Method for identifying producing area and variety of Yunnan Pu'er tea based on multi-modal analysis

Abstract

The invention discloses a method for identifying the producing area and variety of Yunnan Pu'er tea based on multi-modal analysis, which comprehensively utilizes physicochemical property detection, elemental analysis and GC * GC-TOFMS detection to detect the nutritional ingredients, elemental composition and volatile aroma substances of a sample, and combines PCA analysis, PLS-DA analysis and OPLS-DA analysis to identify the producing area and variety of the Yunnan Pu'er tea. Constructing a Pu'er tea production place and variety classification model based on multi-modal analysis; and according to the clustering position of the sample to be detected in the classification model, carrying out cross validation, and carrying out origin and variety identification on the sample to be detected. Through multi-dimensional data fusion, the accuracy and reliability of Pu'er tea production place traceability and variety identification are remarkably improved, the defect of insufficient information of a single detection method is overcome, and technical support is provided for Pu'er tea quality evaluation and market specification supervision.

Description

Technical Field

[0001] The present invention relates to the technical field of tea detection, and in particular to a method for identifying the origin and variety of Pu'er tea based on the fusion of physical, chemical, elemental and aroma multidimensional data. Background Art

[0002] Pu'er tea, a unique dark tea variety from China, has become highly sought-after in the global tea market in recent years due to its unique post-fermentation process, rich flavor, and potential health benefits. According to statistics, the annual output value of Pu'er tea has exceeded 10 billion yuan, making it a major category of Chinese tea exports. However, with surging market demand and the premium brand value created, the Pu'er tea industry faces severe challenges with adulteration and counterfeiting of origin. Some unscrupulous vendors profit from mixing in low-priced tea leaves, adding artificial colors or flavors, and falsifying origin labels, severely damaging consumer rights and the industry's reputation. Research indicates that approximately 30% of Pu'er tea on the market exhibits varying degrees of quality or origin fraud, posing a threat to consumer health and hindering the sustainable development of the Pu'er tea industry.

[0003] Currently, the authentication and traceability of Pu'er tea primarily relies on single-dimensional testing techniques, including sensory evaluation, physical and chemical analysis, elemental analysis, and aroma component analysis. However, these methods all have significant limitations when applied independently. Due to their single-dimensional nature and fragmented data, they struggle to address the complex adulteration challenges facing the Pu'er tea market, resulting in an overall authentication success rate of less than 80%. Sensory evaluation is a core method for traditional tea quality assessment, using a comprehensive score based on factors such as appearance, tea soup color, aroma, flavor, and tea leaf. However, this method is highly dependent on the evaluator's experience and is susceptible to individual preferences and environmental factors. For example, the bitterness and astringency of the same tea sample can be misinterpreted due to differences in brewing time or water temperature. Furthermore, sensory evaluation cannot quantify specific components in tea, making it difficult to address the sophisticated counterfeiting methods used by adulterated tea. Physical and chemical indicators such as moisture, ash content, water extract, and tea polyphenols are widely used in tea quality assessment. For example, water extract content reflects the body of the tea soup, while total catechin content is positively correlated with bitterness and astringency. However, these indicators are susceptible to processing and storage conditions. For example, artificial fermentation or high-temperature treatment may alter the oxidation state of tea polyphenols, leading to distorted test results. Furthermore, Pu'er tea from different origins exhibits minimal differences in conventional physical and chemical indicators, making accurate traceability difficult.

[0004] Elemental analysis can indirectly reflect the soil environmental characteristics of the origin by detecting the mineral and heavy metal content in tea. Studies have shown that differences in the dissolution rates of elements such as potassium, calcium, and manganese can help distinguish different origins. However, single element detection is easily interfered with by environmental pollution or fertilization measures. For example, adulterators may simulate high dissolution characteristics by adding exogenous potassium salts. In addition, although trace residues of heavy metals such as lead and cadmium pose a health risk, their content fluctuates greatly, making them difficult to use as stable traceability markers. Aroma is one of the core characteristics of Pu'er tea quality, and its volatile compounds can be accurately identified using technologies such as comprehensive two-dimensional gas chromatography time-of-flight mass spectrometry. However, aroma molecules are easily affected by factors such as storage humidity and microbial activity, resulting in a shift in the aroma profile of tea from the same origin. In addition, the addition of artificial flavors can mimic natural aroma molecules, making it difficult to distinguish authenticity using traditional gas chromatography-mass spectrometry techniques.

[0005] Existing research shows that a single detection technology can only cover one dimension of Pu'er tea's characteristics and cannot cope with the diversity and concealment of adulteration methods. For example, the combination of sensory and physical and chemical analysis can increase the accuracy of identification to 70%-75%, but it still cannot solve the problem of elemental or aroma imitation. In addition, the traditional combination of multiple methods often makes the interpretation of results difficult due to data redundancy or inconsistent analysis standards. For example, statistical models such as principal component analysis and partial least squares discriminant analysis perform well in single data sets, but the integration of cross-modal data lacks systematicity.

[0006] Therefore, it is necessary to provide a Pu'er tea identification method that can integrate multiple detection technologies and perform multimodal cross-validation to achieve complementary verification of multi-dimensional data, improve the accuracy of identification, and expand the scope of identification. Summary of the Invention

[0007] This paper proposes a Pu'er tea identification and traceability technology based on dual-system multimodal cross-validation. By integrating a physical and chemical elemental analysis system with an aroma and flavor analysis system, combined with a machine learning algorithm, this technology achieves complementary verification of multi-dimensional data, enabling accurate identification and classification of raw Pu'er tea.

[0008] The technical solution adopted in the present invention is:

[0009] A method for identifying the origin and variety of Yunnan Pu'er tea based on multimodal analysis, the method comprising the following steps:

[0010] (1) Data collection: Pu'er tea samples of various known origins or varieties were collected, and then physical and chemical tests, elemental analysis, and volatile aroma compound tests were performed according to the following steps; the samples to be tested were tested according to the same steps;

[0011] (2) Elemental analysis: Inductively coupled plasma tandem mass spectrometry (ICP-MS / MS) was used to detect the metal elements and their contents in tea leaves and tea soup respectively;

[0012] (3) Analysis of volatile aroma compounds: headspace solid phase microextraction (HS-SPME) combined with comprehensive two-dimensional gas chromatography-time of flight mass spectrometry (GC×GC-TOFMS) technology was used to detect volatile aroma compounds and their contents;

[0013] (4) Construct a classification model and analyze and process the sample data to be tested and the known sample data together:

[0014] The element content data detected in step (2) and the volatile aroma substance content data detected in step (3) were subjected to principal component analysis, partial least squares discriminant analysis (PLS-DA) and OPLS-DA analysis respectively, and a Pu'er tea origin and variety classification model based on multimodal analysis was constructed; according to the cluster position of the sample to be tested in the classification model, cross-validation was performed to identify the origin and variety of the sample.

[0015] In step (1), the number of known Pu'er tea samples of different origins or varieties is preferably more than 10, 10 to 30, or 10 to 20. The more types of known samples, the more accurate the classification of the samples to be tested.

[0016] The number of samples to be tested is not limited, for example, 10, 20, or even 100 samples to be tested are applicable. The classification model of the present invention can process data of multiple samples to be tested simultaneously.

[0017] Furthermore, a Pu'er tea origin and variety classification model based on multimodal analysis was constructed, including: PCA analysis of the element content data of tea leaves and tea soup, respectively, to construct PCA classification models of the element content of tea leaves and tea soup, respectively; OPLS-DA analysis of the content data of volatile aroma substances, to construct an OPLS-DA cluster analysis model.

[0018] Furthermore, in step (4), the step of analyzing the element content data and the volatile aroma substance content data includes:

[0019] a) Perform PCA dimensionality reduction, calculate the variance within the principal component groups and use confidence ellipse scatter plots to assess the intra-group variability of different origins, clarify the inherent distribution patterns between samples; and construct a PCA classification model for element content;

[0020] b) Based on PLS-DA, a classification model was constructed to screen key difference indicators with VIP values ​​greater than the screening threshold, and a hierarchical clustering heat map of key difference indicators was constructed to reveal the expression patterns and clustering characteristics of different regions or varieties;

[0021] The screening threshold is generally 1 to 2, which can be set freely according to actual conditions, with 1.5 being the preferred value;

[0022] c) Conduct OPLS-DA analysis; construct an OPLS-DA cluster analysis model.

[0023] Furthermore, element content data and volatile aroma substance content data need to be standardized before analysis and processing. Preferably, the standardization process uses Quantile normalization and automatic scaling to eliminate dimensional differences.

[0024] Furthermore, in step (4), for the sample to be tested, the identification step is:

[0025] A) Observe cluster positions: Observe the cluster positions of the test samples in the PCA classification model for the elemental content of tea leaves and tea soup, or the OPLS-DA cluster analysis model for the content of volatile aroma compounds. Accurately classify and identify the samples based on their clustering relationship with samples of known origin or variety.

[0026] B) If, in any classification model, the sample to be tested has highly consistent clustering characteristics with samples of a known origin or variety, and has discrete cluster locations with no intersection with other sample clusters, then the sample can be directly classified and identified;

[0027] C) If in any classification model, the cluster positions of the sample to be tested and two or more known samples overlap, the two classification models are cross-validated and classified according to their cluster positions.

[0028] Furthermore, in step B), if the sample to be tested is located in the distribution range of samples of a known origin or variety in any classification model and has no intersection with the distribution range of other samples, the sample can be directly classified and identified.

[0029] Furthermore, if the sample to be tested is located in the distribution range of samples of a known origin or variety in any classification model and has no intersection with the distribution range of other samples, and in other classification models, the sample to be tested is also located in the distribution range of the same known sample, then the variety or origin of the sample to be tested is the same as that of the known sample.

[0030] If the sample to be tested is located in the distribution range of samples of a known origin or variety in any classification model and has no intersection with the distribution range of other samples, but in other classification models, the distribution range of the sample to be tested does not intersect with the above-mentioned known samples, then the variety or origin of the sample to be tested does not belong to all the known samples involved in the analysis.

[0031] Further, for step C), if in the element content PCA classification model, the cluster positions of the sample to be tested and more than two known samples intersect, the OPLS-DA cluster analysis model of the volatile aroma substance content is further analyzed and classified according to its cluster position; if in the OPLS-DA cluster analysis model of the volatile aroma substance content, the cluster positions of the sample to be tested and more than two known samples intersect, the element content PCA classification model is further analyzed and classified according to its cluster position.

[0032] Generally speaking, any sample can be accurately classified by combining the PCA classification model of element content and the OPLS-DA cluster analysis model of volatile aroma substance content and cross-validation.

[0033] Furthermore, in step C), if in any classification model, the distribution range of the sample to be tested intersects with the distribution ranges of two or more known samples, the two classification models are cross-validated and classified according to their cluster positions.

[0034] Further, in a preferred embodiment, step C) can be performed as follows:

[0035] In the OPLS-DA cluster analysis diagram of the volatile aroma substance content, if the distribution range of the sample to be tested intersects with the distribution ranges of the known samples A1 and B1; in the PCA cluster analysis diagram of the tea element content, the distribution range of the sample to be tested intersects with the distribution ranges of the known samples A2 and B2; in the PCA cluster analysis diagram of the tea element content, the distribution range of the sample to be tested intersects with the distribution ranges of the known samples A3 and B3;

[0036] (a) There is one and only one common element X in the sets {A1, B1}, {A2, B2}, and {A3, B3}, i.e., {A1, B1}∩{A2, B2}∩{A3, B3}={X}, then the variety or origin of the identified sample to be tested is the same as the variety or origin of sample X; A1, B1, A2, B2, A3, and B3 represent known samples whose distribution ranges intersect with the distribution range of the sample to be tested in different cluster analysis diagrams, and they may be the same or different. However, there is only one sample X whose distribution ranges intersect with the distribution range of the sample to be tested in all three cluster analysis diagrams;

[0037] (b) There is no common intersection among the sets {A1, B1}, {A2, B2}, {A3, B3}, i.e. It is determined that the variety or origin of the sample to be tested does not belong to the known samples involved in the joint analysis, that is, there is no same sample whose distribution range intersects with the sample to be tested in the three cluster analysis diagrams.

[0038] Furthermore, the identification step further comprises step (6) physical and chemical testing: testing the moisture, ash, water extract, soluble protein, amino acids, sugars, tea polyphenols, catechins and caffeine contents of Pu'er tea samples of various known origins or varieties and the sample to be tested;

[0039] Based on the physical and chemical index data obtained from the testing of known samples, calculate the mean value of each component of the known samples;

[0040] The results of moisture, ash, water extract, soluble protein, amino acids, sugars, tea polyphenols, catechins and caffeine content of the test samples should fall within the range of ±50% of the mean value of the same components of known samples to assist in judgment. If the physical and chemical test indicators of the test samples deviate seriously from the mean value, it can be considered that there is an abnormality.

[0041] The Pu'er tea sample can be Pu'er raw tea or cooked tea. The present invention can identify the origin and / or variety of Pu'er raw tea or cooked tea.

[0042] In step (6), the physical and chemical tests include moisture, ash, water extract, soluble protein, amino acids, sugars and tea polyphenols content, and are performed according to conventional testing methods in the art.

[0043] In the step (2), in the elemental analysis, the metal elements and their contents in the tea leaves and tea soup are respectively detected, and the metal elements include one or more of Li, B, Na, Mg, Al, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Sr, Cd, Sn, Sb, Ba, Hg, and Pb.

[0044] Furthermore, the element content in the tea is generally determined by microwave digestion of tea samples and then ICP-MS / MS.

[0045] The element content in the tea soup is obtained by extracting the tea sample with boiling water and then filtering it to obtain the tea soup, and then detecting the element content therein, which is also called the element dissolution rate.

[0046] In the step (3), in the analysis of volatile aroma substances, HS-SPME is performed using a DVB / CAR / PDMS composite coating extraction head;

[0047] The processing steps of headspace solid phase microextraction are as follows: 0.25 g sample is added into 5 mL saturated saline solution, headspace extraction is carried out in a water bath at 95 °C for 30 minutes, the decomposition temperature is 250 °C, and the decomposition time is 10 minutes.

[0048] The conditions for comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry were:

[0049] The first-dimension column Rxi5sil Ms-5MS is connected in series with the second-dimension column Rtx-200;

[0050] Chromatographic conditions: carrier gas: high-purity helium (purity ≥99.999%), flow rate 1.0 mL / min; injection port temperature: 250°C; injection mode: splitless injection;

[0051] Temperature programming: First-dimension column (30 m × 0.25 mm × 0.25 μm): initial temperature 35 °C, hold for 5 min; increase to 65 °C at a rate of 2 °C / min; then increase to 140 °C at a rate of 6 °C / min; then increase to 180 °C at a rate of 3 °C / min; then increase to 200 °C at a rate of 10 °C / min; then increase to 300 °C at a rate of 30 °C / min and hold for 5 min;

[0052] Second-dimension column (1.79 m × 0.18 mm × 0.2 μm): parallel heating at a temperature 5°C higher than the first-dimension column; modulator compensation temperature: 10°C; modulation period: 7 s;

[0053] Mass spectrometry conditions:

[0054] Ionization mode: electron impact ionization source (EI); electron energy: 70 eV; ion source temperature: 240°C; transfer line temperature: 300°C; detector voltage: 1600 V; mass scan range: 35 to 550 u; acquisition frequency: 200 spectra / s.

[0055] In the step (3), headspace solid phase microextraction (HS-SPME) is combined with comprehensive two-dimensional gas chromatography-time of flight mass spectrometry (GC×GC-TOFMS) technology to perform qualitative and semi-quantitative analysis on volatile aroma substances;

[0056] The qualitative analysis refers to searching for compounds in a standard spectral library using mass spectrometry data and performing correction based on the retention index (RI); the RI value of the compound is calculated by injecting a mixed standard sample of normal alkanes (C7-C30);

[0057] The semi-quantitative analysis refers to calculating the relative content of volatile aroma substances using the peak area normalization method.

[0058] The present invention uses inductively coupled plasma mass spectrometry (ICP-MS / MS) to determine the macro- and trace elements in tea leaves and tea soup, accurately analyzes the trace elements, significantly improves the detection sensitivity, and analyzes their mineral composition and dissolution characteristics to reveal the characteristics of the production environment; at the same time, the volatile aroma components in the tea leaves are identified through headspace solid phase microextraction-comprehensive two-dimensional gas chromatography-time of flight mass spectrometry (HS-SPME-GC×GC-TOFMS) technology, screens key difference markers, and constructs an aroma characteristic map; finally, principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) are used to model the multi-source data of elements and aroma, generate a classification heat map, and realize high-precision identification and origin traceability of Pu'er raw tea.

[0059] The present invention can also use physical and chemical analysis methods to determine basic indicators such as moisture, ash, and water extract of tea, and combine it with high performance liquid chromatography (HPLC) to detect catechin and caffeine content, revealing the distribution of nutritional components and flavor substances in tea, and detecting abnormal deviations.

[0060] This invention organically combines physical and chemical testing, elemental analysis, and comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry (GC×GC-TOFMS) technology for the first time, constructing a multimodal analysis system. This system breaks through the limitations of traditional single detection methods and achieves a leap from a single dimension to a multi-dimensional collaboration in the identification and traceability of Pu'er tea. Using ICP-MS / MS technology to accurately determine the elemental distribution in tea leaves and tea soup, combined with PCA and heat map clustering, it achieves origin traceability based on elemental characteristics; using HS-SPME-GC×GC-TOFMS technology to identify volatile aroma molecules, screen out characteristic markers, and provide a scientific basis for aroma classification; through multi-source data fusion and PLS-DA modeling, a high-precision Pu'er tea classification model is constructed, which significantly improves the accuracy of identification and traceability reliability.

[0061] In this study, the physicochemical-elemental system accurately measures the nutritional composition and elemental dissolution characteristics of tea leaves, while the aroma-flavor system captures tens of thousands of volatile molecules through comprehensive two-dimensional gas chromatography, revealing unique aroma markers and accurately identifying trace adulterants such as artificial flavors. Cross-validation of the data from the two systems effectively distinguishes natural ingredients from artificial additives. Through complementary validation, the elemental data corrects for batch bias in aroma analysis, while the physicochemical indicators detect abnormal deviations, ensuring the reliability of the test results. Multidimensional data fusion effectively eliminates the limitations of a single technique, such as environmental interference or artifacts, and enhances interference resistance. For example, adulterated tea may resemble authentic tea in elemental profiles, but its aroma molecule clustering may deviate significantly. Processing multimodal data using omics analysis tools can extract key differentiating variables and construct an origin classification model. Experiments demonstrate that the origin classification model constructed in this study has high classification accuracy for 12 Pu'er tea varieties from Yunnan, significantly exceeding the results of single element or aroma analysis. The dual-system design effectively offsets interference from environmental or artifacts. For example, an adulterated sample simulates the characteristics of a specific origin by adding exogenous manganese elements, but the content of specific components in its aroma molecules is abnormally low, and the system automatically marks it as a suspicious sample.

[0062] The technology of the present invention is widely applicable to the analysis of tea leaves with different processes (raw Pu'er / cooked Pu'er), aging years and origins, and has high adaptability.

[0063] The present invention increases the success rate of identification and traceability from less than 80% to over 95% through dual-system multimodal cross-validation, providing efficient and accurate technical support for industry standardization supervision, quality certification, geographical indication protection, and consumer rights protection. Through the deep integration and intelligent analysis of multi-dimensional data, this technology can not only accurately identify adulteration, but also trace the true origin of tea, promoting the development of the Pu'er tea industry towards standardization and transparency. The present invention provides comprehensive and efficient technical support for the quality evaluation, authenticity identification, and origin certification of Pu'er tea, and has important scientific value and industrial application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 Figure 1 is a graph showing the element content in tea leaves and tea soup. Figure A shows the element content in tea leaves (ng / g); Figure B shows the element content in tea soup (ng / g).

[0065] Figure 2 This is the PCA cluster analysis diagram of element content in tea.

[0066] Figure 3 This is the PCA cluster analysis diagram of the element content in tea soup.

[0067] Figure 4Total ion current chromatograms of GC×GC-TOFMS after solid phase microextraction of tea leaves using four extraction heads.

[0068] Figure 5 This is the hierarchical heat map cluster analysis diagram of GC×GC-TOFMS analysis of 12 tea aroma compounds.

[0069] Figure 6 OPLS-DA cluster analysis diagram of omics analysis of 12 tea aroma compounds by GC×GC-TOFMS. DETAILED DESCRIPTION

[0070] The technical solution of the present invention is further described below in conjunction with embodiments, but the protection scope of the present invention is not limited thereto.

[0071] Example 1

[0072] Comprehensive characterization and classification of Yunnan Pu'er raw tea from 12 different origins: Baishahe, Baotang, Banpen, Manlashan, Yibang, Manzhuan, Zhuziqing, Luoshuidong, Fenghuangwo, Pusanghe, Wangong, and Hekai. Three replicate samples were collected from each origin.

[0073] 1. Physical and chemical properties

[0074] Physical and chemical analysis included determinations of moisture, ash, water extract, and major nutrients. Moisture content was determined by direct drying (drying at 103°C for 4 hours to constant weight), with an average of 8.09% for the 12 tea samples. Ash content was determined by calcining in a muffle furnace at 550°C for 8 hours, with an average of 5.51% for the 12 tea samples. Water extracts were extracted in a boiling water bath for 45 minutes, dried, and weighed, with a range of 17.52% to 41.76% for the 12 tea samples, with an average of 30.72%.

[0075] Nutritional testing includes:

[0076] 1. Soluble protein: Coomassie Brilliant Blue G-250 staining method, using bovine serum albumin (BSA) as the standard, 595 nm wavelength colorimetry, average value 4.49 mg / g;

[0077] 2. Soluble amino acids: ninhydrin colorimetric method, 579 nm wavelength, average value 25.25 mg / g;

[0078] 3. Soluble sugar: sucrose sulfate method, 620 nm wavelength colorimetry, average value 52.43 mg / g;

[0079] 4. Tea polyphenols: Folin phenol reduction method, 765nm wavelength determination, average value 139.81mg / g;

[0080] 5. Catechins and caffeine: HPLC-PDA method, Waters Arc column (C18, 5μm), mobile phase acetonitrile-acetic acid water gradient system, 278nm wavelength detection, the average total content of six catechins was 224.51μg / mL, caffeine was 87.95μg / mL. The average values ​​of the individual components in catechins were GA 2.13μg / mL, EGC 45.31μg / mL, C 18.16μg / mL, EGCG 66.56μg / mL,

[0081] EC 43.33μg / mL, ECG 49.03μg / mL.

[0082] The experimental results show that through the collaborative analysis of multiple indicators, such as the catechin / caffeine ratio, the robustness of identification can be enhanced, the influence of absolute content fluctuations can be eliminated, and the relative proportion characteristics can be highlighted. The ratio of caffeine to catechins (CCR) can be used as an important indicator for tea quality identification, origin traceability, and variety identification. The sample calculates the concentration ratio of caffeine / catechin = [caffeine] / (Σcatechin). Yunnan Pu'er tea has a higher catechin content and its CCR is generally lower. This is closely related to the altitude and light in the production environment of Pu'er tea. Fujian Wuyi Rock Tea has a higher caffeine content and its CCR is generally higher.

[0083] 2. Elemental analysis

[0084] Inductively coupled plasma mass spectrometry (ICP-MS / MS) was used to determine the element content in tea leaves and tea soup, revealing their mineral composition and dissolution characteristics.

[0085] The sample preparation process is as follows: take 0.3g of tea leaves, add 50mL of deionized water and boil for 40 minutes, filter with filter paper, take the supernatant, add 65% nitric acid to a final concentration of 2%, acidify, and prepare for detection on the machine.

[0086] A 0.5g sample of tea solids was added to 5mL of 65% nitric acid and subjected to microwave digestion using the following gradient heating program: 0-5min, heating to 120°C; 5-10min, maintaining at 120°C; 10-15min, heating to 150°C; 15-25min, maintaining at 150°C; 25-30min, heating to 190°C; and 30-50min, maintaining at 190°C. The digestion solution was then heated at 140°C to remove the acid, and the volume of the digestion solution was diluted to 25mL with water for analysis.

[0087] The ICP-MS / MS instrument parameters were set as follows: high-frequency transmission power 1550 W, sampling depth 7 mm, plasma gas flow rate 14.8 L / min, and nebulization chamber temperature 2 °C.

[0088] The results of element content in tea leaves and tea soup are as follows Figure 1 As shown in Figures A and B of the . Test results show that tea leaves contain the highest potassium (K) content (2.21 mg / g), followed by calcium (442.55 μg / g), magnesium (223.73 μg / g), and manganese (85.59 μg / g). The amount of potassium dissolved in tea soup reached 1.97 mg / g, while the dissolution rates of calcium, magnesium, and manganese were 66.2 μg / g, 150.07 μg / g, and 40.99 μg / g, respectively. Among heavy metal elements, lead (Pb) dissolved in tea soup at a high level (332.21 ng / g), but the dissolution rate of overall harmful elements (such as cadmium and arsenic) was lower than the residual amount in tea leaves, indicating that it is safe for daily consumption.

[0089] MetaboAnalyst 6.0 software was used to standardize the element content data of tea leaves and tea soups, and PCA cluster analysis was performed to obtain PCA cluster analysis diagrams of tea leaves and tea soups from different origins, as shown in the following figure. Figure 2 and Figure 3 MetaboAnalyst 6.0 can be used to perform Quantile normalization and automatic scaling on element and aroma data to eliminate dimensional differences.

[0090] Principal component analysis (PCA) and heatmap clustering revealed significant differences in the elemental profiles of teas from different origins. For example, Pusang River tea exhibits an outlier distribution in the PCA due to its high vanadium (V), iron (Fe), and chromium (Cr) content. Meanwhile, the prominent dissolution levels of zinc (Zn) and selenium (Se) in Manla Mountain tea soup serve as its classification markers. Elemental data, combined with environmental characteristics of the origin, can be used to construct a Pu'er tea traceability model, providing reliable indicators for origin identification. The technical advantage of elemental signature analysis is that it can effectively distinguish soil environmental differences between different origins, but single elements are susceptible to interference from fertilization or environmental pollution. For example, adulterators may add exogenous potassium salts to simulate high dissolution profiles, but aberrant samples can be identified using multi-element ratios (such as K / Ca). In normal samples, the ratios of potassium (K) and calcium (Ca) typically remain within a certain range due to the natural correlation between elements. This is due to their relatively stable occurrence and chemical behavior in natural sources. If adulterators add exogenous potassium salts, this will lead to an abnormally elevated K / Ca ratio.

[0091] 3. Aroma analysis

[0092] The volatile aroma compounds were detected by headspace solid phase microextraction-comprehensive two-dimensional gas chromatography-time of flight mass spectrometry (HS-SPME-GC×GC-TOFMS).

[0093] Four extraction heads were used for solid phase microextraction, combined with GC×GC-TOFMS, and the total ion current chromatogram was as follows: Figure 4Figure 2 shows ion chromatograms of a QC tea sample extracted using solid-phase microextraction (SPME) using four extraction tips on a comprehensive two-dimensional GC-MS / MS system. Figure A shows a 50 / 30 μm DVB / CAR / PDMS StableFlex / SS (2 cm) 57348-U; Figure B shows a 65 μm PDMS / DVB Fused Silica / SS 57310-U; Figure C shows a 100 μm PDMS Fused Silica / SS 57300-U; and Figure D shows a 75 μm Carboxen / PDMS Fused Silica / SS 57318.

[0094] The sample processing procedure is as follows: weigh 0.25 g of tea leaves and add 5 mL of saturated salt water, perform headspace extraction in a water bath at 95 °C for 30 minutes, and perform desorption at 250 °C for 10 minutes.

[0095] The conditions for comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry were:

[0096] The first-dimension column Rxi5sil Ms-5MS is connected in series with the second-dimension column Rtx-200;

[0097] Chromatographic conditions: carrier gas: high-purity helium (purity ≥99.999%), flow rate 1.0 mL / min; injection port temperature: 250°C; injection mode: splitless injection;

[0098] Temperature programming: First-dimension column (30 m × 0.25 mm × 0.25 μm): initial temperature 35 °C, hold for 5 min; increase to 65 °C at a rate of 2 °C / min; then increase to 140 °C at a rate of 6 °C / min; then increase to 180 °C at a rate of 3 °C / min; then increase to 200 °C at a rate of 10 °C / min; then increase to 300 °C at a rate of 30 °C / min, hold for 5 min; Second-dimension column (1.79 m × 0.18 mm × 0.2 μm): parallel temperature increase at a temperature 5 °C higher than that of the first-dimension column; modulator compensation temperature: 10 °C; modulation period: 7 s;

[0099] Mass spectrometry conditions:

[0100] Ionization mode: electron impact ionization source (EI); electron energy: 70 eV; ion source temperature: 240°C; transfer line temperature: 300°C; detector voltage: 1600 V; mass scan range: 35 to 550 u; acquisition frequency: 200 spectra / s.

[0101] The results show that Figure A presents a uniform peak during the desorption time of the entire heating program, indicating that compared with the other three solid phase extraction heads, the DVB / CAR / PDMS composite coating adsorbs a richer variety of tea aroma substances.

[0102] The DVB / CAR / PDMS composite coating combines the properties of three materials: DVB (divinylbenzene) has high adsorption capacity, CAR (carboxyl) provides selectivity, and PDMS (polydimethylsiloxane) has versatility. This composite coating can simultaneously extract volatile and semi-volatile compounds and is suitable for C3-C20 compounds with a molecular weight between 40-275. This makes it excellent in analyzing complex samples such as tea, capable of simultaneously extracting polar and non-polar compounds. After optimizing the extraction conditions, a 50 / 30μm DVB / CAR / PDMS composite coating extraction head was selected. Its wide polarity range can efficiently capture acids, alcohols, aldehydes, ketones, and ester compounds in Pu'er tea.

[0103] Data processing and analysis

[0104] Data processing: ChromaTOF software (version: 5.50, Leco Corp. for Windows) was used to process GC×GC-TOFMS data, including peak extraction, deconvolution, baseline correction, and automatic integration.

[0105] Qualitative analysis: Compound search was performed based on the NIST spectral library (version: NIST17) and calibration was performed using retention index (RI). RI values ​​for the compounds were calculated by injecting a standard mixture of normal alkanes (C7-C30).

[0106] Semi-quantitative analysis: The relative content of volatile aroma compounds was calculated using the peak area normalization method.

[0107] 4. Build a classification model

[0108] A total of 11,393 volatile molecules were identified. The elemental and aroma data were Quantile normalized and automatically scaled using MetaboAnalyst 6.0 to eliminate dimensional differences. PLS-DA analysis was then performed, and 25 characteristic markers were screened out based on the criteria of P < 0.05 and VIP > 1.5.

[0109] In this embodiment, the characteristic markers of the 12 tea samples are the following 25 compounds: cyclopentaneacetic acid; 2-methyl-benzoic acid; Benzylnitrile; 2-methyl-nonadecane; Acetophenone; Ethyl citrate; 2,2-dimethylpropionic acid; Dodecan-1-yl acetate; acetate); octadecanenitrile; 2,6-dimethyl-phenol; 1,3-pentadiene, 2-methyl-; 1-methoxy-benzene; (1-butylheptyl)-benzene; 3-ethyl-1-hexene-4-yne Bicyclo[3.1.0]hexane; 2,4-Dimethyl-1-heptene; Undecane, 2,6-dimethyl-; Benzaldehyde, 2-hydroxy-; 2H-1-Benzopyran-2-one; Cyclolongifolene oxide; Spiro[4.5]dec-6-ene; Propanoic acid, 3-methyl-; Butanedioic acid; Pentane, 1-nitro-; N-Dodecylmethylamine.

[0110] Based on the above 25 characteristic markers, a hierarchical clustering heat map is drawn, such as Figure 5 As shown, the expression patterns and clustering characteristics of tea leaves from different origins are revealed.

[0111] Furthermore, the volatile aroma compound data were subjected to OPLS-DA analysis, and the OPLS-DA cluster analysis diagram is shown in the figure below. Figure 6 shown.

[0112] Figure 6The results showed that through heat map clustering and VIP scoring, molecules such as 5,5-dimethyl-1,3-hexadiene and cyclohexanone derivatives were further confirmed as key classification indicators, significantly improving the accuracy of aroma tracing.

[0113] Comprehensive two-dimensional chromatography offers over 10 times the separation power of traditional GC-MS, enabling the detection of over 10,000 volatile molecules. However, aroma is susceptible to storage conditions, which can lead to variations between samples from the same origin but different batches. This paper further incorporates elemental data (such as elements characteristic of origin) and applies a multimodal, dual-scale model to correct for batch variations in aroma compounds.

[0114] Specifically, through the element markers, element PLS-DA cluster analysis diagram, aroma characteristic markers, hierarchical clustering heat map of aroma substance analysis and OPLS-DA cluster analysis diagram, the two classification models of different scales were combined to classify the origins of the 12 tea samples involved in the analysis.

[0115] The present invention has found through verification that combining the classification models of the two analysis methods and cross-validation can achieve mutual separation of all tea samples with high classification accuracy.

[0116] For example, in the OPLS-DA dimensionality reduction analysis of aroma compounds, the Poussin River and Luoshuidong, as well as the Poussin River and Hekai, overlap. However, in the elemental PCA analysis of tea leaves and tea soup, the Poussin River and Luoshuidong, as well as the Poussin River and Hekai, are clearly separated. Therefore, combining aroma and elemental data can significantly improve classification accuracy.

[0117] 5. The samples to be tested are tested and analyzed in the same way as steps 1, 2, 3, and 4 for known samples. When performing principal component analysis, partial least squares discriminant analysis (PLS-DA), and OPLS-DA analysis, the data of the samples to be tested are input together with the data of all known samples for cluster analysis. The origin and variety of the samples to be tested are identified based on their cluster position in the classification model.

[0118] The specific identification steps are:

[0119] A) Observe cluster positions: Observe the cluster positions of the test samples in the PCA cluster analysis diagram of the element content of tea leaves and tea soup, or the OPLS-DA cluster analysis diagram of the volatile aroma compound content; accurately classify and identify the samples based on their clustering relationship with samples of known categories.

[0120] B) If, in any classification model, the sample to be tested has highly consistent clustering characteristics with a certain class of known samples and has no intersection with other sample clusters, it can be preliminarily judged to belong to that class;

[0121] C) If in any classification model, the cluster positions of the sample to be tested and two or more known samples overlap, the two classification models are cross-validated and classified according to their cluster positions.

[0122] Furthermore, in step B), if the sample to be tested is located in the distribution range of samples of a known origin or variety in any classification model and has no intersection with the distribution range of other samples, the sample can be directly classified and identified.

[0123] Furthermore, if the sample to be tested is located in the distribution range of samples of a known origin or variety in any classification model and has no intersection with the distribution range of other samples, and in other classification models, the sample to be tested is also located in the distribution range of the same known sample, then the variety or origin of the sample to be tested is the same as that of the known sample.

[0124] If the sample to be tested is located in the distribution range of samples of a known origin or variety in any classification model and has no intersection with the distribution range of other samples, but in other classification models, the distribution range of the sample to be tested does not intersect with the above-mentioned known samples, then the variety or origin of the sample to be tested does not belong to all the known samples involved in the analysis.

[0125] Furthermore, in step C), if in the PCA classification model of the element content of tea leaves and tea soup, the cluster positions of the sample to be tested and more than two known samples overlap, the OPLS-DA cluster analysis model of the volatile aroma substance content is further analyzed and classified according to its cluster position; if in the OPLS-DA cluster analysis model of the volatile aroma substance content, the cluster positions of the sample to be tested and more than two known samples overlap, the PCA classification model of the element content of tea leaves and tea soup is further analyzed and classified according to its cluster position.

[0126] Furthermore, in step C), if in any classification model, the distribution range of the sample to be tested intersects with the distribution ranges of two or more known samples, the two classification models are cross-validated and classified according to their cluster positions.

[0127] Further, in a preferred embodiment, step C) can be performed as follows:

[0128] In the OPLS-DA cluster analysis diagram of the volatile aroma substance content, if the distribution range of the sample to be tested intersects with the distribution ranges of the known samples A1 and B1; in the PCA cluster analysis diagram of the tea element content, the distribution range of the sample to be tested intersects with the distribution ranges of the known samples A2 and B2; in the PCA cluster analysis diagram of the tea element content, the distribution range of the sample to be tested intersects with the distribution ranges of the known samples A3 and B3;

[0129] (a) There is one and only one common element X in the sets {A1, B1}, {A2, B2}, and {A3, B3}, i.e., {A1, B1}∩{A2, B2}∩{A3, B3}={X}, then the variety or origin of the identified sample to be tested is the same as the variety or origin of sample X; A1, B1, A2, B2, A3, and B3 represent known samples whose distribution ranges intersect with the distribution range of the sample to be tested in different cluster analysis diagrams, and they may be the same or different. However, there is only one sample X whose distribution ranges intersect with the distribution range of the sample to be tested in all three cluster analysis diagrams;

[0130] (b) There is no common intersection among the sets {A1, B1}, {A2, B2}, {A3, B3}, i.e. It is determined that the variety or origin of the sample to be tested does not belong to the known samples involved in the joint analysis, that is, there is no same sample whose distribution range intersects with the sample to be tested in the three cluster analysis diagrams.

[0131] Generally speaking, any sample can be accurately classified by combining the elemental content PCA classification model with the OPLS-DA cluster analysis model for volatile aroma compound content, and cross-validating the results. For example, if there is an intersection in the elemental PCA cluster analysis of tea leaves and tea soup, combining it with the OPLS-DA cluster analysis of volatile aroma compound content will eliminate the intersection, thus enabling accurate classification.

[0132] D) The test results of the moisture, ash, water extract, soluble protein, amino acids, carbohydrates, tea polyphenols, catechins, and caffeine content of the test sample should all fall within the range of ±50% of the mean of the known samples with the same components to assist in judgment. If the physical and chemical test indicators of the test sample deviate significantly from the mean, it can be considered an abnormality.

[0133] For example, if in the PCA classification model of element content, the clustering characteristics of the sample to be tested are highly consistent with those of a certain type of known sample A, and the cluster positions with other samples are discrete and have no intersection; and in the OPLS-DA cluster analysis model of volatile aroma substance content, the clustering characteristics of the sample to be tested are highly consistent with those of a certain type of known sample B, and the cluster positions with other samples are discrete and have no intersection; A and B are different known samples; then the sample is judged to be a new tea sample and cannot be classified into the origin or variety of the currently available known samples.

[0134] The present invention's research shows that combining the classification models of these two analysis methods can significantly improve the accuracy of classification and achieve accurate identification of various tea samples.

[0135] For example, in the identification of Pu'er tea, if the sample to be tested is classified into the same category as a sample of a known origin or variety in the cluster analysis diagram, it indicates that they may have similar origin or variety characteristics.

[0136] The implementation case effect takes Baisha River and Manzhuan tea samples as examples, and the multimodal analysis results show that according to Figure 1 and Figure 6 , which can be directly classified. And according to the results of OPLS-DA cluster analysis, the characteristic aroma molecule of Baisha River is acetophenone, and the characteristic aroma molecule of Manzhuan is 2,4-dimethyl-1-heptene.

[0137] (2,4-Dimethyl-1-heptene); the two can be quickly distinguished using the combined element-aroma-physicochemical model, validating the practicality and reliability of the proposed method. Furthermore, the results of physical and chemical testing for water extract, soluble protein, amino acids, carbohydrates, tea polyphenols, catechins, and caffeine all fell within ±50% of the mean of the known samples for the same component.

[0138] Example of sample testing:

[0139] Take 15 types of Pu'er tea samples and analyze the physical and chemical indicators, metal element content and volatile aroma substances according to steps 1, 2 and 3. Then, follow step 4 to perform cluster analysis on all 15 test sample data and the data of 12 known origin samples. The identification results and the actual sample results are shown in the following table:

[0140]

[0141]

[0142]

[0143] It can be seen that the multimodal analysis identification and classification model of the present invention has a 100% accuracy rate in origin identification, and has identified two abnormal situations and detected a new origin.

[0144] 6. Multimodal Analysis Integration and Conclusion This paper uses multimodal analysis technology to integrate elemental, aroma, and physical and chemical data to construct a comprehensive characteristic map of Pu'er tea. The specific process is as follows:

[0145] 1) Data standardization: MetaboAnalyst 6.0 was used to perform Quantile normalization and automatic scaling on element and aroma data to eliminate dimensional differences;

[0146] 2) Dimensionality reduction analysis: Principal component analysis (PCA) revealed clustering trends in tea elements and aroma. For example, the Pusang River deviated from the main group due to its heavy metal characteristics, while the Manla Mountain formed an independent branch due to its prominent zinc and selenium dissolution.

[0147] 3) Discriminant model: Partial least squares discriminant analysis (PLS-DA) was used to screen out aroma VIP molecules and elemental markers, and an OPLS-DA cluster analysis model was established for 12 types of tea. This was cross-validated with the elemental PCA cluster analysis model for tea leaves and tea soups, and the classification accuracy was over 95%.

[0148] The present invention uses multimodal collaborative analysis technology to achieve a leap from a single dimension to multi-dimensional collaboration in Pu'er tea identification and traceability. Through complementary verification, elemental data can correct batch deviations in aroma analysis, and physical and chemical indicators can detect abnormal deviations, ensuring the reliability of the test results. High-sensitivity full two-dimensional gas chromatography-time-of-flight mass spectrometry (GC×GC-TOFMS) is used to detect 11,393 aroma molecules, accurately identifying trace adulteration ingredients such as artificial flavors. At the same time, inductively coupled plasma mass spectrometry (ICP-MS / MS) is combined to accurately analyze trace elements, significantly improving detection sensitivity. Multi-dimensional data fusion effectively offsets the limitations of a single technology, such as environmental interference or human forgery, and enhances anti-interference capabilities. This technology is widely applicable to the analysis of tea leaves with different processes (raw / cooked), aging years, and origins, and is highly adaptable. By integrating physical and chemical, elemental and aroma data, combining multimodal analysis and cross-validation, the identification accuracy has been improved by more than 20%, providing efficient and precise technical support for Pu'er tea market supervision, quality certification and geographical indication protection, providing technical support for tea industry standardization and brand protection, and effectively promoting the standardization and sustainable development of the tea industry.

Claims

1. A method for identifying the origin and variety of Yunnan Pu'er tea based on multimodal analysis, characterized in that The method comprises the following steps: (1) Data collection: Pu'er tea samples of various known origins or varieties were collected, and then elemental analysis and volatile aroma compound detection were performed according to the following steps; the samples to be tested were tested according to the same steps; (2) Elemental analysis: Inductively coupled plasma tandem mass spectrometry was used to detect the metal elements and their contents in tea leaves and tea soup; (3) Analysis of volatile aroma compounds: headspace solid phase microextraction combined with comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry was used to detect volatile aroma compounds and their contents; (4) Construct a classification model and analyze and process the sample data to be tested and the known sample data together: The element content data of tea leaves and tea soup detected in step (2) and the content data of volatile aroma substances detected in step (3) were subjected to PCA analysis, PLS-DA analysis and OPLS-DA analysis respectively, and a Pu'er tea origin and variety classification model based on multimodal analysis was constructed; according to the cluster position of the sample to be tested in the classification model, cross-validation was performed to identify the origin and variety of the sample.

2. The method according to claim 1, wherein In the step (4), constructing a Pu'er tea origin and variety classification model based on multimodal analysis includes: performing PCA analysis on the element content data of tea leaves and tea soup, respectively, and constructing PCA classification models of the element content of tea leaves and tea soup, respectively; performing OPLS-DA analysis on the content data of volatile aroma substances, and constructing an OPLS-DA cluster analysis model.

3. The method according to claim 2, wherein For the sample to be tested, the identification steps are as follows: A) Observe cluster positions: Observe the cluster positions of the test samples in the PCA classification model for the elemental content of tea leaves and tea soup, or the OPLS-DA cluster analysis model for the content of volatile aroma compounds; classify and identify the samples based on their clustering relationships with samples of known origin or variety; B) If, in any classification model, the sample to be tested has highly consistent clustering characteristics with samples of a known origin or variety, and has discrete cluster locations with no intersection with other sample clusters, the sample is directly classified and identified; C) If in any classification model, the cluster positions of the sample to be tested and two or more known samples overlap, the two classification models are cross-validated and classified according to their cluster positions.

4. The method according to claim 3, wherein In the step C), if the cluster positions of the test sample and two or more known samples overlap in the PCA classification model of the element content of tea leaves and tea soup, the OPLS-DA cluster analysis model of the volatile aroma substance content is further analyzed and classified according to their cluster positions; if the cluster positions of the test sample and two or more known samples overlap in the OPLS-DA cluster analysis model of the volatile aroma substance content, the PCA classification model of the element content of tea leaves and tea soup is further analyzed and classified according to their cluster positions.

5. The method according to claim 1, wherein In the step (2), in the elemental analysis, the metal elements and their contents in the tea leaves and tea soup are respectively detected, and the metal elements include one or more of Li, B, Na, Mg, Al, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Sr, Cd, Sn, Sb, Ba, Hg, and Pb.

6. The method according to claim 1, wherein In the step (3), in the analysis of volatile aroma substances, a DVB / CAR / PDMS composite coating extraction head is used for headspace solid phase microextraction; The processing steps of headspace solid phase microextraction are as follows: 0.25 g sample is added into 5 mL saturated saline solution, headspace extraction is carried out in a water bath at 95 °C for 30 minutes, the decomposition temperature is 250 °C, and the decomposition time is 10 minutes.

7. The method according to claim 1, wherein In step (3), the conditions for comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry are: The first-dimension column Rxi5sil Ms-5MS is connected in series with the second-dimension column Rtx-200; Chromatographic conditions: carrier gas: high-purity helium, flow rate 1.0 mL / min; injection port temperature: 250 °C; injection mode: splitless injection; Temperature programming: First-dimension column: initial temperature 35°C, hold for 5 min; increase to 65°C at a rate of 2°C / min; then increase to 140°C at a rate of 6°C / min; then increase to 180°C at a rate of 3°C / min; then increase to 200°C at a rate of 10°C / min; then increase to 300°C at a rate of 30°C / min, hold for 5 min; Second-dimension column: increase the temperature in parallel at a temperature 5°C higher than that of the first-dimension column; Modulator compensation temperature: 10°C; Modulation period: 7 s; Mass spectrometry conditions: Ionization mode: electron bombardment ionization source; electron energy: 70 eV; ion source temperature: 240°C; transfer line temperature: 300°C; detector voltage: 1600 V; mass scan range: 35 to 550 u.

8. The method according to claim 1, wherein In the step (3), headspace solid phase microextraction combined with comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry is used to perform qualitative and semi-quantitative analysis on the volatile aroma substances; The qualitative analysis refers to searching for compounds in a standard spectral library using mass spectrometry data and performing correction based on retention indices; The semi-quantitative analysis refers to calculating the relative content of volatile aroma substances using the peak area normalization method.

9. The method according to claim 1, wherein The method further comprises step (6) physical and chemical testing: testing the moisture, ash, water extract, soluble protein, amino acids, sugars, tea polyphenols, catechins and caffeine contents of Pu'er tea samples of known origins or varieties and the sample to be tested; Based on the physical and chemical index data obtained from the testing of known samples, calculate the mean value of each component of the known samples; The moisture, ash, water extract, soluble protein, amino acids, sugars, tea polyphenols, catechins and caffeine content results of the test samples should fall within the range of ±50% of the mean value of the same components of the known samples to assist in judgment.

Images