A method and system for identifying storage years of white tea based on spectral peak fitting

Abstract

This invention discloses a method and system for identifying the storage age of white tea based on spectral peak fitting, belonging to the field of nuclear magnetic resonance detection technology. The construction steps of the system include: (1) analyzing the one-dimensional nuclear magnetic resonance hydrogen spectrum of the white tea sample. 1 (1) Preprocess H NMR data; (2) Obtain white tea metabolite standards. 1 H NMR spectrum and its peak information; (3) using the peak information of white tea metabolite standards to analyze white tea samples 1 (4) Fit the H NMR spectrum of the white tea sample; 1 (5) Fit the remaining peaks in the H NMR spectrum that do not match the peaks of the metabolite standard; merge the fitting results of the metabolite standard and the unmatched peaks to construct a storage year identification model for each variety of white tea. This system maximizes the use of the hydrogen spectrum information of the sample by fitting the one-dimensional nuclear magnetic resonance hydrogen spectrum of the white tea sample, providing more comprehensive and reliable information for the identification of the storage year of white tea, and at the same time providing a new method for the analysis of biological nuclear magnetic resonance spectra.

Description

Technical Field

[0001] This invention relates to the field of nuclear magnetic resonance metabolomics, and in particular to a method and system for identifying the storage age of white tea based on peak fitting. Background Technology

[0002] Nuclear magnetic resonance (NMR) is a very powerful analytical tool. It uses the behavior of atomic nuclei in an applied magnetic field to obtain information about the structure and properties of matter. NMR technology can perform unbiased and non-destructive testing on biological samples, and it has many advantages such as simple sample preparation, high quantitative accuracy, and good reproducibility. It has been widely used in chemistry, biomedicine, materials science, and other fields.

[0003] 1H NMR spectrum ( 1 ¹H NMR is an important analytical technique for the composition of complex biological samples. However, complex biological samples contain a wide variety of biomolecules, leading to one-dimensional... 1 The ¹H NMR spectra exhibited peak crowding and severe peak overlap. Furthermore, factors such as pH, ion concentration, and experimental temperature caused random chemical shifts in the compound peaks across different samples. These inconsistencies and instabilities in sample and experimental conditions presented significant challenges to metabolite attribution and quantification, as well as subsequent statistical analysis.

[0004] Currently based on 1 H NMR data preprocessing mainly falls into two categories: The first is targeted analysis, which uses characteristic peaks to quantify known metabolites. However, in practical applications, due to peak overlap, it is difficult to find characteristic peaks for many metabolites. Furthermore, due to noise, the accuracy of quantification based on individual characteristic peaks is difficult to guarantee. The second method is segmented integration. This method, based on peak alignment, integrates the sample data at equal intervals. 1High-resolution one-dimensional proton NMR spectra are segmented and integrated, with each integration bin considered as a feature for subsequent statistical analysis [Mercier P, Lewis MJ, Chang D, et al. Towards automatic metabolomic profiling of high-resolution one-dimensional proton NMR spectra [J]. Journal of biomolecular NMR, 2011, 49: 307-323]. However, due to the complex composition of biological samples, the wide variety of compounds, and the inconsistent chemical shifts of resonance peaks among different compounds, current peak alignment methods based on the assumption that "peak shift does not change the chemical shift order of peaks" cannot completely correct for peak shifts of compounds. This results in the same resonance peak being segmented into different bins in different samples, making features incomparable and affecting the interpretability of subsequent analytical results. Furthermore, when a segment contains multiple overlapping peaks, their information interferes with each other, also significantly impacting subsequent analytical results.

[0005] White tea is rich in beneficial bioactive substances, which have antioxidant, anti-inflammatory, anti-tumor, and cardiovascular protective effects. Furthermore, some active ingredients in white tea increase with storage time, hence the saying, "One year tea, three years medicine, seven years treasure." The market price of white tea varies significantly depending on its storage year, leading some unscrupulous merchants to take advantage of this by selling inferior products. Therefore, establishing a reliable model for determining storage age is urgent and important for tea consumers, traders, and regulatory authorities. White tea contains a large number of primary and secondary metabolites, and the content of these metabolites changes subtly with increasing storage time, which is closely related to the quality of white tea. One-dimensional nuclear magnetic resonance hydrogen spectroscopy, with its unbiasedness, robustness, speed, and efficiency, has become an indispensable tool for tea quality analysis [Maria de Fátima CS, Rech KS, Dutra LM, et al.]. 1 HR-MAS NMR chemical profile and chemometric analysis as a tool for quality control of different cultivars of green tea (Camellia sinensis) [J]. Food Chemistry, 2023, 408: 1350-16.]. However, due to problems such as peak drift, peak overlap, and noise, traditional methods based on targeted analysis or piecewise integration are limited.1 The H NMR spectral data processing method is extremely challenging for identifying and predicting the storage age of white tea. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing methods and propose a strategy based on overall spectrum fitting of known metabolites and single-peak fitting of unknown metabolites. Based on the spectrum peak fitting of known and unknown metabolites in the sample, regression models of fitting data and storage years are established for different varieties of white tea, forming a storage year identification system for multiple varieties of white tea.

[0007] The technical solution adopted in this invention is:

[0008] A system for identifying the storage age of white tea based on spectral peak fitting, the construction method of which includes the following steps:

[0009] (1) Preprocessing of one-dimensional nuclear magnetic resonance (NMR) hydrogen spectrum data: The one-dimensional nuclear magnetic resonance (NMR) hydrogen spectrum of the white tea sample was preprocessed. 1 The free induction decay (FID) signals from H NMR were preprocessed using fast Fourier transform, phase correction, baseline correction, chemical shift calibration, and peak alignment to obtain samples of white tea from different storage years and varieties. 1 H NMR spectral data;

[0010] (2) Metabolite standards 1 H NMR spectral information acquisition: Acquire or download standards of known metabolites of white tea. 1 HNMR spectroscopy is used to obtain information on the splitting mode, chemical shift, linewidth, and peak height of each metabolite spectrum, and to statistically analyze the area ratio between each metabolite peak.

[0011] (3) Sample 1 Abundance fitting of known metabolites in H NMR spectra: matching the peaks of known metabolite standards with those of white tea samples. 1 By comparing, matching, and fitting the H NMR spectra, the white tea samples were calculated. 1 Abundance of known metabolites in H NMR spectra;

[0012] (4) Sample 1 Abundance fitting of mismatched peaks in H NMR spectra: for white tea samples 1 The unassigned peaks in the H NMR spectrum were fitted one by one to calculate the abundance of the unmatched peaks in the white tea sample.

[0013] (5) Construction of storage year identification system: Combine the fitting abundance of known metabolites and unmatched peaks, construct the storage year regression model of each variety of white tea, and complete the construction of white tea storage year identification system based on spectral peak fitting.

[0014] Further, in step (1), the white tea sample was processed as follows: the finished white tea was ground into fine powder using a grinder and passed through a 100-mesh sieve; in the sample preparation for the NMR experiment, 1g of tea powder was mixed with 5 mL of methanol / water solution (70 / 30, v / v), stirred at 70℃ for 30 min, and non-volatile compounds were retained; after cooling to room temperature, it was centrifuged at 4000 rpm for 5 min, and 2.5 mL of supernatant was transferred to a 10 mL centrifuge tube, and residual methanol was removed in a nitrogen blower; finally, it was freeze-dried in a vacuum dryer at -80 ℃ to obtain dry powder of tea metabolites; the dry powder of metabolites was dissolved in 700 μL of DMSO-d6 solution, vortexed for 30s, rotated at 1000 rpm, and centrifuged at 4℃ for 10 min; 600 μL of supernatant was transferred to a nuclear magnetic resonance tube and stored in a 4°C refrigerator until the NMR experiment.

[0015] Further, in step (1), the nuclear magnetic resonance conditions are: equipped with a TXI suitable for reverse detection. 1 H / 13 C / 15 Samples were acquired on an 850 MHz Bruker Avance III nuclear magnetic resonance spectrometer with an N 5mm cryogenic probe. 1 H NMR spectra were obtained using the NOESY pulse sequence. After calibrating the 90-degree pulse width in the sequence, the experimental temperature was set to 298 K, the relaxation delay τ = 1.0 s, the acquisition time was 1.8 seconds, the spectral width was 17000 Hz, and sampling was performed with 16 accumulations and 2 empty scans. The sampling time for each spectrum was 3 minutes.

[0016] Further, the one-dimensional nuclear magnetic resonance hydrogen spectrum data preprocessing in step (1) includes: padding the collected white tea sample nuclear magnetic resonance FID data with zeros to integer powers of 2, and then performing a fast Fourier transform to obtain the white tea sample's... 1 The ¹H NMR spectrum was then processed, including phase correction, baseline correction, chemical shift calibration, peak alignment, and sample normalization, to obtain the white tea sample. 1 H NMR spectral data.

[0017] Furthermore, the metabolite standards in step (2) 1 The acquisition of H NMR spectral information includes: collecting standards for known metabolites that can be attributed to white tea samples. 11H NMR spectra; for metabolites for which standards could not be obtained, their chromatograms can be downloaded from public databases such as HMDB, NP-MRD, and FooDB. 1 1H NMR experimental spectra; by fitting the peaks, information on the composition of the sample is obtained, including the splitting mode, chemical shift, linewidth and peak height of each peak, and the integral area ratio of each peak is calculated.

[0018] Furthermore, the sample in step (3) 1 Fitting the abundance of known metabolites in ¹H NMR spectra includes: fine-tuning the chemical shifts and linewidths of each peak in the metabolite standard while keeping the peak areas constant, so that they match those of the white tea sample. 1 Alignment and matching of corresponding peaks in the HNMR spectrum; and fitting of white tea samples by amplifying or reducing the overall signal intensity of the metabolite standard spectrum. 1 The corresponding matching peaks in the H NMR spectrum are used as the abundance of the corresponding metabolites in the white tea sample.

[0019] Furthermore, the sample in step (4) 1 Abundance fitting of mismatched peaks in H NMR spectra includes: for white tea samples 1 Peaks in the H NMR spectrum that fail to match the standard spectrum of the metabolite are considered as unknown metabolites; Lorentz line peaks are sampled and fitted one by one to the unmatched peaks, and the fitted area is used as the abundance of the corresponding unknown metabolite.

[0020] Furthermore, in step (4), for the peaks of the hydrogen spectrum of the white tea samples that could not be assigned, the signal-to-noise ratio of each unknown peak is calculated according to the following formula.

[0021] SNR=SI / SD

[0022] Where SI is the signal intensity of the unmatched peak, and SD is the standard deviation of the noise region in the same sample spectrum;

[0023] Using the Lorentz linear function, i.e. Where A and B are constants, the unmatched peaks with SNR>5 are fitted, and the unmatched peaks with SNR≤5 are not considered. The peak area is obtained as the abundance of the unmatched peak.

[0024] Furthermore, the construction of the storage year identification system in step (5) includes: merging the fitted abundance of known and unknown metabolites obtained in steps (3) and (4) to form the abundance matrix of white tea samples; and using the abundance matrix of each variety of white tea to construct a multivariate regression model to form a white tea storage year identification system.

[0025] This invention also provides a method for identifying the storage age of white tea based on spectral peak fitting, comprising the following steps:

[0026] Step 1: Following the aforementioned method, establish a white tea storage year identification system based on spectral peak fitting;

[0027] Step Two: Perform nuclear magnetic resonance (NMR) analysis on the white tea sample to be identified to obtain the sample's... 1 The H NMR spectrum was then fitted to the sample to be identified using the method described above. 1 Abundance of known and unknown metabolites in H NMR spectra;

[0028] Step 3: Use the data obtained in Step 2 as input to the system obtained in Step 1 to calculate the year of the white tea sample to be identified.

[0029] Compared with the prior art, this technical solution has the following advantages:

[0030] This invention proposes a strategy based on fitting the overall spectrum of known metabolites and fitting single peaks of unknown metabolites. For different varieties of white tea, regression models are established between the fitted data and the storage year, forming a multi-variety white tea storage year identification system. For known metabolites, the method first fits the standard spectrum of the metabolite to obtain information such as the chemical shift, peak height, linewidth, and area of ​​each peak. Then, while maintaining the integral area ratio of each peak of the metabolite, the standard spectrum is matched, aligned, and fitted with the sample spectrum. For possible peak chemical shift drift and linewidth variations, fine-tuning of the chemical shift and linewidth is allowed. For unassigned peaks, single-peak fitting is performed using the Lorentz line shape. Finally, the spectral data of the standard and unmatched peaks are merged, and multiple storage year regression models are established for different varieties of white tea samples. This method utilizes more complete sample composition information and effectively eliminates and reduces the influence of peak drift and overlap, constructing a system for identifying the storage year of white tea. It is of great significance for correctly assigning metabolite peaks, accurate quantitative analysis, and prediction of the storage year of white tea.

[0031] This invention proposes a strategy based on fitting the overall spectrum of known metabolites and fitting the single peak of unknown metabolites to comprehensively and accurately extract metabolite composition information from the NMR spectrum of samples. Based on the fitted sample composition information, the storage years of different varieties of white tea are modeled and analyzed one by one, forming a multi-variety white tea age identification system. Compared with existing methods, this strategy effectively solves the problems of peak overlap and peak drift, and can obtain more comprehensive sample composition information, providing a more accurate storage year identification system for different varieties of white tea.

[0032] 1. The nuclear magnetic resonance hydrogen spectrum feature extraction method of the present invention can make more comprehensive use of complete sample composition information while minimizing the introduction of baseline and noise.

[0033] 2. The method for extracting features from proton nuclear magnetic resonance (NMR) spectra of the present invention uses the standard spectrum of metabolites as a reference spectrum for peak assignment, matching, and quantification, while allowing for fine-tuning of peak chemical shifts and linewidths, which can effectively reduce the influence of random peak drift and overlap.

[0034] 3. The method for extracting proton nuclear magnetic resonance (NMR) spectral features of the present invention, based on more accurate peak assignment and more comprehensive sample composition information, obtains more differential features and can construct a more reliable analytical model.

[0035] 4. The white tea storage year identification system of the present invention has better applicability because it builds a storage year regression model for different varieties of white tea. Attached Figure Description

[0036] Figure 1 The PLS scores of white tea samples based on the peak fitting strategy are shown in the image. (A) Peony, R2Y=98.5%, Q2=0.96; (B) Shoumei, R2Y=98.6%, Q2=0.95; (C) Silver Needle, R2Y=97.2%, Q2=0.94. Dots: 1-year samples, triangles: 3-year samples, rhombuses: 7-year samples.

[0037] Figure 2 These are metabolites or unmatched peaks that are significantly related to the year of origin in three varieties of white tea: Peony (A), Shoumei (B), and Silver Needle (C).

[0038] Figure 3 The results of white tea storage age identification based on peak fitting strategy. (A) Peony; (B) Shoumei; (C) Silver Needle. (Dark black: white tea sample stored for 1 year; medium black: white tea sample stored for 3 years; light black: white tea sample stored for 7 years). Detailed Implementation

[0039] This invention discloses a system for identifying the storage age of white tea based on spectral peak fitting. Its specific construction method includes the following steps:

[0040] (1) Preprocessing of one-dimensional nuclear magnetic resonance hydrogen spectrum data: one-dimensional nuclear magnetic resonance hydrogen spectrum of white tea sample ( 1 The free induction decay (FID) signals from H NMR were preprocessed using fast Fourier transform, phase correction, baseline correction, chemical shift calibration, and peak alignment to obtain samples of white tea from different storage years and varieties. 1 H NMR spectral data.

[0041] (2) Metabolite standards 1 1H NMR spectral information acquisition: Acquire (or download) white tea metabolite standards. 11H NMR spectroscopy is used to obtain information such as the splitting mode, chemical shift, linewidth, and peak height of metabolite spectra, and to statistically analyze the area ratio between the peaks of each metabolite.

[0042] (3) Sample 1 Abundance fitting of known metabolites in H NMR spectra: matching the peaks of known metabolite standards with those of white tea samples. 1 By comparing, matching, and fitting the H NMR spectra, the white tea samples were calculated. 1 Abundance of known metabolites in H NMR spectra.

[0043] (4) Sample 1 Abundance fitting of mismatched peaks in H NMR spectra: for white tea samples 1 The unassigned peaks in the 1H NMR spectrum were fitted one by one to calculate the abundance of unmatched peaks in the white tea sample.

[0044] (5) Construction of storage year identification system: Combine the fitting abundance of known metabolites and unmatched peaks, construct the storage year regression model of each variety of white tea, and complete the construction of white tea storage year identification system based on spectral peak fitting.

[0045] The one-dimensional nuclear magnetic resonance hydrogen spectrum data preprocessing in step (1) includes: padding the collected white tea sample nuclear magnetic resonance FID data with zeros to integer powers of 2, and then performing a fast Fourier transform to obtain the white tea sample's... 1 The ¹H NMR spectrum was then processed, including phase correction, baseline correction, chemical shift calibration, peak alignment, and sample normalization, to obtain the white tea sample. 1 H NMR spectral data.

[0046] The metabolite standards in step (2) 1 The acquisition of H NMR spectral information includes: collecting standards for known metabolites that can be attributed to white tea samples. 1 1H NMR spectra; for metabolites for which standards could not be obtained, their chromatograms can be downloaded from public databases such as HMDB, NP-MRD, and FooDB. 1 1H NMR experimental spectra; by fitting the peaks, information on the composition of the sample is obtained, including the splitting mode, chemical shift, linewidth and peak height of each peak, and the integral area ratio of each peak is calculated.

[0047] The sample in step (3) 1 Fitting the abundance of known metabolites in ¹H NMR spectra includes: fine-tuning the chemical shifts and linewidths of each peak in the metabolite standard while keeping the peak areas constant, so that they match those of the white tea sample. 1 Alignment and matching of corresponding peaks in the H NMR spectrum; and fitting of white tea samples by amplifying or reducing the intensity of the overall signal of the metabolite standard spectrum.1 The corresponding matching peaks in the H NMR spectrum are used as the abundance of the corresponding metabolites in the white tea sample.

[0048] The sample in step (4) 1 Abundance fitting of mismatched peaks in H NMR spectra includes: for white tea samples 1 Peaks in the H NMR spectrum that fail to match the standard spectrum of the metabolite are considered as unknown metabolites; Lorentz line peaks are sampled and fitted one by one to the unmatched peaks, and the fitted area is used as the abundance of the corresponding unknown metabolite.

[0049] The construction of the storage year identification system in step (5) includes: merging the fitted abundance of known and unknown metabolites obtained in steps (3) and (4) to form the abundance matrix of white tea samples; and using the abundance matrix of each variety of white tea to construct a multivariate regression model to form a white tea storage year identification system.

[0050] In one embodiment of the present invention, the white tea storage age identification system based on spectral peak fitting described in the present invention is constructed by the following steps: preprocessing the one-dimensional hydrogen spectrum data of the white tea sample, including fast Fourier transform, phase correction, baseline correction, chemical shift calibration, and peak alignment; and performing metabolite standardization. 1 ¹H NMR spectral information acquisition: obtaining information such as splitting mode, chemical shift, linewidth, and peak height of the standard; statistical analysis of the area ratio between each peak of the metabolite; abundance fitting of known metabolites in the sample: matching the peaks of the standard with those of the white tea sample. 1 The ¹H NMR spectra were compared and matched, and the peak values ​​of metabolites were optimized and fitted as a whole to calculate the abundance of each metabolite in the white tea sample. The abundance of unmatched peaks in the sample was fitted one by one to obtain the abundance of each unmatched peak. An age identification system was constructed, and the fitted abundance of known metabolites and unmatched peaks was combined. Regression models for the storage years of white tea were constructed according to different varieties of white tea.

[0051] Example 1

[0052] 1. White tea sample processing

[0053] The white tea samples used in this invention were collected by the Fujian Academy of Agricultural Sciences in Fuding City, Fujian Province. They included three different varieties: Silver Needle, Shoumei, and Peony, with 36 samples of each variety, totaling 108 samples. Each variety also included three different storage years: 1 year, 3 years, and 7 years, with 12 samples for each storage year.

[0054] The finished white tea was ground into a fine powder using a grinder and passed through a 100-mesh sieve. For NMR sample preparation, 1 g of tea powder was mixed with 5 mL of methanol / water solution (70 / 30, v / v) and stirred at 70 °C for 30 min, retaining non-volatile compounds. After cooling to room temperature, the mixture was centrifuged at 4000 rpm for 5 min, and 2.5 mL of the supernatant was transferred to a 10 mL centrifuge tube. Residual methanol was removed using a nitrogen blower. Finally, the mixture was freeze-dried in a vacuum dryer at -80 °C to obtain dry powder of tea metabolites. The dry powder of metabolites was dissolved in 700 μL of DMSO-d6 (dimethyl sulfoxide, containing 0.03% TMS tetramethylsilane) solution, vortexed for 30 s at 1000 rpm at 4 °C, and centrifuged for 10 min. 600 μL of the supernatant was transferred to an NMR tube and stored at 4 °C until NMR analysis.

[0055] 2. Nuclear Magnetic Resonance Experiment

[0056] Equipped with a TXI suitable for reverse detection 1 H / 13 C / 15 Samples were acquired on an 850 MHz Bruker Avance III nuclear magnetic resonance spectrometer with an N 5 mm cryogenic probe. 1 H NMR spectra were obtained using the NOESY (nuclear Overhauser effect spectroscopy) pulse sequence. After calibrating the 90-degree pulse width in the sequence, the experimental temperature was set to 298 K, the relaxation delay τ = 1.0 s, the acquisition time was 1.8 seconds, the spectral width was 17000 Hz, and sampling was performed with 16 accumulations and 2 blank scans. The sampling time for each spectrum was approximately 3 minutes.

[0057] 3. Preprocessing of proton NMR data

[0058] The acquired nuclear magnetic resonance (NMR) FID data of the samples were zero-padded to 32 K data points, smoothed using a Gaussian window function with a widening factor of 0.3 Hz, and then subjected to a fast Fourier transform to obtain the white tea sample data. 1H NMR spectra; based on this, manual phase and baseline corrections were performed on the spectra [Long Yichi, Deng Lingli, Xu Jingjing, et al. An automatic baseline correction algorithm for nuclear magnetic resonance spectra based on elliptical motion [C]. Proceedings of the 21st National Conference on Spectroscopy. 2021:404-405]. Chemical shift referenced internal standard TMS (δ0.00). All the above steps were completed using MestReNova software (Version 14, Mestrelab Research, Santiago de Compostela, Spain); the icoshift method [Savorani F, Tomasi G, Engelsen S B. icoshift: A versatile tool for the rapid alignment of 1D NMR spectra [J]. Journal of magnetic resonance, 2010, 202(2): 190-202] was used to align the spectral peaks between different samples; the PQN method [Dieterle F, Ross A, Schlotterbeck G, et al. Probabilistic quotient normalization as robust method to account fordilution of complex biological mixtures. Application in 1 H NMR metabonomics[J]. Analytical chemistry, 2006, 78(13): 4281-4290] Sample normalization was performed to reduce the influence of sample concentration differences and experimental condition differences on the sample signal intensity, and finally the white tea sample was obtained. 1 H NMR spectral data.

[0059] 4. Metabolite standards 1 H NMR spectral information acquisition

[0060] Eleven compounds, including catechin (C), epicatechin gallate (GCG), gallocatechin (GC), epigallocatechin (EGC), epicatechin (EC), epigallocatechin gallate (EGCG), epicatechin gallate (ECG), caffeine (CAF), gallic acid (GA), and theanine (The), were purchased as standards from Nanjing Daosi Biotechnology Co., Ltd. Based on the NMR experimental parameters of white tea samples, NMR samples of these eleven compounds were collected using an 850 MHz Bruker Avance III nuclear magnetic resonance spectrometer. 1 1H NMR spectra. In addition, 10 metabolites, including leucine, valine, isoleucine, alanine, fatty acids, chlorogenic acid, acetic acid, succinic acid, glucose, and sucrose, were downloaded from public databases such as HMDB (https: / / hmdb.ca / ), NP-MRD (https: / / np-mrd.org / ), and FooDB (https: / / foodb.ca / ). 1 1H NMR experimental spectra. The 21 metabolite standards were analyzed using SpectFitting software [Xu Xiangnan, Dong Jiyang. Quantitative Analysis Software for 1H NMR Spectrum (NMRSpectFitting Ver.1.0). Computer Software Copyright Registration Certificate (Software Copyright Registration No. 1743627, Registration No.: 2017SR158343) 2017.05.04], developed by our research group. 1 Peak fitting of H NMR spectra yielded 21 metabolites. 1 Information such as peak splitting mode, chemical shift, linewidth, peak height, and peak area ratio of H NMR spectrum.

[0061] 5. Abundance fitting of known metabolites in the sample

[0062] While maintaining the area ratio between each metabolite peak, the chemical shift and linewidth of each metabolite peak were finely adjusted to match those of the white tea sample. 1 The corresponding peaks in the H NMR spectrum were aligned with each other; then, the overall signal intensity of the metabolite spectrum was amplified or reduced to match that of the white tea sample. 1 When the corresponding peaks in the 1H NMR spectrum reach optimal matching, the total area of ​​the metabolite spectrum is taken as the abundance of that metabolite in the white tea sample. This process is performed using the SpectFitting software independently developed by our research group.

[0063] 6. Abundance fitting of mismatched peaks in samples

[0064] For the peaks in the hydrogen spectrum of white tea samples that could not be assigned in steps 2 and 3, they were all considered as unknown single-peak metabolites; the signal-to-noise ratio (SNR) of each unknown metabolite peak was calculated according to the following formula [Hyberts SG, Robson SA, Wagner G. Exploring signal-to-noise ratio and sensitivity in non-uniformly sampled multi-dimensional NMR spectra [J]. Journal of biomolecular NMR, 2013, 55:167-178.].

[0065] SNR=SI / SD

[0066] Where SI is the signal intensity of the unmatched peak, and SD is the standard deviation of the noise region in the same sample spectrum.

[0067] Using the Lorentz linear function, i.e. (Where A and B are constants), unmatched peaks with SNR>5 are fitted (unmatched peaks with SNR≤5 are not considered), and their peak areas are used as the abundance of unmatched peaks. This process is completed on the SpectFitting software independently developed by our research group.

[0068] 7. Construction of a Year Determination System

[0069] The fitted abundance data of metabolites and mismatched peaks were combined. Partial least square (PLS) regression models were constructed for each white tea variety based on its storage year, as follows: Figure 1 As shown. Meanwhile, potential biomarkers were screened based on the variable projection importance analysis value (VIP>1, variable importance for the projection) [Akarachantachote N, Chadcham S, Saithanu K. Cutoff threshold of variable importance in projection for variable selection [J]. Int J Pure Appl Math,2014, 94(3): 307-322.] and the absolute value of the correlation coefficient (|r|>0.6) of the PLS regression model. Figure 2 As shown.

[0070] Using cross-validation, a 70% validation test was conducted on the white tea storage age identification model. The predicted age results for 36 samples from each of the three white tea varieties are as follows: Figure 3 As shown in the figure, the average absolute errors in identifying the storage years of the three types of white tea (1, 3, and 7 years) were 0.200, 0.277, and 0.267 (years), respectively. The relative errors for identifying the storage years of the seven-year-old white tea were only 3.39% (Peony), 3.87% (Shoumei), and 4.11% (Silver Needle). This indicates that this method has high reliability in identifying high-quality white tea with longer storage periods.

[0071] The above description is merely a preferred embodiment of the present invention, and therefore should not be construed as limiting the scope of the present invention. All equivalent changes and modifications made in accordance with the scope of the patent and the contents of the specification should still fall within the scope of the present invention.

Claims

1. A system for identifying the storage age of white tea based on spectral peak fitting, characterized in that, The construction method of this system includes the following steps: (1) Preprocessing of one-dimensional nuclear magnetic resonance hydrogen spectrum data: Fast Fourier transform, phase correction, baseline correction, chemical shift calibration and peak alignment were performed on the free induction decay signal of the one-dimensional nuclear magnetic resonance hydrogen spectrum of white tea samples to obtain one-dimensional nuclear magnetic resonance hydrogen spectrum data of white tea samples with different storage years and different varieties. (2) Acquisition of one-dimensional nuclear magnetic resonance hydrogen spectrum information of metabolite standards: Acquiring or downloading known metabolite standards of white tea. 1 H NMR spectroscopy is used to obtain information on the splitting mode of metabolite spectra, chemical shifts, linewidths, and peak heights of each peak, and to statistically analyze the area ratios between the peaks of each metabolite. (3) Abundance fitting of known metabolites in one-dimensional nuclear magnetic resonance hydrogen spectrum of sample: The peaks of known metabolite standards are compared with the one-dimensional nuclear magnetic resonance hydrogen spectrum of white tea sample. Under the condition of keeping the integral area ratio between the peaks of the same metabolite unchanged, the chemical shift and linewidth of each peak of the known metabolite standards are finely adjusted to align and match with the corresponding peaks in the one-dimensional nuclear magnetic resonance hydrogen spectrum of white tea sample. The signal intensity of the peaks of the known metabolite standards is amplified or reduced as a whole to achieve optimal matching with the corresponding peaks in the one-dimensional nuclear magnetic resonance hydrogen spectrum of white tea sample. The total area of ​​the peaks of the known metabolite standards when optimal matching is achieved is taken as the abundance of the corresponding known metabolite in white tea sample. (4) Abundance fitting of mismatched peaks in the one-dimensional proton NMR spectrum of the sample: The peaks that failed to match the spectral peaks of the known metabolite standards in step (3) are considered as unmatched peaks, and the signal-to-noise ratio of each unmatched peak is calculated according to the following formula: ； Where SI is the signal intensity of the unmatched peak, and SD is the standard deviation of the noise region in the same sample spectrum; Using Lorentz linear function ； For unmatched peaks with an SNR greater than 5, peak-by-peak fitting was performed, where A and B are constants. The area of ​​the fitted peak was used as the abundance of the corresponding unmatched peak. Unmatched peaks with an SNR less than or equal to 5 were not included in the subsequent analysis. (5) Construction of storage year identification system: Combine the fitting abundance of known metabolites obtained in step (3) and the fitting abundance of unmatched peaks obtained in step (4) to form the abundance matrix of white tea samples, and use the abundance matrices of different varieties of white tea to construct multivariate regression models of storage year, forming a white tea storage year identification system based on spectral peak fitting.

2. The white tea storage age identification system according to claim 1, characterized in that: In step (1), the white tea sample was processed as follows: the finished white tea was ground into fine powder using a grinder and passed through a 100-mesh sieve; in the sample preparation for the NMR experiment, 1g of tea powder was mixed with 5 mL of methanol / water solution (70 / 30, v / v), stirred at 70℃ for 30 min, and non-volatile compounds were retained; after cooling to room temperature, it was centrifuged at 4000 rpm for 5 min, and 2.5 mL of supernatant was transferred to a 10 mL centrifuge tube and the residual methanol was removed in a nitrogen blower; finally, it was freeze-dried in a vacuum dryer at -80 ℃ to obtain dry powder of tea metabolites; the dry powder of metabolites was dissolved in 700 μL of DMSO-d6 solution, vortexed for 30 s, rotated at 1000 rpm, and centrifuged at 4℃ for 10 min; 600 μL of supernatant was transferred to a nuclear magnetic resonance tube and stored in a 4°C refrigerator until the NMR experiment.

3. The white tea storage year identification system according to claim 1, characterized in that: In step (1), the nuclear magnetic resonance conditions are: equipped with a TXI suitable for reverse detection. 1 H / 13 C / 15 Samples were acquired on an 850 MHz Bruker Avance III nuclear magnetic resonance spectrometer with an N 5 mm cryogenic probe. 1 H NMR spectra were obtained using the NOESY pulse sequence. After calibrating the 90-degree pulse width in the sequence, the experimental temperature was set to 298 K, the relaxation delay τ = 1.0 s, the acquisition time was 1.8 seconds, the spectral width was 17000 Hz, and sampling was performed with 16 accumulations and 2 empty scans. The sampling time for each spectrum was 3 minutes.

4. The white tea storage year identification system according to claim 1, characterized in that: The one-dimensional nuclear magnetic resonance (NMR) hydrogen spectrum data preprocessing in step (1) includes: padding the acquired white tea sample NMR FID data with zeros to integer powers of 2, and then performing a fast Fourier transform to obtain the white tea sample's... 1 The ¹H NMR spectrum was then processed, including phase correction, baseline correction, chemical shift calibration, peak alignment, and sample normalization, to obtain the white tea sample. 1 H NMR spectral data.

5. The white tea storage year identification system according to claim 1, characterized in that: The metabolite standards in step (2) 1 The acquisition of H NMR spectral information includes: collecting standards for known metabolites that can be attributed to white tea samples. 1 1H NMR spectra; for metabolites for which standards could not be obtained, their chromatograms can be downloaded from public databases such as HMDB, NP-MRD, and FooDB. 1 1H NMR experimental spectra; by fitting the peaks, information on the composition of the sample is obtained, including the splitting mode, chemical shift, linewidth and peak height of each peak, and the integral area ratio of each peak is calculated.

6. The white tea storage year identification system according to claim 1, characterized in that: The sample in step (3) 1 Fitting the abundance of known metabolites in HNMR spectra includes: fine-tuning the chemical shifts and linewidths of each peak in the metabolite standard while keeping the peak areas constant, so that they match those of the white tea sample. 1 Alignment and matching of corresponding peaks in the H NMR spectrum; and fitting of white tea samples by amplifying or reducing the intensity of the overall signal of the metabolite standard spectrum. 1 The corresponding matching peaks in the H NMR spectrum are used as the abundance of the corresponding metabolites in the white tea sample.

7. The white tea storage year identification system according to claim 1, characterized in that: The sample in step (4) 1 Abundance fitting of mismatched peaks in HNMR spectra includes: for white tea samples 1 Peaks in the H NMR spectrum that do not match the peaks of the metabolite standard are considered unknown metabolites; Lorentz line peaks are sampled and fitted one by one to the unmatched peaks, and the fitted area is used as the abundance of the corresponding unknown metabolite.

8. The white tea storage year identification system according to claim 7, characterized in that: In step (4), for the peaks of the hydrogen spectrum of the white tea samples that could not be assigned, the signal-to-noise ratio of each unknown peak is calculated according to the following formula. SNR = SI / SD; Where SI is the signal intensity of the unmatched peak, and SD is the standard deviation of the noise region in the same sample spectrum; Using the Lorentz linear function, i.e. Where A and B are constants, the unmatched peaks with SNR>5 are fitted, and the unmatched peaks with SNR≤5 are not considered. The peak area is obtained as the abundance of the unmatched peak.

9. The white tea storage age identification system according to claim 1, characterized in that: The construction of the storage year identification system in step (5) includes: merging the fitted abundance of known and unknown metabolites obtained in steps (3) and (4) to form the abundance matrix of white tea samples; using the abundance matrix of each variety of white tea to construct a multivariate regression model to form the white tea storage year identification system.

10. A method for identifying the storage age of white tea based on spectral peak fitting, comprising the following steps: Step 1: Establish a white tea storage year identification system based on spectral peak fitting according to the method described in any one of claims 1 to 9; Step 2: Perform nuclear magnetic resonance (NMR) detection on the white tea sample to be identified to obtain a one-dimensional NMR hydrogen spectrum of the white tea sample to be identified, and preprocess the one-dimensional NMR hydrogen spectrum using the same preprocessing method as in the construction process of the white tea storage year identification system described in Step 1. Using the known metabolite standard spectra in the white tea storage age identification system described in step one, the known metabolite standard spectra are compared with the one-dimensional nuclear magnetic resonance hydrogen spectrum of the white tea sample to be identified. While keeping the integral area ratio between the peaks of the same metabolite constant, the chemical shift and linewidth of each peak in the known metabolite standard spectra are finely adjusted to align and match with the corresponding peaks in the one-dimensional nuclear magnetic resonance hydrogen spectrum of the white tea sample to be identified. The signal intensity of the known metabolite standard spectra is then amplified or reduced to achieve optimal matching with the corresponding peaks in the one-dimensional nuclear magnetic resonance hydrogen spectrum of the white tea sample to be identified. The total area of ​​the known metabolite standard spectra at which optimal matching is achieved is taken as the abundance of the corresponding known metabolite in the white tea sample to be identified. Peaks that fail to match the standard spectra of known metabolites during the overall fitting process are considered unmatched peaks. The signal-to-noise ratio of each unmatched peak is calculated using the following formula: ； in, SI is the signal intensity of the unmatched peak, and SD is the standard deviation of the noise region in the same sample spectrum. Using Lorentz linear function ； For unmatched peaks with an SNR greater than 5, peak-by-peak fitting was performed, where A and B are constants. The area of ​​the fitted peak was used as the abundance of the corresponding unmatched peak. Unmatched peaks with an SNR less than or equal to 5 were not included in the subsequent analysis. The known metabolite fit abundance and the unmatched peak fit abundance of the white tea sample to be identified are combined to obtain the fit abundance data of the white tea sample to be identified. Step 3: Based on the variety of the white tea sample to be identified, input the fitted abundance data obtained in Step 2 into the corresponding variety white tea storage year regression model established in Step 1 to calculate the storage year of the white tea sample to be identified.

Images