, including partial least squares regression (PLSR), ridge regression (RR), support vector machine (SVM) and random forest (RF), were used to construct the prediction model of conventional chemical components of tobacco leaves during tobacco curing. In the study, the L* value, a* value, b* value, H* value, H° value and C* value were used as independent variables, and the tobacco leaf to be predicted during the curing process was used as a dependent variable. The system grid search method was used to optimize the model parameters. Through 10-fold cross-validation, the best prediction model was determined when the root mean square error of training set (RMSET) was the smallest.

Model evaluation

Coefficient of determination (R2), RMSET, root mean square error of validation (RMSEV), and residual prediction deviation (RPD) were used to evaluate the performance of the model. The smaller the root mean square error is, the closer the determination coefficient R2is to 1, indicating the higher the accuracy of the model. RPD is an index to evaluate the overall prediction performance of the model. When RPD ≥ 2, it shows that the model can predict the predicted components more accurately. When 1.4 ≤ RPD < 2, it shows that the model can only predict the predicted components roughly34,35.

graphic file with name M1.gif 1
graphic file with name M2.gif 2
graphic file with name M3.gif 3
Note

Yi, and are the true value, estimated value and average value of sample “i”, respectively, and “n” is the number of samples.

Data analysis

All experiments were performed in triplicate, with results reported as mean value ± standard deviation. Significance testing for differences was conducted using SPSS Statistics 21.0 (SPSS Inc., Shanghai, China), and a probability value (P ≤ 0.05) indicated that differences between means were statistically significant. Origin 2021 software (OriginLab Corporation, USA) was used to draw correlation heatmap. Python 3.9 was used to build the model.

Results and discussion

Significance analysis

Variations of color values of tobacco leaves

The change trend of color values of middle and upper tobacco leaves during curing process showed high consistent (Table 1). The L*, b* and C* values continued to increase, and then gradually decreased. The a* value and H* value increased significantly first and then increased trend has slowed. The H° value increased rapidly and then tended to be stable. The L*, b* and C* values of tobacco increased significantly with the increase of curing time, and then decreased significantly after 60–72 h. The a*, H* and H° values increased significantly in the curing time of 0–48 h.

Table 1.

The variations of color values in tobacco leaves under different curing time.

Location Curing time (h) Indicators
L* a* b* H* H° C*
Upper leaves 0 55.94 ± 1.92ef -8.11 ± 0.80f 43.43 ± 2.68d -0.19 ± 0.03e -1.38 ± 0.03c 44.20 ± 2.48c
24 66.80 ± 2.15ab 2.41 ± 2.29e 56.22 ± 1.63a 0.04 ± 0.04d 0.83 ± 0.57b 56.32 ± 1.73a
48 68.60 ± 0.59a 6.64 ± 0.72d 57.97 ± 1.12a 0.11 ± 0.01c 1.46 ± 0.01a 58.36 ± 1.12a
60 66.80 ± 1.19ab 8.12 ± 0.96cd 54.48 ± 2.46a 0.15 ± 0.02bc 1.42 ± 0.02a 55.10 ± 2.30a
72 65.82 ± 1.17ab 8.60 ± 0.79bcd 50.69 ± 1.42b 0.17 ± 0.02b 1.40 ± 0.02a 51.42 ± 1.37b
84 63.15 ± 2.47bc 9.89 ± 1.06abc 50.43 ± 2.83b 0.20 ± 0.030ab 1.38 ± 0.03a 51.41 ± 2.61b
96 60.65 ± 2.02cd 10.90 ± 1.45ab 47.86 ± 1.54b 0.23 ± 0.04a 1.35 ± 0.04a 49.11 ± 1.16b
108 58.13 ± 1.77de 11.70 ± 1.62a 47.18 ± 1.54bc 0.25 ± 0.04a 1.33 ± 0.04a 48.64 ± 1.33b
120 58.12 ± 0.14de 11.05 ± 1.03a 47.72 ± 1.81bc 0.23 ± 0.03a 1.34 ± 0.03a 49.00 ± 1.55b
132 52.76 ± 2.32f 10.40 ± 0.90abc 44.08 ± 2.52cd 0.24 ± 0.02a 1.34 ± 0.02a 45.30 ± 2.58c
Middle leaves 0 55.85 ± 0.92f -9.79 ± 0.21g 42.16 ± 1.14f -0.23 ± 0.01h -1.34 ± 0.01b 43.31 ± 1.11g
24 70.03 ± 2.1ab 1.59 ± 2.04f 58.34 ± 0.65e 0.03 ± 0.04g 0.50 ± 1.78a 58.39 ± 0.63a
48 70.37 ± 0.09a 5.17 ± 0.30de 53.75 ± 1.84d 0.10 ± 0f 1.48 ± 0a 54.00 ± 1.84b
60 69.93 ± 0.69ab 4.17 ± 0.46e 50.79 ± 2.13d 0.08 ± 0.01f 1.49 ± 0.01a 50.97 ± 2.08cd
72 70.63 ± 2.10a 5.13 ± 1.36de 51.78 ± 0.39d 0.10 ± 0.03f 1.47 ± 0.03a 52.05 ± 0.36bc
84 68.06 ± 0.44b 5.68 ± 0.80cde 49.89 ± 0.71cd 0.11 ± 0.01ef 1.46 ± 0.01a 50.22 ± 0.79cde
96 65.85 ± 1.22c 6.61 ± 0.37bcd 47.69 ± 0.68bc 0.14 ± 0.01de 1.43 ± 0.01a 48.15 ± 0.66ef
108 63.74 ± 0.80d 7.29 ± 0.41abc 48.21 ± 1.20b 0.15 ± 0.01d 1.42 ± 0.01a 48.77 ± 1.24def
120 64.68 ± 0.50cd 7.75 ± 0.84ab 49.77 ± 0.41ab 0.16 ± 0.02acd 1.42 ± 0.02a 50.38 ± 0.29cde
132 61.50 ± 0.42e 8.58 ± 1.14a 46.66 ± 2.04a 0.18 ± 0.02ac 1.39 ± 0.02a 47.45 ± 2.21f
Open in a new tab

Note: Different letters indicate significant differences (P ≤ 0.05) at different curing times for the same part of the tobacco.

There were some differences between the color values of middle and upper tobacco leaves. The L* value of the upper leaves reached the maximum at 48 h, and the L* value of the middle leaves reached the maximum at 72 h. The b* value and C* value reached the maximum at 24–48 h of curing. Maximum values were reached after 24 h and 48 h of curing for the upper and middle leaves, respectively. During curing process, the b* and C* values of the two sites differed significantly at 48 h and 60 h. The b* and C*values of the upper leaves were significantly higher than those of the middle leaves. The color change of tobacco leaves during curing was divided into two stages. The first stage was that the color of tobacco leaves changed from yellowish green to light yellow (0–72 h), and the second stage was that the color of tobacco leaves changed from light yellow to dark orange (72–132 h). This is consistent with the results reported by Meng et al36.. Combined with the change of moisture content, in the early stage of curing, the moisture content decreased slowly, and the leaves contained high moisture content (50-80%), which gave a brighter feeling visually, so the L* value reached the maximum value. In the later stage of curing, the moisture content decreased rapidly, the yellowing degree of tobacco leaves slowed down, and the color depth continued to accelerate. Therefore, the a* value increased slightly and the b*value decreased6.

The changes of pigments content and polyphenol content in tobacco leaves

Overall, the trends of the pigments content and polyphenols content of the middle and upper tobacco during curing process were generally consistent (Table 2). The results demonstrated that the pigments content decreased significantly at the beginning of the curing process and did not change significantly from 48 h to the end of the curing process. During curing process, the tobacco leaves were oxidized and decomposed under the action of lipoxygenase to form intermediate products such as violaxanthin, geraniol and ionone8. This may be the reason for the decrease of pigments content.

Table 2.

The changes of pigments content and polyphenols content in tobacco leaves under different curing time.

Location Curing time (h) Indicators
Pigments contents (µg/g) Polyphenols contents (mg/g)
Lutein β-Carotene Neochlorogenic acid Caffequinic acid Chlorogenic acid Scopoletin Rutin Kaempferol glycoside Total polyphenols
Upper leaves 0 225.09 ± 4.52a 109.51 ± 7.19a 1.50 ± 0.15d 1.66 ± 0.14e 12.06 ± 0.51d 0.20 ± 0.01c 17.84 ± 0.44c 2.07 ± 0.52a 35.34 ± 0.98d
24 149.18 ± 7.59b 75.26 ± 8.04b 1.67 ± 0.17cd 1.89 ± 0.19de 15.00 ± 1.12abc 0.24 ± 0.02b 18.19 ± 1.49c 2.46 ± 0.44a 39.45 ± 2.78bcd
48 114.77 ± 9.35c 55.44 ± 7.09c 1.85 ± 0.19abcd 2.01 ± 0.17cde 14.14 ± 2.04bcd 0.21 ± 0.04bc 18.95 ± 3.68bc 2.19 ± 0.29a 39.35 ± 5.72bcd
60 117.75 ± 8.91c 55.30 ± 6.56c 1.97 ± 0.42abc 2.20 ± 0.39abcd 13.98 ± 1.67bcd 0.19 ± 0.01c 19.00 ± 1.11bc 2.15 ± 0.20a 39.38 ± 3.23bcd
72 106.61 ± 8.25cd 50.03 ± 2.86c 1.80 ± 0.11bcd 1.99 ± 0.17de 13.03 ± 1.15cd 0.21 ± 0.02bc 18.77 ± 1.84bc 2.25 ± 0.30a 37.96 ± 3.27cd
84 105.63 ± 11.64cd 51.79 ± 9.34c 2.05 ± 0.08ab 2.21 ± 0.10abcd 14.78 ± 0.78abc 0.22 ± 0bc 19.45 ± 1.52abc 2.22 ± 0.09a 40.83 ± 1.94bc
96 89.02 ± 9.35de 44.92 ± 5.32c 1.94 ± 0.07abc 2.11 ± 0.06bcd 14.78 ± 1.16abc 0.26 ± 0a 20.90 ± 0.92abc 2.36 ± 0.26a 42.25 ± 2.35abc
108 88.36 ± 7.04de 44.50 ± 4.73c 2.19 ± 0.08a 2.42 ± 0.07ab 15.93 ± 0.61ab 0.28 ± 0.01a 21.67 ± 2.07ab 2.24 ± 0.10a 44.54 ± 2.59ab
120 77.38 ± 8.63e 41.01 ± 5.09c 2.19 ± 0.28a 2.36 ± 0.28abc 15.21 ± 1.46abc 0.27 ± 0.01a 20.56 ± 0.97abc 2.21 ± 0.14a 42.61 ± 1.45abc
132 78.52 ± 8.08e 42.68 ± 3.82c 2.18 ± 0.11ab 2.48 ± 0.07a 16.77 ± 0.17a 0.27 ± 0a 22.38 ± 0.46a 2.35 ± 0.18a 46.24 ± 0.09a
Middle leaves 0 231.98 ± 15.99b 112.47 ± 7.59a 1.42 ± 0.20d 1.97 ± 0.35d 10.77 ± 0.96c 0.19 ± 0.02c 16.15 ± 0.49e 1.38 ± 0.08e 31.89 ± 2.07d
24 167.43 ± 15.62a 64.36 ± 7.81b 1.82 ± 0.05cd 2.43 ± 0.13cd 16.16 ± 0.60b 0.27 ± 0.02ab 18.17 ± 0.59de 1.52 ± 0.06de 40.38 ± 0.75c
48 110.49 ± 14.43c 41.28 ± 9.82c 1.79 ± 0.19cd 2.47 ± 0.28cd 16.36 ± 0.23b 0.30 ± 0.02a 18.86 ± 1.14cd 1.60 ± 0.13de 41.37 ± 0.59c
60 114.64 ± 11.08c 45.77 ± 6.95bc 2.17 ± 0.20bc 2.64 ± 0.21bc 16.2 ± 1.05b 0.24 ± 0b 18.37 ± 1.47de 1.67 ± 0.21cd 41.28 ± 2.40c
72 114.42 ± 11.76cd 48.22 ± 5.98bc 2.16 ± 0.39bc 2.78 ± 0.38bc 17.24 ± 0.59b 0.26 ± 0.03ab 20.07 ± 1.34bcd 1.86 ± 0.08bc 44.36 ± 1.80c
84 92.35 ± 12.53cd 38.87 ± 7.49c 2.56 ± 0.11ab 3.16 ± 0.14ab 19.99 ± 1.10a 0.30 ± 0.01a 21.15 ± 0.89abc 1.97 ± 0.03ab 49.13 ± 1.25b
96 97.15 ± 10.83cd 39.98 ± 5.57c 2.73 ± 0.12a 3.15 ± 0.18ab 19.77 ± 0.69a 0.28 ± 0.02ab 22.45 ± 2.74ab 2.09 ± 0.19ab 50.47 ± 3.59ab
108 93.17 ± 11.61cd 41.81 ± 3.43c 2.52 ± 0.22ab 2.99 ± 0.21abc 20.69 ± 0.51a 0.27 ± 0.03ab 22.62 ± 2.04ab 2.16 ± 0.23a 51.26 ± 2.30ab
120 93.41 ± 10.15cd 45.64 ± 4.39bc 2.92 ± 0.19a 3.53 ± 0.34a 22.03 ± 1.78a 0.30 ± 0.04a 23.18 ± 1.50a 2.22 ± 0.16a 54.19 ± 3.23a
132 85.77 ± 13.11d 45.18 ± 3.41bc 2.90 ± 0.47a 3.48 ± 0.69a 21.35 ± 2.57a 0.29 ± 0.01a 21.77 ± 0.82ab 1.97 ± 0.07ab 51.77 ± 4.27ab
Open in a new tab

Note: Different letters indicate significant differences (P ≤ 0.05) at different curing times for the same part of the tobacco.

The polyphenols content increased significantly from the beginning to the end of curing. No significant change was observed in neochlorogenic acid, caffequinic acid, chlorogenic, rutin and kaempferol glycoside content at the curing time of 84–132 h. Total polyphenols content were not significantly different from 84 h to 132 h of curing. The changes of polyphenols in tobacco leaves were very severe during the curing process, and the total polyphenols increased significantly due to the cleavage and enzymatic decomposition of phenolic glycosides. Under the action of peroxidase and polyphenol oxidase, polyphenols are easily oxidized to light red to dark brown quinones and their polymers, so that the color of tobacco leaves changes from yellow to different degrees of tan37.

Moisture content and conventional chemical components changes of tobacco leaves

The moisture content of the upper, middle leaves of tobacco in tobacco baking room was significantly different (Table 3). During curing process, the moisture content decreased in turn. From the beginning to the end of curing, the moisture content of the tobacco leaves decreased to 15-20%. Before the curing time (48 h), the moisture content changed little. The moisture content decreased substantially from the curing time 48 h to 96 h. After that, moisture content tended to be stable. These results indicated that the water loss of tobacco leaves was slow in the early stage (from 0 h to 48 h), accelerated in the middle stage (from 48 h to 96 h), and stabilized in the later stage (from 96 h to 132 h). This is basically consistent with the research results of Condorí et al2.. The results were divided into 55 h and 117 h.

Table 3.

The changes of moisture content and conventional chemical components in tobacco leaves under different curing time.

Location Curing time (h) Indicators (%)
Moisture
content		 Nicotine
content		 Total nitrogen
content		 Reducing sugar
content		 Total sugar
content		 Starch
content
Upper leaves 0 73.77 ± 2.51a 2.51 ± 0.23b 1.67 ± 0.16b 4.95 ± 0.85b 7.35 ± 1.07d 31.61 ± 1.08a
24 74.18 ± 0.63a 3.15 ± 0.16a 1.71 ± 0.06ab 17.27 ± 1.77a 20.38 ± 2.05c 18.70 ± 2.44b
48 69.00 ± 2.73a 3.27 ± 0.52a 1.86 ± 0.21ab 18.46 ± 0.50a 24.47 ± 1.39b 11.06 ± 1.34c
60 57.55 ± 5.17b 3.53 ± 0.31a 1.93 ± 0.08a 16.87 ± 1.53a 24.81 ± 0.68b 7.71 ± 1.08cd
72 38.72 ± 3.28c 3.22 ± 0.48a 1.79 ± 0.13ab 18.64 ± 1.31a 28.59 ± 1.24a 6.06 ± 1.71d
84 25.79 ± 4.78d 3.18 ± 0.15a 1.85 ± 0.10ab 17.70 ± 1.73a 28.89 ± 1.42a 5.90 ± 0.70d
96 15.65 ± 3.08e 3.34 ± 0.36a 1.84 ± 0.08ab 19.40 ± 1.20a 28.28 ± 1.27a 5.86 ± 1.03d
108 9.71 ± 0.78f 3.34 ± 0.42a 1.82 ± 0.07ab 19.06 ± 2.10a 28.00 ± 1.51a 5.28 ± 0.23d
120 7.84 ± 0.73f 3.00 ± 0.29ab 1.80 ± 0.08ab 19.18 ± 0.82a 29.07 ± 0.56a 6.29 ± 0.66d
132 6.34 ± 0.27f 3.35 ± 0.25a 1.91 ± 0.09a 17.69 ± 0.54a 27.79 ± 0.75a 4.82 ± 1.00d
Middle leaves 0 72.31 ± 1.13ab 1.85 ± 0.14d 1.37 ± 0.06b 5.99 ± 1.34c 7.56 ± 1.34d 35.22 ± 0.93a
24 75.22 ± 2.43a 1.93 ± 0.30cd 1.40 ± 0.05b 22.07 ± 0.92b 25.77 ± 1.75c 22.44 ± 1.43b
48 68.61 ± 2.35b 2.12 ± 0.49bcd 1.41 ± 0.19b 27.12 ± 1.44a 35.58 ± 1.59b 10.38 ± 2.76c
60 59.61 ± 1.70c 2.18 ± 0.17abcd 1.53 ± 0.06ab 25.90 ± 0.11a 36.30 ± 2.25b 6.64 ± 1.30d
72 55.48 ± 0.84c 2.61 ± 0.13a 1.63 ± 0.08a 25.09 ± 1.56ab 35.77 ± 2.10b 6.54 ± 0.19de
84 34.57 ± 1.49d 2.32 ± 0.30abc 1.51 ± 0.10ab 27.86 ± 2.34a 40.99 ± 1.33a 4.28 ± 0.36de
96 26.50 ± 3.78e 2.35 ± 0.17abc 1.58 ± 0.09ab 26.15 ± 0.89a 38.82 ± 1.33ab 4.29 ± 0.19de
108 21.17 ± 1.97f 2.44 ± 0.13ab 1.68 ± 0.11a 24.49 ± 1.59ab 37.91 ± 2.16ab 5.14 ± 1.02de
120 20.46 ± 2.23f 2.28 ± 0.09abcd 1.65 ± 0.07a 25.38 ± 1.43ab 39.21 ± 1.57ab 4.06 ± 0.75de
132 18.26 ± 2.42f 2.38 ± 0.15abc 1.63 ± 0.22a 24.76 ± 2.67ab 37.71 ± 3.20ab 3.56 ± 1.15e
Open in a new tab

Note: Different letters indicate significant differences (P ≤ 0.05) at different curing times for the same part of the tobacco.

According to Table 3, the content of nicotine and total nitrogen in the two parts of tobacco leaves showed a basically stable trend during curing process. The content of reducing sugar and total sugar increased rapidly and then tended to be stable. Starch showed a trend of rapid decline first and then stabilized. Among them, the content of reducing sugar, total sugar and starch tended to be stable after curing for 60 h. Compared with the two parts, the content of nicotine and total nitrogen in the upper leaves were higher than those in the middle leaves, and the content of reducing sugar and total sugar in the middle leaves were higher than those in the upper leaves, while the changes of starch content in the two parts were basically the same. The content of nicotine, total nitrogen, reducing sugar and total sugar in different parts of tobacco leaves were significantly different during curing process. The content of nicotine and total nitrogen in the upper leaves were about 1% and 0.5% higher than those in the middle leaves, respectively, which may be related to the different nitrogen content, water content and sunlight exposure29. The difference of reducing sugar and total sugar content gradually increased from 0 to 10% and 15% respectively, which indicated that the content of reducing sugar and total sugar was greatly affected during the curing process of tobacco leaves, which may be related to the accumulation of internal macromolecules in the two parts, the tightness of the organizational structure and the degree of influence by curing32.

Free amino acid content changes of tobacco leaves

A total of 21 free amino acids were detected in the tobacco leaves (Tables 4, 5 and 6), which include 18 protein amino acids and 3 non-protein amino acids. The 18 protein amino acids are categorized according to their side chains: (1) aromatic amino acids, including phenylalanin (Phe), tryptophan (Try), tyrosine (Tyr); (2) acidic amino acids, including aspartic acid (Asp) and glutamic acid (Glu); (3) alkaline amino acids including lysine (Lys), histidine (His), and arginine (Arg); (4) aliphatic amino acids, including alanine (Ala), glycine (Gly), isoleucine (Ile), leucine (Leu), and valine (Val); (5) Hydroxy amino acids, including threonine (Thr) and serine (Ser); (6) sulfur-containing amino acids: cystine (Cys); (7) amide amino acids: asparagine (Asn); and (8) sublethionic amino acids: proline (Pro). Three non-protein amino acids, including β-alanine (β-Ala), β-Aminoisobutyric acid (β-AiBA), and γ-aminobutyric acid (γ-ABA).

Table 4.

The changes of aromatic, acidic, alkaline amino acid content in tobacco leaves under different curing time.

Location Curing time (h) Indicators (mg/g)
Aromatic amino acid content Acidic amino acid content Alkaline amino acid content
Phenylalanin Tryptophan Tryptophan Aspartic Glutamic Histidine Arginine
Upper leaves 0 0.25 ± 0.04f 0.26 ± 0.01f 0.17 ± 0.03a 0.50 ± 0.07a 0.28 ± 0.11a 0.09 ± 0.02e 0.05 ± 0.01abc
24 0.54 ± 0.02cde 0.55 ± 0cd 0.16 ± 0.05a 0.19 ± 0.11b 0.16 ± 0.05b 0.25 ± 0.01d 0.08 ± 0.03a
48 0.81 ± 0.17ab 0.74 ± 0.13ab 0.13 ± 0ab 0.16 ± 0.04b 0.13 ± 0.03b 0.42 ± 0.05ab 0.07 ± 0.01a
60 0.84 ± 0.11a 0.80 ± 0.08a 0.12 ± 0.01bc 0.20 ± 0.02b 0.16 ± 0.05b 0.49 ± 0.07a 0.07 ± 0.01ab
72 0.63 ± 0.09bc 0.69 ± 0.13abc 0.10 ± 0.02bcd 0.15 ± 0.02b 0.13 ± 0.02b 0.40 ± 0.08abc 0.05 ± 0.02abc
84 0.58 ± 0.15cd 0.64 ± 0.08bc 0.10 ± 0.02bcd 0.20 ± 0.06b 0.14 ± 0.02b 0.39 ± 0.05abc 0.05 ± 0.01abc
96 0.47 ± 0.08cdef 0.56 ± 0.05cd 0.09 ± 0.02bcd 0.16 ± 0.04b 0.13 ± 0.02b 0.35 ± 0.05bcd 0.04 ± 0.01bc
108 0.43 ± 0.12cdef 0.46 ± 0.09de 0.09 ± 0.02cd 0.18 ± 0.05b 0.17 ± 0.02b 0.30 ± 0.09cd 0.05 ± 0.02abc
120 0.40 ± 0.10def 0.44 ± 0.09de 0.09 ± 0.01bcd 0.16 ± 0.03b 0.19 ± 0.03b 0.26 ± 0.06d 0.04 ± 0.01c
132 0.33 ± 0.09ef 0.39 ± 0.03ef 0.07 ± 0.02d 0.18 ± 0.02b 0.18 ± 0.03b 0.27 ± 0.05d 0.05 ± 0.02abc
Middle leaves 0 0.19 ± 0.04c 0.22 ± 0.07e 0.10 ± 0.02a 0.32 ± 0.03a 0.21 ± 0.07a 0.26 ± 0e 0.03 ± 0ab
24 0.38 ± 0.03abc 0.58 ± 0.01ab 0.09 ± 0.02ab 0.10 ± 0.01b 0.13 ± 0.01b 0.22 ± 0.01cd 0.03 ± 0ab
48 0.37 ± 0.04abc 0.60 ± 0.05ab 0.09 ± 0.01ab 0.10 ± 0.02b 0.09 ± 0.02b 0.28 ± 0.02bcd 0.04 ± 0ab
60 0.46 ± 0.08ab 0.67 ± 0.07a 0.09 ± 0.01ab 0.13 ± 0.01b 0.11 ± 0.01b 0.33 ± 0.04ab 0.05 ± 0.01a
72 0.50 ± 0.12a 0.70 ± 0.08a 0.09 ± 0.02ab 0.14 ± 0.04b 0.12 ± 0.03b 0.38 ± 0.08a 0.05 ± 0.02ab
84 0.27 ± 0.02bc 0.52 ± 0.04bc 0.07 ± 0.02ab 0.12 ± 0.02b 0.10 ± 0.02b 0.29 ± 0.04abc 0.03 ± 0b
96 0.31 ± 0.11bc 0.48 ± 0.11bcd 0.07 ± 0.02ab 0.14 ± 0.03b 0.09 ± 0.01b 0.27 ± 0.09bcd 0.03 ± 0.01ab
108 0.28 ± 0.03bc 0.42 ± 0.04cd 0.07 ± 0.01ab 0.13 ± 0.02b 0.10 ± 0.03b 0.25 ± 0.04bcd 0.03 ± 0ab
120 0.27 ± 0.05bc 0.45 ± 0.03cd 0.08 ± 0.01ab 0.14 ± 0.02b 0.10 ± 0.03b 0.22 ± 0.03cd 0.03 ± 0ab
132 0.21 ± 0.14c 0.39 ± 0.11d 0.06 ± 0.03b 0.13 ± 0.06b 0.12 ± 0.04b 0.18 ± 0.07d 0.03 ± 0.01ab
Open in a new tab

Note: Different letters indicate significant differences (P ≤ 0.05) at different curing times for the same part of the tobacco.

Table 5.

The changes of amino acid content with aliphatic, hydroxyl and amide groups in tobacco leaves under different curing time.

Location Curing time (h)  Indicators (mg/g)
Aliphatic amino acid content Amino acid content with hydroxyl and amide groups
Glycine Isoleucine Leucine Valine Threonine Serine Asparagine
Upper leaves 0 0.03 ± 0c 0.09 ± 0.01abc 0.18 ± 0.02a 0.17 ± 0.03d 0.21 ± 0.03a 0.28 ± 0.06a 0.37 ± 0.22d
24 0.03 ± 0.01c 0.07 ± 0.02bc 0.13 ± 0.03bc 0.21 ± 0.05cd 0.24 ± 0.06a 0.30 ± 0.04a 0.97 ± 0.07cd
48 0.04 ± 0.01bc 0.07 ± 0.01bc 0.15 ± 0.02abc 0.18 ± 0.04d 0.20 ± 0.02ab 0.21 ± 0.03b 2.47 ± 0.58ab
60 0.04 ± 0.01bc 0.07 ± 0c 0.13 ± 0.01bc 0.17 ± 0.03d 0.15 ± 0.03bc 0.16 ± 0.03bc 2.87 ± 0.41a
72 0.05 ± 0abc 0.07 ± 0.01c 0.12 ± 0.02bc 0.16 ± 0.02d 0.10 ± 0.03cd 0.11 ± 0.03cd 2.06 ± 0.58ab
84 0.06 ± 0.01ab 0.07 ± 0.01c 0.11 ± 0.01c 0.17 ± 0.03d 0.08 ± 0.02d 0.10 ± 0.02d 2.13 ± 0.58ab
96 0.06 ± 0.01ab 0.07 ± 0bc 0.10 ± 0.01c 0.19 ± 0.03cd 0.07 ± 0.01d 0.08 ± 0.02d 1.92 ± 0.41b
108 0.06 ± 0ab 0.09 ± 0.02abc 0.11 ± 0.02c 0.26 ± 0.05c 0.07 ± 0.01d 0.08 ± 0.01d 1.86 ± 0.67b
120 0.05 ± 0abc 0.11 ± 0.01ab 0.12 ± 0.01c 0.34 ± 0.06b 0.07 ± 0.01d 0.08 ± 0.01d 1.72 ± 0.48bc
132 0.07 ± 0.04a 0.12 ± 0.04a 0.16 ± 0.04ab 0.40 ± 0.04a 0.08 ± 0.01d 0.08 ± 0.01d 1.72 ± 0.40bc
Middle leaves 0 0.02 ± 0d 0.06 ± 0.02c 0.11 ± 0.02abc 0.09 ± 0.01d 0.13 ± 0.01a 0.20 ± 0.02b 0.14 ± 0.02d
24 0.03 ± 0cd 0.06 ± 0.01c 0.10 ± 0.01abc 0.14 ± 0.01cd 0.12 ± 0.02ab 0.27 ± 0.03a 0.77 ± 0.08cd
48 0.03 ± 0.01bc 0.06 ± 0.01c 0.11 ± 0.02abc 0.13 ± 0.01d 0.08 ± 0.01cd 0.16 ± 0.03b 2.13 ± 0.75bcd
60 0.04 ± 0a 0.06 ± 0.01c 0.12 ± 0.02ab 0.16 ± 0.02cd 0.09 ± 0.03bc 0.18 ± 0.07b 3.80 ± 0.46ab
72 0.04 ± 0.01a 0.06 ± 0c 0.12 ± 0.02abc 0.15 ± 0.01cd 0.08 ± 0.03cde 0.15 ± 0.06bc 4.57 ± 0.43a
84 0.04 ± 0a 0.05 ± 0.01c 0.09 ± 0.01bc 0.14 ± 0.04cd 0.05 ± 0e 0.09 ± 0.01cd 2.96 ± 1.01abc
96 0.04 ± 0a 0.06 ± 0c 0.09 ± 0.01c 0.16 ± 0.01cd 0.06 ± 0.01de 0.09 ± 0.03cd 3.06 ± 0.68abc
108 0.04 ± 0a 0.09 ± 0.03bc 0.11 ± 0.02abc 0.25 ± 0.10bc 0.06 ± 0.01de 0.09 ± 0.01cd 2.98 ± 1.26abc
120 0.04 ± 0a 0.12 ± 0.03ab 0.13 ± 0.02a 0.32 ± 0.08ab 0.07 ± 0.01cde 0.09 ± 0.01cd 2.69 ± 0.91abc
132 0.04 ± 0ab 0.14 ± 0.04a 0.11 ± 0.03 0.39 ± 0.11a 0.08 ± 0.01cde 0.08 ± 0.02d 2.37 ± 0.85abcd
Open in a new tab

Note: Different letters indicate significant differences (P ≤ 0.05) at different curing times for the same part of the tobacco.

Table 6.

The changes of amino acid content with sulfur, subamino, non-protein and total in tobacco leaves under different curing time.

Location Curing time (h) Indicators (mg/g)
Sulfur, subamino acid content Non-protein amino acid content Total free amino acids
Cystine Proline β-alanine β-Aminoisobutyric γ-aminobutyric
Upper leaves 0 0.10 ± 0.02c 0.15 ± 0.08e 0.16 ± 0.08d 0.11 ± 0.02a 1.10 ± 0.19a 3.68 ± 0.44c
24 0.11 ± 0.02c 1.40 ± 0.59e 0.27 ± 0.06bc 0.08 ± 0.01ab 0.83 ± 0.18b 5.44 ± 0.13c
48 0.12 ± 0.01bc 5.45 ± 0.92d 0.18 ± 0.07cd 0.07 ± 0b 0.73 ± 0.1bc 10.12 ± 1.69b
60 0.13 ± 0.01abc 7.82 ± 0.60c 0.18 ± 0.04cd 0.08 ± 0ab 0.79 ± 0.03bc 15.14 ± 1.85a
72 0.11 ± 0.02bc 8.60 ± 0.85bc 0.15 ± 0d 0.07 ± 0.01b 0.86 ± 0.11b 17.02 ± 1.39a
84 0.12 ± 0.02bc 9.88 ± 0.81ab 0.17 ± 0.01cd 0.07 ± 0.01b 0.78 ± 0.11bc 15.79 ± 0.98a
96 0.13 ± 0.01abc 10.66 ± 0.23a 0.22 ± 0.04cd 0.06 ± 0.01b 0.69 ± 0.1bcd 15.55 ± 1.50a
108 0.13 ± 0.01abc 8.98 ± 0.69abc 0.24 ± 0.02bcd 0.07 ± 0.01b 0.62 ± 0.06bcd 15.68 ± 1.51a
120 0.18 ± 0.03a 8.92 ± 0.71abc 0.36 ± 0.06a 0.08 ± 0.04b 0.55 ± 0.02cd 15.33 ± 0.87a
132 0.17 ± 0.05ab 9.39 ± 0.76abc 0.33 ± 0.05ab 0.07 ± 0.03b 0.46 ± 0.05d 14.21 ± 1.85a
Middle leaves 0 0.05 ± 0.01c 0.07 ± 0.02c 0.17 ± 0.08cde 0.07 ± 0.04ab 1.07 ± 0.09a 4.97 ± 0.68c
24 0.06 ± 0c 1.00 ± 0.22c 0.23 ± 0.05bc 0.08 ± 0.01a 0.82 ± 0.08b 6.86 ± 0.20c
48 0.06 ± 0bc 4.27 ± 0.66b 0.16 ± 0.07cde 0.06 ± 0.02abcd 0.70 ± 0.09bc 12.82 ± 1.09b
60 0.07 ± 0.01bc 6.91 ± 0.30a 0.09 ± 0.03e 0.07 ± 0.02abc 0.80 ± 0.05b 16.11 ± 1.27a
72 0.07 ± 0.01abc 7.89 ± 1.07a 0.13 ± 0.01de 0.05 ± 0bcd 0.69 ± 0.06bc 16.00 ± 1.95a
84 0.06 ± 0bc 8.61 ± 0.17a 0.14 ± 0.02de 0.04 ± 0d 0.65 ± 0.08c 17.45 ± 1.25a
96 0.07 ± 0.01bc 8.35 ± 1.32a 0.20 ± 0.02cd 0.04 ± 0cd 0.60 ± 0.07cd 17.71 ± 0.61a
108 0.07 ± 0.01abc 8.53 ± 0.54a 0.28 ± 0.03b 0.05 ± 0.02bcd 0.57 ± 0.07cd 15.72 ± 1.68ab
120 0.09 ± 0.01ab 8.35 ± 0.30a 0.38 ± 0.05a 0.08 ± 0.01ab 0.49 ± 0.07de 15.53 ± 1.59ab
132 0.09 ± 0.02a 7.83 ± 1.46a 0.41 ± 0.05a 0.08 ± 0.01ab 0.38 ± 0.11e 15.60 ± 1.21ab
Open in a new tab

Note: Different letters indicate significant differences (P ≤ 0.05) at different curing times for the same part of the tobacco.

On the whole, the content of most free amino acids increased first and then decreased during the curing process, which may be due to the degradation and transformation of macromolecular substances (proteins, etc.) in the tobacco leaves during the curing process3,11. The degradation and transformation of macromolecular substances (proteins, etc.) mainly occurred in the yellowing stage and the early color fixing stage. The content of free amino acids decreased to a certain extent during the curing process, mainly due to the non-enzymatic browning reaction between amino acids and reducing sugars, which not only increased the aroma of tobacco leaves, but also increased the proportion of orange leaves33.

With the curing process, Phe, Trp, His and Asn showed a trend of increasing significantly at first and then decreasing gradually. Cys, Val, Pro and β-Ala gradually increased. Tyr and γ-ABA showed a gradual decline, while the acidic amino acids, Lys, Thr and Ser first decreased significantly, and then remained stable. The remaining amino acids were unstable. During the curing process, the aromatic amino acids within the tobacco leaves degrade, which has an impact on the quantity and quality of the tobacco’s aroma. Additionally, they undergo non-enzymatic browning reactions, generating key intermediate products such as Amadori compounds, which subsequently lead to the formation of melanoidins, causing the tobacco leaves to become brown to a certain extent38,39.

Correlation of color values and chemical components of tobacco leaves

As shown in Fig. 2, the L* value was significantly positively correlated with moisture, scopoletin, reducing sugar and total sugar contents, and significantly negatively correlated with β-carotene content. The a* value was significantly negatively correlated with moisture, lutein, β-carotene, scopoletin and starch contents, and significantly positively correlated with other chemical components. It could be due to the decrease in the moisture content of wet basis of tobacco leaves affected the activity of Polyphenol oxidase (PPO) and the content of malondialdehyde (MDA), which caused the change of polyphenol content40. The b*value was significantly positively correlated with moisture and reducing sugar contents. Due to the acceleration of Maillard reaction at higher temperatures, the interaction between sugars and amino acids is easy to form brown compounds, resulting in darker color25.

Fig. 2.

Fig. 2

Open in a new tab

Heat map of the correlation between color values, moisture content, pigments content, polyphenols content and conventional chemical components of tobacco leaves during curing (P ≤ 0.05).

The L* value was significantly positively correlated with Trp (Fig. 3). The a* value was significantly negatively correlated with Asp, Thr and Ser, and significantly positively correlated with Pro and total free amino acids. The b* value was significantly positively correlated with Phe and Trp. Overall, the change in apparent color of tobacco leaves during curing was closely related to the content of chemical components of tobacco leaves.

Fig. 3.

Fig. 3

Open in a new tab

Heat map of the correlation between color values and free amino acids content of tobacco leaves during curing (P ≤ 0.05). Note: Phe: phenylalanin, Try: tryptophan, Tyr: 360 tyrosine, Asp: aspartic acid, Glu: glutamic acid, Lys: lysine, His: histidine, Arg: arginine, Ala: alanine, Gly: glycine, Ile: isoleucine, Leu: leucine, Val: valine, Thr: threonineand, Ser: serine, Cys: cystine, Asn: asparagine, Pro: proline, β-Ala: β-alanine, β-AiBA: β-Aminoisobutyric acid, γ-ABA: γ-aminobutyric acid. 

Construction and validation of prediction model for chemical components of tobacco leaves during curing process

The PLSR, RR, SVM and RF algorithms were used to construct the prediction model of chemical components of tobacco leaves during flue-cured tobacco curing. It can be seen from Table 7 that the moisture, lutein, β-carotene, total polyphenols, reducing sugar, total sugar, starch and partial free amino acids prediction models constructed by four modeling methods based on color values are all good. Among them, the RF algorithm was used to construct the moisture, total polyphenols, reducing sugar, total sugar, starch and partial free amino acids prediction model of tobacco leaves during curing process, the accuracies of the models were the highest. The RR algorithm was used to construct the prediction model of lutein and β-carotene in tobacco curing process with the highest accuracy. The model validation set R2of these indicators was high, and the RPD is basically higher than 2.0. This indicted that it was feasible to predict the content of some chemical components in tobacco leaves by color quantification. Zhu15proposed that a tobacco chemical component analysis method based on a neural network (TCCANN) performed simultaneous quantitative analysis of multiple chemical compositions of tobacco by using near-infrared (NIR) hyperspectroscopy imagery. However, the proposed TCCANN cannot determine the completely accurate determination of chemical components of tobacco. In this study, a more accurate starch prediction model was constructed by color quantization machine learning algorithm. The reason may be that RF is a non-linear ensemble algorithm, which is suitable for modeling high-dimensional data samples41.

Table 7.

Construction and validation of PLSR, RR, SVM and RF models based on color values.

Indicators to be predicted PLSR RR SVM RF
Train set Validation set Train set Validation set Train set Validation set Train set Validation set
R 2 RMSET R 2 RMSEV RPD R 2 RMSET R 2 RMSEV RPD R 2 RMSET R 2 RMSEV RPD R 2 RMSET R 2 RMSEV RPD
Moisture content 0.87 9.11 0.86 9.64 2.68 0.87 9.19 0.86 9.3 2.7 0.9 8.01 0.89 8.53 3.03 0.98 3.81 0.94 6.13 4.22
Lutein content 0.78 22.41 0.79 20.94 2.17 0.78 22.48 0.8 20.51 2.21 0.76 23.42 0.77 21.54 2.11 0.88 16.73 0.75 22.67 2
β-Carotene content 0.75 11.75 0.79 10.41 2.16 0.75 11.64 0.79 10.2 2.21 0.73 12.09 0.77 10.83 2.08 0.86 8.82 0.79 10.26 2.19
Neochlorogenic acid content 0.44 0.35 0.45 0.37 1.34 0.43 0.35 0.44 0.37 1.34 0.45 0.35 0.46 0.36 1.36 0.65 0.28 0.59 0.32 1.56
Caffequinic acid content 0.33 0.47 0.28 0.51 1.18 0.34 0.47 0.31 0.5 1.2 0.36 0.46 0.32 0.5 1.21 0.67 0.33 0.54 0.41 1.47
Chlorogenic acid content 0.46 2.35 0.39 2.62 1.28 0.45 2.35 0.39 2.61 1.28 0.5 2.26 0.44 2.5 1.34 0.88 1.11 0.66 1.94 1.72
Scopoletin content 0.46 0.06 0.44 0.06 1.33 0.45 0.06 0.44 0.06 1.33 0.45 0.06 0.36 0.07 1.25 0.77 0.04 0.66 0.05 1.71
Rutin content 0.44 1.89 0.31 1.76 1.2 0.44 1.89 0.31 1.75 1.21 0.47 1.84 0.36 1.7 1.25 0.62 1.55 0.43 1.6 1.32
Kaempferol glycoside content 0.18 0.35 0.25 0.26 1.16 0.34 0.31 0.3 0.26 1.19 0.32 0.32 0.36 0.24 1.25 0.34 0.31 0.23 0.27 1.14
Total polyphenols content 0.49 4.44 0.42 4.65 1.31 0.49 4.45 0.42 4.65 1.31 0.51 4.35 0.46 4.48 1.36 0.87 2.29 0.69 3.43 1.78
Nicotine content 0.43 0.46 0.32 0.47 1.21 0.43 0.46 0.34 0.46 1.23 0.51 0.43 0.4 0.44 1.29 0.84 0.25 0.44 0.42 1.34
Total nitrogen content’ 0.48 0.15 0.35 0.16 1.24 0.48 0.15 0.35 0.16 1.24 0.84 0.08 0.4 0.16 1.29 0.64 0.12 0.31 0.17 1.21
Reducing sugar content 0.75 3.13 0.74 3.24 1.96 0.75 3.13 0.74 3.24 1.96 0.97 1.16 0.84 2.58 2.46 0.97 1.14 0.91 1.92 3.31
Total sugar content 0.8 4.15 0.8 4.19 2.25 0.8 4.16 0.8 4.18 2.26 0.98 1.2 0.82 3.98 2.37 0.96 1.86 0.93 2.5 3.77
Starch content 0.88 3.2 0.85 3.6 2.61 0.92 2.67 0.9 2.98 3.15 0.89 3.12 0.85 3.59 2.62 0.97 1.68 0.92 2.64 3.57
Asp 0.48 0.07 0.67 0.06 1.74 0.61 0.06 0.55 0.07 1.48 0.58 0.06 0.74 0.05 1.95 0.89 0.03 0.78 0.05 2.12
Thr 0.57 0.04 0.59 0.04 1.56 0.60 0.04 0.56 0.03 1.50 -0.49 0.07 -0.46 0.07 0.83 0.71 0.03 0.77 0.03 2.07
Ser 0.74 0.04 0.77 0.04 2.10 0.78 0.04 0.78 0.04 2.13 0.40 0.06 0.41 0.06 1.30 0.88 0.03 0.84 0.03 2.51
Asn 0.27 1.16 0.21 1.19 1.12 0.44 1.01 0.44 1.01 1.34 0.36 1.08 0.34 1.09 1.23 0.48 0.98 0.38 1.05 1.27
Glu 0.40 0.04 0.43 0.05 1.33 0.39 0.05 0.51 0.04 1.43 -0.19 0.06 0.06 0.06 1.03 0.55 0.04 0.41 0.05 1.30
Gly 0.39 0.01 0.50 0.01 1.41 0.59 0.01 0.25 0.02 1.16 -8.29 0.05 -21.40 0.05 0.21 0.57 0.01 0.59 0.01 1.56
Ala 0.68 0.29 0.74 0.26 1.97 0.76 0.25 0.66 0.31 1.72 0.75 0.26 0.81 0.22 2.31 0.91 0.15 0.88 0.18 2.85
Val 0.48 0.07 0.44 0.08 1.34 0.52 0.07 0.44 0.07 1.34 0.43 0.07 0.48 0.07 1.38 0.70 0.05 0.51 0.07 1.42
Cys 0.56 0.03 0.33 0.04 1.22 0.50 0.03 0.48 0.03 1.39 -1.61 0.06 -0.74 0.06 0.76 0.94 0.01 0.43 0.04 1.32
Ile 0.21 0.03 0.12 0.03 1.07 0.20 0.03 0.16 0.03 1.09 -0.78 0.04 -0.72 0.04 0.76 0.34 0.03 0.17 0.03 1.10
Leu 0.06 0.03 0.08 0.03 1.05 0.24 0.03 0.12 0.03 1.07 -2.00 0.05 -1.78 0.05 0.60 0.34 0.02 0.08 0.03 1.04
Tyr 0.31 0.03 0.39 0.03 1.28 0.44 0.03 0.24 0.03 1.15 -0.64 0.04 -0.55 0.05 0.80 0.52 0.02 0.44 0.03 1.34
Phe 0.33 0.17 0.32 0.17 1.21 0.50 0.15 0.38 0.14 1.27 0.84 0.08 0.45 0.15 1.35 0.91 0.06 0.61 0.13 1.61
β-Ala 0.29 0.08 0.14 0.09 1.08 0.31 0.08 0.27 0.09 1.17 0.32 0.08 0.25 0.08 1.15 0.67 0.06 0.32 0.08 1.21
β-AiBA 0.16 0.02 -0.06 0.02 0.97 0.26 0.02 0.14 0.02 1.08 0.00 0.02 -0.02 0.02 0.99 0.34 0.02 0.04 0.02 1.02
γ-ABA 0.51 0.14 0.48 0.15 1.38 0.54 0.14 0.59 0.13 1.56 0.58 0.13 0.53 0.14 1.46 0.82 0.09 0.56 0.14 1.51
Trp 0.59 0.10 0.58 0.11 1.55 0.65 0.10 0.59 0.10 1.56 0.65 0.10 0.65 0.10 1.69 0.94 0.04 0.69 0.09 1.80
Lys 0.52 0.03 0.57 0.03 1.52 0.57 0.03 0.51 0.03 1.43 -0.62 0.06 -0.65 0.06 0.78 0.79 0.02 0.63 0.03 1.65
His 0.49 0.08 0.48 0.08 1.38 0.57 0.08 0.52 0.08 1.45 0.58 0.07 0.57 0.08 1.53 0.93 0.03 0.71 0.06 1.85
Arg 0.09 0.02 0.14 0.02 1.08 0.27 0.02 0.08 0.02 1.04 -1.52 0.03 -2.64 0.03 0.52 0.31 0.02 0.26 0.01 1.16
Pro 0.78 1.57 0.74 1.80 1.97 0.85 1.35 0.82 1.41 2.36 0.83 1.37 0.81 1.54 2.30 0.93 0.89 0.87 1.29 2.75
Total 0.76 2.28 0.70 2.64 1.84 0.76 2.29 0.70 2.64 1.84 0.75 2.30 0.70 2.65 1.83 0.91 1.37 0.74 2.49 1.95
Note: Phe: phenylalanin, Try: tryptophan, Tyr: tyrosine, Asp: aspartic acid, Glu: glutamic acid, Lys: lysine, His: histidine, Arg: arginine, Ala: alanine, Gly: glycine, Ile: isoleucine, Leu: leucine, Val: valine, Thr: threonineand, Ser: serine, Cys: cystine, Asn: asparagine, Pro: proline, β-Ala: β-alanine, β-AiBA: β-Aminoisobutyric acid, γ-ABA: γ-aminobutyric acid.
Open in a new tab

The scatterplots represent the RR models for two chemical components (lutein and β-carotene) in Fig. 4, and the RF models for the fourteen chemical components (Moisture, total polyphenols, reducing sugar, total sugar, starch and partial free amino acids) in Fig. 5. A small difference between estimated and measured values is shown, and most of the points are evenly and compactly distributed along the diagonal. The closer these points are to the diagonal, the higher their predicted values are, and the better the model fits. The clear linear relationships between the predicted and measured values of the sixteen chemical components are observed, which can predict some chemical components by color quantization.

Fig. 4.

Fig. 4

Open in a new tab

The RR of modeling and verification of content prediction model in curing process.

Fig. 5.

Fig. 5

Open in a new tab

The RF of modeling and verification of content prediction model in curing process. Note: Thr: threonineand, Asp: aspartic acid, Ala: alanine, Ser: serine, Pro: proline.

The estimation model of moisture content, reducing sugar, total sugar and starch content in RF algorithm constructed by Fujian Nanping tobacco samples has high accuracy (model validation set: R2 > 0.90, RPD > 2.0). On the one hand, the color value in correlation analysis is significantly correlated with moisture, reducing sugar, total sugar and starch. On the other hand, the RF algorithm can better deal with the multicollinearity problem, and successfully identify moisture content, reducing sugar, total sugar and starch content as strong predictors, thereby improving the accuracy of the model. Therefore, the samples other than these four high-precision models (Qujing, Yunnan) are tested. It can be seen from Table 8 that the R2 and RPD of the test set of the prediction model of moisture content in tobacco leaves during curing were 0.84 and 3.51, respectively. The R2 and RPD of the test set of reducing sugar content estimation model were 0.73 and 3.91, respectively. The R2 and RPD of the test set of total sugar content estimation model were 0.77 and 3.95, respectively. The test set R2 and RPD of the starch content estimation model were 0.82 and 5.27, respectively. It showed that the RF algorithm had high accuracy and stable fitting effect in estimating the contents of reducing sugar, total sugar and starch in tobacco leaves outside the model during curing process.

Table 8.

Construction and test of RF model based on color values.

Indicators to be predicted Train set Test set
R 2 RMSET R 2 RMSEV RPD
Moisture content 0.98 3.81 0.84 8.84 3.51
Reducing sugar content 0.97 1.14 0.73 3.97 3.93
Total sugar content 0.96 1.86 0.77 4.62 3.95
Starch content 0.97 1.68 0.82 5.79 5.27
Open in a new tab

From Fig. 6, it can be seen that the measured values and estimated values of the test set of the RF algorithm estimation model with higher accuracy are evenly distributed near the 1 : 1 line. The performance of the estimation model constructed in Fujian is considered good, as indicated by the R²and RPD values of the four chemical components in the validation set of samples from Yunnan that were external to the model. The model can thus make fairly accurate estimations of these four chemical components. Some samples that were not accurately estimated can be attributed to regional differences in cultivation conditions and climate, among other factors12. This also indicates that there is room for improvement in the model’s performance. Going forward, adjustments to model parameters, the addition of more relevant features, and the use of more advanced algorithms will be employed to address these issues.

Fig. 6.

Fig. 6

Open in a new tab

Test scatter plot of RF prediction model of moisture, reducing sugar, total sugar and starch content estimation during the curing process.

Conclusions

Changes in the surface color of tobacco leaves are related to the internal chemical content of the tobacco. The correlationship between color values and chemical components of tobacco leaves was analyzed and the results showed that the a* value of tobacco leaves was significantly negatively correlated with moisture, lutein, β-carotene, scopoletin, starch, Asp, Thr and Ser. The b* value of tobacco leaves was significantly positively correlated with moisture content and reducing sugar content. The PLSR, RR, SVM and RF algorithms were used to establish the prediction models of chemical components of tobacco leaves during flue-cured tobacco curing. The results indicated that the RF algorithms performed best to predict tobacco moisture, reducing sugar, total sugar and starch in the curing process, with the R2 values of the model validation sets were more than 0.90 and the RPD values were more than 2.0. These results confirm that it is possible to monitor the chemical components of tobacco in real time during curing process. Based on the color values of tobacco, machine learning algorithm can be used to establish the prediction model of tobacco chemical components in the curing process, which can quickly and accurately predict the moisture, lutein, β-carotene, total polyphenols, reducing sugar, total sugar, starch and partial free amino acids in tobacco, provide theoretical basis and methodological reference for intelligently monitoring of tobacco curing status and improving the quality of tobacco. Further researches would focus on the application of the predicted model on the devices monitoring the images of tobacco leaves to realize the rapid acquisition of chemical components of tobacco leaves during curing process.

Acknowledgements

This work was supported by the major science and technology program of CNTC (NO.110202101084) and the major science and technology program of CNTC (NO.110202201051).

Author contributions

Y. M. and Q. X.: Writing - original draft, Writing - review & editing. G. C., J. L. and S. Z.: Validation. Y.Z., A.W., J.W.: Formal analysis. D.Y., X.C., Q.L. and Q.Z.: Investigation. J.L. and X.C.: Data curation. W.G. and Y.W.: Conceptualization, Methodology, Resources. All authors have read and agreed to the published version of the manuscript.

Data availability

The data presented in this study are available on request from the corresponding author (guoweimin1984@sina.com).

Declarations

Competing interests

The authors declare no competing interests. Yang Meng, Qiang Xu, Yanling Zhang, Aiguo Wang, Jianwei Wang and Weimin Guo were employed by Zhengzhou Tobacco Research Institute of CNTC (Zhengzhou, China). Guangqing Chen, Jianjun Liu and Shuoye Zhou were employed by Henan Provincial Tobacco Company (Zhengzhou, China). Ding Yan and Xianjie Cai were employed by Shanghai Tobacco Company (Shanghai, China). Junying Li and Xuchu Chen were employed by Pingdingshan Branch of Henan Provincial Tobacco Company (Pingdingshan, China). Qiuying Li and Qiang Zeng were employed by Nanping Branch of Fujian Provincial Tobacco Company (Nanping, China). The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Ethics declarations

The authors declare that the collection of plant material (tobacco leaves) and the experimental studies complied with the relevant institutional, national and international guidelines and legislation. The authors confirm that all methods were carried out in accordance with relevant guidelines in the method section. The authors comply with the IUCN Policy Statement on Research Involving Species at Risk of Extinction and the Convention on the Trade in Endangered Species of Wild Fauna and Flora. The authors ensure that the collection of tobacco leaves samples had been licensed by the local tobacco production regulatory agency. 

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Yang Meng and Qiang Xu contributed equally.

Contributor Information

Weimin Guo, Email: guoweimin1984@sina.com.

Yuanhui Wang, Email: wangyuanhui2014@haut.edu.cn.

References

  • 1.Hać, P. et al. Evaluation of mercury content in combustible tobacco products by employing cold vapor atomic absorption spectroscopy and considering the moisture content: a comprehensive study. Monatshefte für Chemie - Chem. Monthly. 153, 829–836. 10.1007/s00706-022-02965-1 (2022). [Google Scholar]
  • 2.Condorí, M. et al. Image processing for monitoring of the cured tobacco process in a bulk-curing stove. Comput. Electron. Agric.16810.1016/j.compag.2019.105113 (2020).
  • 3.Chen, Y. et al. Dynamic changes in physiological and biochemical properties of flue-cured tobacco of different leaf ages during flue-curing and their effects on yield and quality. BMC Plant Biol.1910.1186/s12870-019-2143-x (2019). [DOI] [PMC free article] [PubMed]
  • 4.Wang, Y. & Qin, L. Research on state prediction method of tobacco curing process based on model fusion. J. Ambient Intell. Humaniz. Comput.13, 2951–2961. 10.1007/s12652-021-03129-5 (2021). [Google Scholar]
  • 5.Zou, C. et al. Different yellowing degrees and the industrial utilization of flue-cured tobacco leaves. Scientia Agricola. 76, 1–9. 10.1590/1678-992x-2017-0157 (2019). [Google Scholar]
  • 6.Zong, J. et al. Effect of two drying methods on chemical transformations in flue-cured tobacco. Drying Technol.40, 188–196. 10.1080/07373937.2020.1779287 (2020). [Google Scholar]
  • 7.Rochester, M. The chemical changes that occur during the curing of tobacco leaves. Science, 397–399 (1931). [DOI] [PubMed]
  • 8.Tegan, A. M. et al. Highly resolved systems Biology to dissect the etioplast-to-chloroplast transition in Tobacco leaves. Plant Physiol.180, 654–681. 10.1104/pp.18.01432 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Song, Z. et al. The mechanism of Carotenoid Degradation in Flue-cured Tobacco and Changes in the related enzyme activities at the Leaf-Drying Stage during the bulk curing process. Agricultural Sci. China. 9, 1381–1388. 10.1016/s1671-2927(09)60229-7 (2010). [Google Scholar]
  • 10.Matheis, G. & R, W. J. Modification of proteins by polyphenol oxidase and peroxidase and their products. J. Food Biochem.8, 137–162. 10.1111/j.1745-4514.1984.tb00322.x (1983). [Google Scholar]
  • 11.Zhang, Q. et al. Microbial and enzymatic changes in cigar tobacco leaves during air-curing and fermentation. Appl. Microbiol. Biotechnol.107, 5789–5801. 10.1007/s00253-023-12663-5 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Li, Y. et al. Cold stress in the harvest period: effects on tobacco leaf quality and curing characteristics. BMC Plant Biol.2110.1186/s12870-021-02895-w (2021). [DOI] [PMC free article] [PubMed]
  • 13.Tang, Z. et al. Climatic factors determine the yield and quality of Honghe flue-cured tobacco. Sci. Rep.1010.1038/s41598-020-76919-0 (2020). [DOI] [PMC free article] [PubMed]
  • 14.Wang, D. et al. A lightweight convolutional neural network for nicotine prediction in tobacco by near-infrared spectroscopy. Front. Plant Sci.1410.3389/fpls.2023.1138693 (2023). [DOI] [PMC free article] [PubMed]
  • 15.Zhu, Z. et al. A Long Short-Term Memory Neural Network Based Simultaneous Quantitative Analysis of Multiple Tobacco Chemical Components by Near-Infrared Hyperspectroscopy images. Chemosensors. 1010.3390/chemosensors10050164 (2022).
  • 16.Wei, K. et al. On-Line monitoring of the Tobacco Leaf Composition during Flue-Curing by Near-Infrared spectroscopy and deep transfer learning. Anal. Lett.55, 2089–2107. 10.1080/00032719.2022.2046021 (2022). [Google Scholar]
  • 17.Huang, J. et al. Application and comparison of several machine learning algorithms and their integration models in regression problems. Neural Comput. Appl.32, 5461–5469. 10.1007/s00521-019-04644-5 (2019). [Google Scholar]
  • 18.Qian, Y. Exploration of machine algorithms based on deep learning model and feature extraction. Math. Biosci. Eng.18, 7602–7618. 10.3934/mbe.2021376 (2021). [DOI] [PubMed] [Google Scholar]
  • 19.Zhang, X. et al. Sensory evaluation and prediction of bulk wine by physicochemical indicators based on PCA-PSO‐LSSVM method. Food Process. Preservation. 3, 46. 10.1111/jfpp.16343 (2022). [Google Scholar]
  • 20.Zhang, H. et al. Determination of soluble solids content in oranges using visible and near infrared full transmittance hyperspectral imaging with comparative analysis of models. Postharvest Biol. Technol.163, 111148 (2020). [Google Scholar]
  • 21.Liu, H., Duan, S. & Luo, H. Design and Temperature Modeling Simulation of the Full Closed Hot Air Circulation Tobacco Bulk Curing Barn. Symmetry 14, doi: (2022). 10.3390/sym14071300
  • 22.Meng, Y. et al. Analysis of the relationship between color and natural pigments of tobacco leaves during curing. Sci. Rep.14, 166 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Su, H. Data Research on Tobacco Leaf Image Collection Based on Computer Vision Sensor. J. Sens.2021 (1-11). 10.1155/2021/4920212 (2021).
  • 24.Ligor, M. & Buszewski, B. Study of Xanthophyll Concentration in Spinach leaves by Means of HPLC coupled with UV–VIS and Corona CAD detectors. Food. Anal. Methods. 5, 388–395. 10.1007/s12161-011-9256-7 (2011). [Google Scholar]
  • 25.Long, M. et al. Effect of different combined moistening and redrying treatments on the physicochemical and sensory capabilities of smoking food tobacco material. Drying Technol.36, 52–62. 10.1080/07373937.2017.1299752 (2017). [Google Scholar]
  • 26.Ji, X. et al. Quantitative determination of polyphenols in tobacco leaves by HPLC. Agric. Environ.11, 868–870 (2013). [Google Scholar]
  • 27.Nirmaan, A. M. C., Prasantha, R., Peiris, B. L. & B. D. & Comparison of microwave drying and oven-drying techniques for moisture determination of three paddy (Oryza sativa L.) varieties. Chem. Biol. Technol. Agric.710.1186/s40538-019-0164-1 (2020).
  • 28.Crandell, C. Continuous flow analysis the Auto-Analyzer. J. Autom. Chem.7, 145–148 (1985). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Mahmoud, B., El-Sayed, A. & Mohame, Y. A. An extractive-spectrophotometric method for the determination of nicotine. Planta Rnedica. 27, 140–144 (1975). [PubMed] [Google Scholar]
  • 30.Koistinen, J., Sjöblom, M. & Spilling, K. in Biofuels from Algae Methods in Molecular Biology 206, 81–86 (2019).
  • 31.Yin, C. et al. Influence of physicochemical characteristics on the effective moisture diffusivity in Tobacco. Int. J. Food Prop.18, 690–698. 10.1080/10942912.2013.845785 (2015). [Google Scholar]
  • 32.Chen, J. et al. Influences of different curing methods on chemical compositions in different types of tobaccos. Ind. Crops Prod.16710.1016/j.indcrop.2021.113534 (2021).
  • 33.Wang, G. et al. Regional differences of free amino acids during aging and their relationship with sensory and appearance quality of tobacco strips. J. ofSouthern Agric.21, 1176–1184. 10.3969/j.issn.2095-1191.2020.05.024 (2020). [Google Scholar]
  • 34.Huang, J. et al. Predictive modeling of blood pressure during hemodialysis: a comparison of linear model, random forest, support vector regression, XGBoost, LASSO regression and ensemble method. Comput. Methods Programs Biomed.195, 1–30. 10.1016/j.cmpb.2020.105536 (2020). [DOI] [PubMed] [Google Scholar]
  • 35.Yan, J. et al. Prediction of retention indices for frequently reported compounds of plant essential oils using multiple linear regression, partial least squares, and support vector machine. J. Sep. Sci.36, 2464–2471. 10.1002/jssc.201300254 (2013). [DOI] [PubMed] [Google Scholar]
  • 36.Meng, Y. et al. Relationship between heat/mass transfer and color change during drying process. J. Food Meas. Charact.16, 4151–4160. 10.1007/s11694-022-01497-w (2022). [Google Scholar]
  • 37.McGrath, T. E. et al. Phenolic compound formation from the low temperature pyrolysis of tobacco. J. Anal. Appl. Pyrol.84, 170–178. 10.1016/j.jaap.2009.01.008 (2009). [Google Scholar]
  • 38.Banožić, M. et al. Carbohydrates-Key players in Tobacco Aroma formation and quality determination. Molecules. 25, 1734. 10.3390/molecules25071734 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.JIN, Y. et al. Simultaneous Detection and Analysis of free amino acids and glutathione in different shrimp. Foods. 17, 11 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Zhao, S. et al. Determination of optimum humidity for air-curing of cigar tobacco leaves during the browning period. Ind. Crops Prod.18310.1016/j.indcrop.2022.114939 (2022).
  • 41.Zhang, Y. et al. Quantitative analysis of routine chemical constituents in tobacco by near-infrared spectroscopy and support vector machine. Spectrochim. Acta Part A Mol. Biomol. Spectrosc.71, 1408–1413. 10.1016/j.saa.2008.04.020 (2008). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The data presented in this study are available on request from the corresponding author (guoweimin1984@sina.com).

Articles from Scientific Reports are provided here courtesy of Nature Publishing Group

RESOURCES