Screening method of mango pulp processing mode marker and identification method of mango pulp processing mode

Abstract

The invention discloses a screening method of mango pulp processing mode markers and an identification method of mango pulp processing modes, and relates to the technical field of analytical chemistry. According to the method, non-targeted metabonomics is combined with quasi-targeted metabonomics and targeted quantitative analysis, so that the characteristic markers capable of accurately distinguishing mango pulp in different processing modes are systematically screened and verified for the first time. The feature marker screened by the method can be used for constructing a discrimination model of the mango pulp processing mode, high-precision and high-reliability processing mode discrimination is realized based on the established multi-model fusion discrimination method, the discrimination efficiency and accuracy of the mango pulp processing mode are greatly improved, and the method is suitable for popularization and application. And a new technical means and theoretical basis are provided for product quality control and market supervision.

Description

Technical Field

[0001] The present invention relates to the technical field of analytical chemistry, and in particular to a method for screening markers of mango pulp processing methods and a method for identifying mango pulp processing methods. Background Art

[0002] Mango (Mangifera indica L.), a member of the Anacardiaceae family, is native to northern India and the Malay Peninsula, primarily distributed in tropical and subtropical regions. Known for its delicate texture, rich nutrition, and unique flavor, it is known as the "King of Tropical Fruits" (Evans et al. 2017; Ntsoane et al. 2019b) and is beloved by consumers worldwide (Alphonce et al. 2015). It accounts for over 50% of the global tropical fruit market share (Food Outlook 2018). With growing demand for high-end fruits, China's mango industry is gradually moving towards scale, standardization, and high-value-added processing, resulting in a steady increase in both production and value. Currently, China has become the world's second-largest mango producer, with a total output of 4.899 million tons and a total output value of 28.82 billion yuan in 2023. Guangxi, Hainan, Yunnan, and Guangdong are the main mango producing areas in China (National Bureau of Statistics). Despite the continued expansion of the domestic mango industry, China remains the world's third-largest mango importer. In 2024, my country's cumulative mango imports reached 17,313.37 tons, a year-on-year increase of 16% (according to data from the General Administration of Customs). According to the Food and Agriculture Organization of the United Nations (FAO), global mango planting area and production continue to grow, with total production in 2022 estimated at approximately 56 million tons. Asia contributes approximately 75% of global production, with India as the largest producer and China as one of the main consumer markets.

[0003] In recent years, mangoes have become not only sold fresh, but also processed products have gained a significant foothold in the food industry. Key processed products include frozen mango pulp, mango juice, dried mango, and mango preserves. Frozen mango pulp, due to its convenience, sustainability, and clean label characteristics, has become increasingly popular in the new tea, bakery, dairy, and functional food industries. In 2023, the global frozen fruit pulp market was projected to be worth approximately US$64.06 billion, with mango pulp being a key product. With the gradual development of domestic mango pulp production in my country, exports of mango juice and pulp products are gradually increasing, with the country's trade deficit shifting to a surplus in 2024. Frozen mango pulp is widely used in the production of fruit teas, beverages, baked goods, and other products. With rising consumer demand for healthy foods and the rapid development of the catering industry, the frozen mango pulp market continues to expand. In the future, as the trends toward clean label, high-quality, and functional foods continue to grow, innovative base products such as frozen fruit pulp and ready-to-drink fruit and vegetable beverages will drive industry upgrades.

[0004] Microorganisms such as bacteria, yeast, and mold are the primary cause of mango pulp deterioration. Therefore, pasteurization is necessary during processing to extend the shelf life of the pulp. There are three main processing methods for frozen mango pulp: NS (unpasteurized), HPP (ultra-high pressure pasteurized), and HTST (ultra-high temperature transient pasteurization). Both thermal and ultra-high pressure non-thermal pasteurization techniques affect the content of phytochemicals in the product. Because some phenolic compounds, such as hydroxycinnamic acid and its derivatives, are temperature-sensitive and easily degrade at high temperatures, the total phenolic content of mango pulp decreases by 10% after thermal pasteurization (Guan et al. 2016). For example, the content of phenolic compounds in Maria mango pulp decreases after pasteurization, with the chlorogenic acid content dropping from 23.74 ± 0.01 mg / 100 g FW to 14.34 ± 0.17 mg / 100 g FW (Silva et al. 2022). Compared with traditional pasteurization, non-thermal sterilization technologies such as ultra-high pressure have less impact on the phytochemical composition of fruits and vegetables. It is generally believed that small molecular compounds such as amino acids and terpenes are basically unaffected by pressure (Roobab et al. 2021). At the same time, the shear, turbulence and shock waves generated during ultra-high pressure processing release phenolic substances such as ferulic acid that are bound to the food matrix. Therefore, the quantifiable content of total phenols in mango pulp treated with ultra-high pressure increased by up to 21.4% (Kubo et al. 2021; Guan et al. 2016; Kaushik et al. 2015). Carotenoids in mango pulp are easily degraded under high pressure, among which lutein is more easily degraded under high pressure. The total amount of lutein in Keitt mango juice decreased by 17% after high-pressure treatment, from 324.54 to 269.37 μg / 100g FW; carotene is relatively stable under high pressure, and β-carotene only lost 27.61 μg / 100g FW after ultra-high pressure processing, with a loss rate of 6% (Hu et al. 2023).

[0005] In summary, while NS mango pulp can maximize the preservation of mango's original flavor and nutrients, it poses microbial safety risks. HPP technology uses ultra-high pressure processing to achieve sterilization, effectively killing microorganisms at lower temperatures, maximizing the preservation of mango's nutrients and flavor while extending its shelf life. HTST technology uses high-temperature, short-term sterilization, achieving a thorough sterilization effect and extending its shelf life, making it suitable for large-scale production and long-distance transportation. However, high-temperature processing may lead to the loss of some heat-sensitive nutrients and affect flavor. Currently, product quality is mainly evaluated based on conventional physical and chemical indicators such as total phenolics and total sugars. However, the lack of precise markers and corresponding detection methods makes it difficult to accurately distinguish products processed by different methods.

[0006] In view of this, the present invention is proposed. Summary of the Invention

[0007] The object of the present invention is to provide a method for screening markers of mango pulp processing methods and a method for identifying mango pulp processing methods so as to accurately discriminate the processing methods of mango pulp.

[0008] The present invention is achieved in that:

[0009] In a first aspect, the present invention provides a method for screening markers of mango pulp processing methods, comprising at least one of the following methods:

[0010] Screening of non-targeted differential markers, screening of pseudo-targeted differential markers, and screening and validation of targeted differential markers;

[0011] The screening of non-targeted differential markers includes the following steps:

[0012] S1: Non-targeted metabolomics analysis: The samples were subjected to non-targeted metabolomics analysis using high-resolution chromatography-mass spectrometry to identify the phytochemical components in mango pulp;

[0013] S2: Screening of non-targeted differential markers: Based on the non-targeted metabolomics data of mango pulp under any different processing methods, a first discriminant model of different mango pulp processing methods was established. Through the first discriminant model of different mango pulp processing methods, combined with statistics, the non-targeted differential markers of mango pulp under different processing methods were found; the screening rules for non-targeted differential markers were as follows: VIP value greater than 1, difference fold FC greater than 2 or less than 0.5, and p value less than 0.05.

[0014] The screening of targetable differential markers includes the following steps i or ii:

[0015] i: Based on the identification results of phytochemical components in mango pulp or the results of non-targeted differential marker screening obtained by non-targeted metabolomics analysis, quantitative analysis instruments are used to accurately quantitatively analyze the phytochemical components and collect pseudo-targeted metabolomics data;

[0016] ii: Based on the pseudo-metabolomics data of mango pulp under any different processing methods, a second discriminant model for different mango pulp processing methods was established. Through the second discriminant model for different mango pulp processing methods, combined with statistics, pseudo-targeted differential markers of mango pulp under different processing methods were identified; the screening rules for pseudo-targeted differential markers were as follows: VIP value greater than 1, differential fold greater than 2 or less than 0.5, and p-value less than 0.05;

[0017] The screening and validation of targeted differential markers include the following steps:

[0018] Based on the screened non-targeted differential markers and / or the intended targeted differential markers, marker screening and targeted verification are performed; existing target compound standards are screened from the non-targeted differential markers and / or the intended targeted differential markers as candidate targeted differential markers; the candidate targeted differential markers are qualitatively and quantitatively verified, and through the qualitative and quantitative verification results, targeted metabolome markers with obvious differences under different processing methods are screened.

[0019] In a second aspect, the present invention provides a method for identifying a mango pulp processing method, wherein at least one of the following mango pulp processing method markers is obtained according to the above-mentioned screening method for mango pulp processing method markers: a non-targeted differential marker, a pseudo-targeted differential marker, and a targeted differential marker;

[0020] Then, a third discriminant model was established and verified: Based on the quantitative results of mango pulp processing method markers, multiple classification models were used to independently predict the processing type of the samples. The prediction results of multiple classification models were integrated to construct a third discriminant model for mango pulp processing methods.

[0021] Then, discriminant analysis of mango pulp samples was performed based on the established third discriminant model.

[0022] In a third aspect, the present invention provides an electronic device comprising: a processor and a memory; the processor and the memory are connected, wherein the memory is used to store a computer program, and the processor is used to call the computer program to execute the above-mentioned method for identifying the mango pulp processing method.

[0023] In a fourth aspect, the present invention provides a computer-readable storage medium comprising instructions, which, when executed on a computer, enable the computer to execute the above-mentioned method for identifying a mango pulp processing method.

[0024] In a fifth aspect, the present invention provides the use of a reagent for detecting a mango pulp processing method marker in the identification of mango pulp processing methods, wherein the mango pulp processing method marker is selected from at least one of the following: isoleucine, valine, valine, arginine, glycyl-L-leucine, paeonol, (-)-catechin, azelaic acid, gulonic acid, epigallocatechin, ferulic acid and methyl gallate.

[0025] The present invention has the following beneficial effects:

[0026] This study systematically screened and validated, for the first time, characteristic markers that can accurately distinguish mango pulp processed by different methods by combining non-targeted metabolomics with quasi-targeted metabolomics and targeted quantitative analysis. The characteristic markers screened by this method can be used to construct a discriminant model for mango pulp processing methods. The established multi-model fusion discrimination method achieves high-precision and high-reliability processing method discrimination, greatly improving the efficiency and accuracy of identifying mango pulp processing methods and providing new technical means and theoretical basis for product quality control and market supervision. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0028] Figure 1 The figure shows the classification and quantity results of qualitative phytochemical components in mango pulp;

[0029] Figure 2 The PCA scores of phytochemical components in mango pulp under different processing methods are shown in Figure 1. a: PCA score plot of three mango pulps under HILIC chromatography mode; b: PCA score plot of three mango pulps under RPLC chromatography mode; c: PCA score plot of three mango pulps extracted with different solvents under HILIC chromatography analysis; d: PCA score plot of three mango pulps extracted with different solvents under RPLC chromatography analysis.

[0030] Figure 3 Cluster heat map of components of frozen mango pulp processed by different methods;

[0031] Figure 4 OPLS-DA model for annotating components of three mango pulps (a, c, e), and 200 permutation test plots (b, d, f);

[0032] Figure 5 Comparison of the fold difference (FC) values ​​between non-targeted and targeted compounds; a, glutamic acid; b, ferulic acid;

[0033] Figure 6 Multi-model ensemble learning model framework (a) and decision boundary diagram (b);

[0034] Figure 7 The prediction results of the integrated model training set (a), the prediction results of the integrated model test set (b), the confusion matrix of the integrated model test set (c) and the confusion matrix of the integrated model training set (d);

[0035] Figure 8 This is the decision boundary diagram for discriminating the processing methods of commercially available samples based on multi-model integration. DETAILED DESCRIPTION

[0036] To make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely below. Where specific conditions are not specified in the embodiments, conventional conditions or conditions recommended by the manufacturer are used. Where the manufacturer of the reagents or instruments is not specified, all are conventional products that can be purchased commercially.

[0037] This study systematically screened and validated for the first time the characteristic markers that can accurately distinguish mango pulp processed by different methods by combining non-targeted metabolomics with quasi-targeted metabolomics and targeted quantitative analysis. The details are as follows:

[0038] In a first aspect, the present invention provides a method for screening markers of mango pulp processing methods, which includes at least one of the following methods: screening of non-targeted differential markers, screening of pseudo-targeted differential markers, and screening and verification of targeted differential markers;

[0039] The screening of the non-targeted differential markers comprises the following steps:

[0040] S1: Non-targeted metabolomics analysis: The samples were subjected to non-targeted metabolomics analysis using high-resolution chromatography-mass spectrometry to identify the phytochemical components in mango pulp;

[0041] S2: Screening of non-targeted differential markers: Based on the non-targeted metabolomics data of mango pulp under any different processing methods, a first discriminant model of different mango pulp processing methods was established. Through the first discriminant model of different mango pulp processing methods, combined with statistics, the non-targeted differential markers of mango pulp under different processing methods were found; the screening rules for non-targeted differential markers were as follows: VIP value greater than 1, difference fold FC greater than 2 or less than 0.5, and p value less than 0.05.

[0042] The screening of the differential markers to be targeted includes the following steps i or ii:

[0043] i: Based on the identification results of phytochemical components in mango pulp or the non-targeted differential marker screening results obtained by the non-targeted metabolomics analysis, a quantitative analysis instrument is used to perform precise quantitative analysis of the phytochemical components to obtain pseudo-targeted metabolomics data;

[0044] ii: Based on the pseudo-metabolomics data of mango pulp under any different processing methods, a second discriminant model for different mango pulp processing methods was established. Through the second discriminant model for different mango pulp processing methods, combined with statistics, pseudo-targeted differential markers of mango pulp under different processing methods were identified; the screening rules for pseudo-targeted differential markers were as follows: VIP value greater than 1, differential fold greater than 2 or less than 0.5, and p-value less than 0.05;

[0045] The screening and verification of the targeted differential markers includes the following steps:

[0046] Based on the screened non-targeted differential markers and / or the intended targeted differential markers, marker screening and targeted verification are performed; existing target compound standards are screened out from the non-targeted differential markers and / or the intended targeted differential markers as candidate targeted differential markers; the candidate targeted differential markers are qualitatively and quantitatively verified, and targeted metabolome markers with obvious differences under different processing methods are screened out through the qualitative and quantitative verification results.

[0047] In non-targeted metabolomics analysis, the samples to be tested obtained by different extraction methods are chromatographed through one or more chromatographic modes and connected in tandem with a mass spectrometer for non-targeted metabolomics analysis;

[0048] Preferably, the chromatography mode is selected from HILIC, RPLC, ion exchange chromatography, hydrophobic interaction chromatography, size exclusion chromatography, supercritical fluid chromatography or a mixed chromatography mode.

[0049] Preferably, the chromatography mode is a combination of HILIC and RPLC chromatography modes.

[0050] In step S1 of the non-targeted metabolomics analysis, preferably, HILIC and / or RPLC columns are used in combination to improve component coverage and detection sensitivity, thereby identifying the phytochemical components in the mango pulp. Connecting the HILIC and RPLC chromatography products to a mass spectrometer for non-targeted metabolomics analysis helps to increase the number of phytochemical components in the mango pulp, providing a technical foundation for subsequent screening of differentially expressed biomarkers.

[0051] In one embodiment, the statistical analysis method in the screening of non-targeted differential markers and the screening of intended targeted differential markers is selected from: analysis of variance, fold difference analysis, Mann-Whitney U test, Wilcoxon signed rank test, Kruskal-Wallis H test, Friedman test, Bonferroni correction or FDR correction;

[0052] The high-resolution chromatography mass spectrometry instrument is selected from UHPLC-QTOF-MS, GC-TOF-MS or LC-Orbitrap-MS;

[0053] Preferably, the quantitative analysis instrument for the analysis of the targeted metabolites is UHPLC-QQQ or DART-QQQ;

[0054] Preferably, the first discriminant model or the second discriminant model is an OPLS-DA or PLS-DA model. When screening non-targeted differential markers, multivariate statistical analysis is performed using MetaboAnalyst 5.0 online platform, Excel, and SIMCA 14.1 software.

[0055] In one embodiment, before step S1, the sample to be tested is pretreated, and the pretreatment comprises the following steps: taking a mango pulp sample, extracting it with a 0.05%-0.20% acetic acid aqueous solution or a 30-70% (v / v) methanol aqueous solution containing 0.05%-0.20% acetic acid, taking the supernatant, and filtering it.

[0056] In step S1, the sample was subjected to non-targeted metabolomics analysis using UHPLC-QTOF-MS, coupled with RPLC and HILIC chromatography. After extraction with the same solvent using both RPLC and HILIC modes, the phytochemical components in mango pulp processed by different methods were clearly separated.

[0057] In one embodiment, those skilled in the art can adaptively select targeted, non-targeted and quasi-targeted differential markers for screening as needed, or can select a combination of targeted differential marker screening and non-targeted differential marker screening methods; or can select a combination of targeted differential marker screening, quasi-targeted differential marker screening, and non-targeted differential marker screening methods; or can select a combination of quasi-targeted differential marker screening and non-targeted differential marker screening methods.

[0058] The candidate targeted differential markers were qualitatively and quantitatively validated using at least one of the following methods: LC-MS, GC-MS, NMR, NIR, DART-MS, matrix-assisted laser desorption ionization (MALDI)-MS, and ion mobility mass spectrometry (IMS-MS).

[0059] In a preferred embodiment of the present invention, the non-targeted differential marker screening further comprises: performing unsupervised pattern recognition by principal component analysis (PCA) before establishing a prediction model for different mango pulp processing methods; in a more preferred embodiment, the model parameters (R 2 , Q 2 ) and at least 200 permutation tests were performed to verify the reliability of the model and determine the different components of mango pulp under different processing methods.

[0060] If the principal component analysis results show that mango pulp prepared by different processing methods can be clearly separated and different samples have differences, it indicates that the differential components may be used as non-targeted differential markers.

[0061] In a second aspect, the present invention provides a method for identifying a mango pulp processing method, wherein at least one of the following mango pulp processing method markers is obtained according to the above-mentioned screening method for mango pulp processing method markers: a non-targeted differential marker, a pseudo-targeted differential marker, and a targeted differential marker;

[0062] Then, a third discriminant model was established and verified: Based on the quantitative results of mango pulp processing method markers, multiple classification models were used to independently predict the processing type of the samples. The prediction results of multiple classification models were integrated to construct a third discriminant model for mango pulp processing methods.

[0063] Then, discriminant analysis of mango pulp samples was performed based on the established third discriminant model.

[0064] In a preferred embodiment of the present invention, the prediction results of integrating multiple classification models include: obtaining the final classification results of mango pulp processing methods by voting or weighting;

[0065] In a preferred embodiment of the present invention, the results of different classification models for processing methods are counted, the results of the processing methods are voted or weighted, and the processing method with the largest number of votes or the largest weight is selected as the final classification result of the mango pulp processing method.

[0066] In a preferred embodiment of the present invention, the classification model is selected from at least three of the following:

[0067] At least one of random forest (RF), support vector machine (SVM), k-nearest neighbor, partial least squares (PLS), least squares support vector machine (LSSVM), BP neural network (BPNN), convolutional neural network (CNN), extreme learning machine (ELM), particle swarm optimization BP neural network (PSO-BP), recurrent neural network, Bayesian classifier, K-nearest neighbor algorithm, K-means algorithm, linear regression, logistic regression, multivariate nonlinear regression fitting method, Adaboost algorithm, hidden Markov model, decision tree algorithm, clustering algorithm, generative adversarial network, stacked autoencoder, fully connected network, unsupervised pre-trained network, deep belief network, deep Boltzmann machine and neural tensor network.

[0068] For example, the classification model is selected from random forest (RF), support vector machine (SVM), partial least squares (PLS), least squares support vector machine (LSSVM), BP neural network (BPNN), convolutional neural network (CNN), extreme learning machine (ELM) and particle swarm optimization BP neural network (PSO-BP).

[0069] In a preferred embodiment of the present invention, by comparing the differences in detection accuracy, stability, sensitivity, linear range and coverage of non-targeted differential markers, quasi-targeted differential markers and targeted differential markers, mango pulp processing method markers are selected according to the comparison results to construct a third discriminant model for mango pulp processing methods.

[0070] In a preferred embodiment of the present invention, discriminant analysis of mango pulp samples based on the established third discriminant model includes the following steps: obtaining the content of mango pulp processing method markers in the samples to be tested, and discriminating the mango pulp processing method of the samples to be tested based on the third discriminant model.

[0071] The first discriminant model, the second discriminant model and the third discriminant model in the present invention are only used to distinguish different discriminant models, and there is no difference in the order of “first”, “second” and “third”.

[0072] In a third aspect, the present invention provides an electronic device comprising: a processor and a memory; the processor and the memory are connected, wherein the memory is used to store a computer program, and the processor is used to call the computer program to execute the above-mentioned method for identifying the mango pulp processing method.

[0073] Specifically, the electronic device may include a memory, a processor, a bus, and a communication interface, wherein the memory, processor, and communication interface are electrically connected to each other directly or indirectly to enable data transmission or interaction. For example, these components may be electrically connected to each other via one or more buses or signal lines. The processor may process information and / or data related to target identification to perform one or more functions described in this application.

[0074] The memory may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0075] A processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including a central processing unit (CPU) or a network processor (NP). It can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0076] In a fourth aspect, the present invention provides a computer-readable storage medium comprising instructions, which, when executed on a computer, enable the computer to execute the above-mentioned method for identifying a mango pulp processing method.

[0077] In a fifth aspect, the present invention provides the use of a reagent for detecting a mango pulp processing method marker in the identification of mango pulp processing methods, wherein the mango pulp processing method marker is selected from at least one of the following: isoleucine, valine, valine, arginine, glycyl-L-leucine, paeonol, (-)-catechin, azelaic acid, gulonic acid, epigallocatechin, ferulic acid and methyl gallate.

[0078] In a preferred embodiment of the present invention, the marker used to determine whether the mango pulp has been sterilized is selected from at least one of the following: isoleucine valine, valine, arginine, glycyl-L-leucine, paeonol, (-)-catechin and azelaic acid;

[0079] In a preferred embodiment of the present invention, the marker used to distinguish mango pulp that has been ultra-high pressure sterilized or mango pulp that has been sterilized at high temperature for a short time is selected from at least one of the following: (-)-catechin, methyl gallate, ferulic acid, and epigallocatechin;

[0080] In a preferred embodiment of the present invention, the application includes: obtaining the content of mango pulp processing method markers of the samples to be tested, and discriminating the mango pulp processing method of the samples to be tested based on the constructed third discriminant model of mango pulp processing method.

[0081] Reagents for detecting markers of mango pulp processing methods include, for example, reagents for liquid chromatography-mass spectrometry detection and standards for markers of mango pulp processing methods.

[0082] The features and performance of the present invention are further described in detail below with reference to the embodiments.

[0083] Example 1

[0084] This example provides a method for screening differential markers of mango pulp processed in different ways, specifically for screening non-targeted differential markers.

[0085] Experimental samples (frozen mango pulp base): collected from the factory in July 2022. The mango pulp was pulped by steam blanching at 100℃ for 1min, and then three different processing methods of mango pulp samples were prepared: non-sterilized mango pulp (NS), ultra-high pressure sterilized mango pulp (High pressure processing, HPP) (500MPa, 5min) and high-temperature short-time sterilization mango pulp (High-Temperature Short-Time Pasteurization, HTST) (90℃, 1min). Three different production batches were collected for each type of sample, and 10 bottles of samples were collected from each batch. They were immediately frozen and stored at -20℃ after collection.

[0086] 1. Liquid chromatography sample extraction method: Four solvents (pure water (H2O), 0.1% acetic acid aqueous solution (H2O+A), 50% methanol aqueous solution (Me), and 0.1% acetic acid 50% methanol aqueous solution (Me+A)) were used to extract the target compound.

[0087] Extraction procedure: Weigh 3.00 g of mango pulp, add 9 mL of extraction solvent, and vortex mix for 10 seconds. Ultrasonic extraction was performed at 4°C for 30 minutes. Centrifuge at 9000 rpm for 15 minutes. The supernatant was filtered through 0.45 μm and then 0.22 μm membranes and stored at -80°C until further use. Six replicates were prepared for each sample group, and QC samples were prepared for data quality control and instrument stability monitoring.

[0088] 2. Metabolomics analysis:

[0089] Liquid chromatography conditions: Hydrophilic interaction chromatography (HILIC) and reversed-phase chromatography (RPLC) analysis methods were established, and samples extracted from aqueous phase (H2O and H2O+A) and alcohol phase (Me and Me+A) were analyzed using HILIC and RPLC, respectively.

[0090] (1) HILIC chromatography mode: Column: Poroshell 120HILIC-Z (3.0×100 mm, 2.7 μm; Agilent, USA); Mobile phase: A: 10 mmol / L ammonium acetate aqueous solution, B: acetonitrile; Elution gradient: 0–2 min (5% A), 2–7 min (5→20% A), 7–13 min (20→32% A), 13–16 min (32→35% A), 16–18 min (5% A); Parameter settings: Injection volume 2.0 μL, column temperature 40°C, autosampler temperature 4°C, flow rate 0.50 mL / min.

[0091] (2) RPLC chromatography mode: Chromatographic column: ACQUITY UPLC BEH C18 (2.1×100 mm, 1.7 μm, (Waters, USA); mobile phase: A: 0.2% formic acid in water, B: acetonitrile; elution gradient: 0-11.5 min (5→30% B), 11.5-15 min (100% B), 15-18 min (5% B); parameter settings: injection volume 2.0 μL, column temperature 40°C, autosampler temperature 4°C, flow rate 0.30 mL / min.

[0092] Mass spectrometry conditions: Based on UHPLC-QTOF-MS (ExionLC-Triple TOF 6600, AB SCIEX, USA), data were collected in information-dependent acquisition (IDA) mode in positive and negative ion modes (ESI+ / -). (1) Scan parameters: TOF MS full scan: mass range: 100-1500 Da (accumulation time 50 ms); product ion scan: mass range: 50-1500 Da (accumulation time 30 ms). IDA cycle time: 551 ms. (2) Ion source parameters: ionization voltage (ISVF): +5500 V / -4500 V; ion source temperature: 500°C; gas parameters: nebulizer gas (Gas 1) 50 psi, auxiliary gas (Gas 2) 50 psi, curtain gas (Curtain Gas) 25 psi. (3) Energy parameters: Declustering potential (DP): ±60 V; Collision energy (CE): primary scan: ±10 eV, secondary scan: ±35 eV; Collision energy range (CES): ±15 eV.

[0093] Quality Control: Instrument stability monitoring: During operation, QC samples were collected every six injections. After every five injections, calibration fluid was automatically delivered via the CDS system for APCI mass calibration. Data acquisition was performed by Analyst TF1.7.1 to ensure experimental reproducibility and data reliability.

[0094] Component Notes:

[0095] Full component analysis: MS-DIAL software (version 5.2.4) was used for component identification. The parameter settings are as follows: the primary mass spectrometry mass tolerance is ±10ppm, the secondary mass spectrometry mass tolerance is ±15ppm, the retention time alignment threshold is 0.1min, the peak extraction MS mass tolerance is 15ppm, and the detection rate DR is ≥80%. Based on multi-dimensional verification of retention time, accurate mass (±10ppm), isotope distribution and MS / MS fragment spectrum matching, compound identification was performed through public databases such as HMDB and local self-built databases, and subsequent statistical analysis of the component characteristic peak table was performed using Microsoft Office Excel 2019. Components were screened based on the relative peak area (>5000) and detection rate (DR>80%) of the compounds in the frozen mango pulp sample. The manual MS / MS spectra corresponded poorly to the reference MS / MS spectra mirror images, and finally the phytochemical components of the frozen mango pulp were identified.

[0096] 3. Non-targeted differential marker screening: Multivariate statistical analysis was performed using the MetaboAnalyst 5.0 online platform (https: / / www.metaboanalyst.ca / ) and SIMCA 14.1 software.

[0097] (1) First, unsupervised pattern recognition was performed using principal component analysis (PCA). PCA analysis was used to evaluate the effects of different extraction and separation methods (such as differences in extraction solvents) on the content of phytochemical components in mango pulp under different processing methods.

[0098] Qualitative results: 311 phytochemical components were preliminarily annotated in mango pulp under HILIC chromatography and RPLC chromatography in positive and negative ion modes. Figure 1As shown in Table 1, the phytochemicals included 57 phenolic components, 53 amino acids, peptides, and their derivatives, 34 carboxylic acids, 25 carbohydrates, 23 lipids, 19 terpenes, 14 alkaloids, 8 vitamins, and 78 other compounds (including nucleosides, purines, and their derivatives, N-heterocyclic compounds, and amines). Fifty-one of these phytochemicals had not been previously reported, as shown in Table 1, including 17 phenols, 10 amino acids, 11 terpenes, 6 alkaloids, 3 carboxylic acids, 1 vitamin, 1 lipid, and 2 diarylheptanes.

[0099] Table 1 Information on phytochemical components not reported in mango pulp

[0100]

[0101]

[0102]

[0103] Different processing methods have a great influence on the content of phytochemical components in mango pulp. The results of principal component analysis show that mango pulp prepared by different processing methods can be clearly separated ( Figure 2 ), Figure 2 The results showed that different extraction solvents and chromatographic separation analysis methods had relatively limited effects on the content of phytochemical components in mango pulp processed by different methods. For example, the HILIC chromatography mode ( Figure 2 a) and RPLC chromatography mode ( Figure 2 In b), after extraction with the same solvent in RPLC chromatography mode, the contents of phytochemical components in mango pulp processed by different methods can be clearly separated. The combined mode of RPLC chromatography mode and HILIC chromatography mode is preferred. The influence of different solvent extraction modes is small ( Figure 2 c) and d) in the above, preferably 0.1% acetic acid 50% methanol aqueous solution is used as the extraction solvent.

[0104] There is similarity between parallel samples of the same sample in the cluster heat map ( Figure 3 ), different samples have differences.

[0105] (2) Non-targeted differential marker screening:

[0106] Orthogonal partial least squares discriminant analysis (OPLS-DA) was then used to establish a prediction model. Specifically, based on the non-targeted metabolomics data of mango pulp under any two processing methods, OPLS-DA prediction models for different mango pulp processing methods were established. The prediction models for different mango pulp processing methods were used to identify non-targeted differential markers of mango pulp under different processing methods.

[0107] The screening criteria for non-targeted differential components were as follows: (1) statistical significance (p < 0.05, Student's t-test); (2) fold difference (FC < 0.5 or FC > 2); and (3) variable projection importance (VIP > 1.0). The reliability of the model was verified by using model parameters (R2, Q2) and 200 permutation tests to determine the differential components of frozen mango pulp under different processing methods.

[0108] Difference Mining: The differences in phytochemical components of frozen mango pulp samples processed by NS, HPP and HTST were studied. OPLS-DA models were established based on the non-targeted metabolomics data of the three frozen mango pulps ( Figure 4 a, c and e) in order to find out the characteristic markers of mango pulp under different processing methods, and conduct 200 permutation tests ( Figure 4 b, d, and f in the figure) to demonstrate that the model fits well and can be used for difference analysis.

[0109] In summary, non-targeted differential markers were obtained through the above-mentioned differential mining.

[0110] Qualitative analysis of differences: Screen the compounds that differed before and after sterilization based on VIP>1, multiple difference>2 or <0.05, p<0.05 and MS-DIAL mirror image. The contents of 11 compounds (Ile-val, Valylvaline, Arginine, Gly-Leu, Paeonol, (-)-Catechin, Azelaic acid, Gulonic acid, Epigallocatechin, Ferulic acid and Methylgallate) showed significant differences before and after sterilization, of which the first 9 compounds were common differential compounds before and after sterilization. There were 4 differentially annotated compounds between the HPP and HTST groups, including (-)-Catechin, Methylgallate, Ferulic acid and Epigallocatechin, and their contents in the HPP group were significantly higher than those in the HTST group.

[0111] Example 2

[0112] This example provides a method for screening differential markers for mango pulp processed by different methods, specifically for the screening and validation of targeted differential markers. Based on the 11 differential compounds screened from the NS, HPP, and HTST sample groups in Example 1, targeted qualitative and quantitative marker validation was performed to further explore the specific changes in their content under different processing methods.

[0113] 1. Sample preparation: same as Example 1.

[0114] 2. Preparation of standards: Weigh 1.00 mg of the standard sample and place it in 1 mL of 0.1% acetic acid and 50% methanol aqueous solution. Oscillate and mix until fully dissolved to prepare a 1 mg / mL standard sample stock solution. Pass the solution through 0.45 μm and 0.22 μm syringe filters in sequence and store at -80°C until use. Prepare a mixed standard for each of the 11 standards. Dilute the mixed standard stock solution with 0.1% acetic acid and 50% methanol aqueous solution to the following concentrations: 50 mg / L, 10 mg / L, 5 mg / L, 1 mg / L, 500 μg / L, 200 μg / L, 100 μg / L, 50 μg / L, 20 μg / L, 10 μg / L, and 1 μg / L test solutions.

[0115] 3. Liquid chromatography conditions: same as in Example 1

[0116] Mass spectrometry conditions: Based on UPLC-QQQ-MS (Acquity Ultra Performance LC-Triple Quad 4500, AB SCIEX, USA), the ion source conditions were the same as in Example 1, and the characteristic ion parameters of the markers are shown in Table 2.

[0117] Table 2 Reference mass spectrometry conditions for marker characteristic ions

[0118]

[0119]

[0120] like Figure 5 As shown, the overall trend of the targeted quantification results is consistent with that of the untargeted metabolomics results, but the fold change (FC) differs slightly. For glutamate, the fold change between the NS and HTST groups in the untargeted analysis was 0.4267, while the FC in the targeted quantification results was 0.6327. For ferulic acid, the fold change between the NS and HPP groups in the untargeted analysis was 0.4875, while the FC in the targeted analysis was 0.4255. Although untargeted metabolomics can provide a broad metabolite profile, it may not be as accurate as targeted analysis in terms of quantitative accuracy.

[0121] Targeted quantitative analysis uses standards and precise quantitative methods, while optimizing sample processing and analytical conditions to improve detection sensitivity and accuracy, providing more accurate content determination compared to non-targeted methods.

[0122] Example 3

[0123] This embodiment provides a method for identifying mango pulp processing methods. Specifically, the method first screens for quasi-targeted differential markers, and then establishes and verifies a discriminant model: based on the quantitative results of the mango pulp processing method markers, multiple classification models are used to independently predict the processing type of the sample, and the prediction results of the multiple classification models are integrated to construct a discriminant model for the mango pulp processing method; and then, based on the established discriminant model, a discriminant analysis of the mango pulp samples is performed. This embodiment is based on a non-targeted differential marker screening method, and reconstructs an OPLS-DA prediction model. Through the prediction models of different mango pulp processing methods, candidate quasi-targeted differential markers of mango pulp under different processing methods are found. Then, multiple classification models are used to predict the processing type of the sample, and the candidate quasi-targeted differential markers are quantitatively verified using the UPLC-QQQ method: the quasi-targeted differential markers are screened and obtained.

[0124] QTOF non-targeted metabolomics technology has a broad spectrum coverage and can comprehensively analyze multiple components; while QQQ-based quasi-targeted metabolomics can provide better instrument stability and accuracy of quantitative analysis.

[0125] This example develops a pseudo-targeted analysis method based on non-targeted data, and systematically compares the stability (RSD), sensitivity (DR), and linear range (R) of the QQQ pseudo-targeted and QTOF non-targeted methods. 2 ) and quantitative accuracy. Based on the comparison results, the optimal discriminant model for mango pulp processing methods was selected.

[0126] 1. Sample preparation: same as Example 1.

[0127] 2. Liquid chromatography conditions: same as in Example 1.

[0128] Data acquisition: IDA acquisition mode was the same as in Example 1. SWATH acquisition mode: QC samples were extracted using an alcohol phase (Me+A) and SWATH data acquisition was performed. Six injections were repeated to improve data reliability. SWATH acquisition mode (cycle time: 551 ms) consisted of a TOF MS scan (accumulation time: 50 ms) and a high-sensitivity product ion scan (accumulation time: 50 ms). Mass spectrometry parameters were set as follows: ion source: electrospray ionization (ESI±); scan range: primary: 100-1500 Da, secondary: 50-1500 Da (15 variable windows); energy parameters: ISVF: +5500 V / -4500 V, DP: ±60 V, CE: ±10 eV, CES: ±15 eV; temperature and pressure: ion source temperature: 500°C, Gas 1: 50 psi, Gas 2: 50 psi, Curtain Gas: 25 psi. SWATH Window Optimization: Peak detection was performed using PeakView 2.2 software based on TOF-MS data acquired in IDA mode. The following parameters were set: LC peak width: 6 s, minimum intensity: 5 counts, chemical noise factor: 1.5, and threshold setting: 5000. After exporting the LC / MS peak data, the swathTuner package in R was used to optimize the SWATH window distribution based on the total ion current balance within each window. The m / z range was set to 100–1500 Da, with a window overlap of 1 Da. The 15 optimized variable SWATH windows were imported into Analyst 1.8.1 software to establish the final acquisition method to reduce cross-fragmentation interference and improve selectivity.

[0129] Targeted MRM ion pair generation: UHPLC-QTOF mode QC data (mzXML format) were converted using MSConvert 3.0. Ion pairs were generated using the MRM-Ion Pair Finder package in R. Parameters were set as follows: m / z deviation 15 ppm, peak width 5–15 s, signal-to-noise ratio (S / N) 15, and MS1 / MS2 intensity limits 400 / 80 counts.

[0130] Targeted data acquisition: A Schedule MRM method was constructed using Analyst 1.6.3 with the following parameters: MRM detection window: 60 s, cycle time: 600 ms. Instrument stability was monitored: a QC sample was collected after every six samples, and APCI automated mass calibration was performed after every five samples. Data acquisition was performed using Analyst TF 1.7.1 to ensure experimental reproducibility and data reliability.

[0131] 4. Discriminant Model Data Processing: A discriminant model was established using the MATLAB 2022 platform. The modeling samples were all from Guangxi, including three groups of HPP samples (experimental samples: Guangxi Pilot (HPP), cooperative enterprise samples: Guangxi Origin (3H), Guangxi Ice Guest (4H)); and three groups of HTST samples (experimental samples: Guangxi Pilot (HTST), cooperative enterprise samples: Guangxi Ice Guest (15R, 16R). Input features were normalized to the [0, 1] interval using the mapminmax function to eliminate dimensionality differences. The randperm function was used to randomly partition the dataset (training set:test set = 7:3). Cross-validation was used to assess stability during model training, and classification performance was ultimately verified using the test set.

[0132] Based on the relative contents of the above-mentioned screening difference components, the classification and discrimination model was constructed using Matlab 2022 software. The modeling sample information and data processing steps are as follows:

[0133] Modeling sample information: A single model was established using the pseudo-targeted data results of three groups of HPP mango pulp (HPP, 3H, 4H) and three groups of HTST mango pulp (HTST, 15R, 16R) produced in Guangxi.

[0134] Data processing: OPLS-DA models were constructed using SIMCA 14.1 software. Differential components with FC > 1.7 or < 0.6 and p < 0.05 were selected. Recursive feature elimination (RFE) was performed on the MATLAB 2022b platform to identify the top 30% of important features. Data were normalized using mapminmax and randomly split into training and test sets with a ratio of 7:3. To address insufficient data, SMOTE oversampling (nearest neighbor number k = 5) was used to enhance training set stability.

[0135] 5. Eight machine learning classification models were established for discriminant analysis of mango pulp processing methods, including random forest (RF), support vector machine (SVM), partial least squares (PLS), least squares support vector machine (LSSVM), backpropagation neural network (BPNN), convolutional neural network (CNN), extreme learning machine (ELM), and particle swarm optimization backpropagation neural network (PSO-BP). The calculated metrics for the different classification modeling results are shown in Table 3. With the exception of the support vector machine (SVM) model, which performed slightly worse, the remaining seven models achieved 100% classification accuracy on both the training and test sets, demonstrating excellent discriminant performance.

[0136] 6. The above classification models were further verified using 12 commercially available samples from Guangxi. It was found that the SVM model performed poorly and showed obvious classification inconsistency. Although the ELM model performed well in the modeling data, its accuracy dropped to 75% in the commercial sample verification, indicating a tendency to overfitting. The BPNN and CNN models performed best in the commercial verification (91.67% accuracy), indicating that they have strong feature extraction capabilities and stability. The discriminative differences between different models in commercial verification may be due to the fact that different models may rely on different feature selection mechanisms. Some models may be sensitive to specific features, while other models may ignore these features, resulting in different classification results. For example, PLS is sensitive to linear features but ignores nonlinear features. In addition, the decline in the accuracy of SVM and PLS commercial sample verification also shows that the models are dependent on training data and have limited generalization capabilities. Based on the limitations of the above-mentioned single classification model in commercial sample verification, in order to further improve the stability and reliability of the classification system, the hard voting method was used to perform multi-model integration analysis on the seven best-performing base classifiers RF, PLS, LSSVM, BPNN, CNN, ELM and PSO-BP to further improve the discrimination consistency of commercial samples.

[0137] Table 3: Calculation indicators of classification model modeling results

[0138]

[0139] 7. Multi-model integration optimization discrimination

[0140] The voting method of multi-model fusion (i.e., ensemble learning method) can effectively address the weaknesses of a single model and improve the accuracy and robustness of discrimination by combining the judgments of multiple models, including voting, stacking, bagging, and boosting (Kuncheva 2014; Shareef and Kurnaz 2023). Figure 6 As shown in Figure a, in order to improve the discrimination accuracy of the classification model, the hard voting method is used to integrate the prediction results of seven base classifiers: random forest (RF), partial least squares (PLS), least squares support vector machine (LSSVM), back propagation neural network (BPNN), convolutional neural network (CNN), extreme learning machine (ELM) and particle swarm optimization BP neural network (PSO-BP). Figure 6 As shown in (b), the decision boundary diagram shows that the HPP and HTST samples have a significant separation trend in the PC1-PC2 space, forming a clear boundary, indicating that the model has high discrimination.

[0141] like Figure 7 A and Figure 7 As shown in b, the accuracy of the integrated model in both the training set and the test set is 100%. Figure 7 Middle C and Figure 7 The ensemble model correctly classified all sample types without misclassification, demonstrating high classification stability and reliability. Therefore, the ensemble model was used to identify the processing type of commercially available samples.

[0142] Example 4

[0143] Application of identification model for mango pulp with different processing methods.

[0144] Based on the comparison results, the integrated model in Example 3 was selected to construct a discriminant model for mango pulp processing methods. The model was then validated using commercially available samples. Commercial mango juice / pulp samples were purchased from 40 e-commerce platforms (Taobao, JD.com, and Meituan), covering 12 production areas domestically and internationally (including two overseas).

[0145] Based on the established multi-model fusion method, a discriminant analysis was performed on 40 purchased commercial samples. As shown in Table 4, among the 26 samples with known processing methods, the model's discriminant results were consistent with the manufacturer's claims in 22 groups, accounting for 84.62%. The discriminant results for four groups, accounting for 15.38%, differed from the manufacturer's claims. Among the samples with discrepant results, a mango juice sample (1H) from Thailand, among the HPP samples, was identified as a HTST sample. Among the heat-pasteurized samples, a mango pulp sample (22R) from Hainan and a mango juice sample (12R, 13R) from Henan were identified as HPP samples.

[0146] like Figure 8 As shown in the decision diagram of the model discrimination results, 12R and 13R are in the overlapping area of ​​the HPP / HTST decision boundary, indicating that the confidence level of the sample processing method discrimination is low. 1H and 23R form clusters far away from the boundary and are respectively located in the HTST / HPP category cluster centers, indicating that there may be errors in the manufacturer's labeling of the processing method.

[0147] Table 4 Model discrimination results of commercially available samples

[0148]

[0149] Note: “None” means that no samples were collected in the treatment group. The superscript symbol ★ after the serial number indicates that the model discrimination result of the group sample is inconsistent with the actual result.

[0150] Based on the established multi-model fusion method, discriminant analysis was performed on 14 groups of samples with unknown processing methods among the 40 purchased commercial samples. The results are shown in Table 5.

[0151] The application results of the model show its potential discriminative ability, but in its actual promotion and application, due to the influence of raw material variety, regional differences and processing parameters on the content of phytochemical components, the applicability of the model still needs further verification and optimization.

[0152] Table 5 Model discrimination results of commercial samples with unknown processing methods

[0153]

[0154]

[0155] Note: NA: Not Available, indicating that relevant information is not provided.

[0156] In summary, this invention systematically screened and verified for the first time the characteristic markers that can accurately distinguish mango pulp processed by different methods by combining non-targeted metabolomics with quasi-targeted metabolomics and targeted quantitative analysis. The characteristic markers screened by this method can be used to construct a discriminant model for mango pulp processing methods. The established multi-model fusion discrimination method achieves high-precision and high-reliability processing method discrimination, greatly improving the efficiency and accuracy of mango pulp processing method identification, and providing new technical means and theoretical basis for product quality control and market supervision.

[0157] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for screening markers of mango pulp processing methods, characterized in that: It includes at least one of the following methods: screening of non-targeted differential markers, screening of proposed targeted differential markers, and screening and verification of targeted differential markers; The screening of non-targeted differential markers comprises the following steps: S1: Non-targeted metabolomics analysis: The samples were subjected to non-targeted metabolomics analysis using high-resolution chromatography-mass spectrometry to identify the phytochemical components in mango pulp; S2: Screening of non-targeted differential markers: Based on the non-targeted metabolomics data of mango pulp under any different processing methods, a first discriminant model of different mango pulp processing methods was established. Through the first discriminant model of different mango pulp processing methods, combined with statistics, the non-targeted differential markers of mango pulp under different processing methods were found; the screening rules for non-targeted differential markers were as follows: VIP value greater than 1, difference fold FC greater than 2 or less than 0.5, and p value less than 0.

05. The screening of the differential markers to be targeted includes the following steps i or ii: i: Based on the identification results of phytochemical components in mango pulp or the non-targeted differential marker screening results obtained by the non-targeted metabolomics analysis, a quantitative analysis instrument is used to perform precise quantitative analysis of the phytochemical components to obtain pseudo-targeted metabolomics data; ii: Based on the pseudo-targeted metabolomics data of mango pulp under any different processing methods, a second discriminant model for different mango pulp processing methods was established. Through the second discriminant model for different mango pulp processing methods, combined with statistics, pseudo-targeted differential markers of mango pulp under different processing methods were identified; the screening rules for pseudo-targeted differential markers were as follows: VIP value greater than 1, differential fold greater than 2 or less than 0.5, and p-value less than 0.05; The screening and verification of the targeted differential markers includes the following steps: Based on the screened non-targeted differential markers and / or the intended targeted differential markers, marker screening and targeted verification are performed; existing target compound standards are screened out from the non-targeted differential markers and / or the intended targeted differential markers as candidate targeted differential markers; the candidate targeted differential markers are qualitatively and quantitatively verified, and targeted differential markers with obvious differences under different processing methods are screened out through the qualitative and quantitative verification results.

2. The method for screening markers of mango pulp processing methods according to claim 1, characterized in that: During the non-targeted metabolomics analysis, the samples to be tested obtained by different extraction methods are subjected to chromatographic separation using one or more chromatographic modes, and are connected in series with a high-resolution mass spectrometer for non-targeted metabolomics analysis; Preferably, the chromatography mode is selected from HILIC, RPLC, ion exchange chromatography, hydrophobic interaction chromatography, size exclusion chromatography, supercritical fluid chromatography or a mixed chromatography mode; Preferably, the chromatography mode is a combination of HILIC and RPLC chromatography modes; Preferably, the statistical analysis method in the screening of non-targeted differential markers and the screening of intended targeted differential markers is selected from: analysis of variance, fold difference analysis, Mann-Whitney U test, Wilcoxon signed rank test, Kruskal-Wallis H test, Friedman test, Bonferroni correction or FDR correction; The high-resolution chromatography mass spectrometer is selected from LC-QTOF-MS, GC-TOF-MS or LC-Orbitrap-MS; Preferably, the quantitative analysis instrument for the analysis of the targeted metabolites is LC-QQQ or DART-QQQ; Preferably, the first discriminant model or the second discriminant model is an OPLS-DA or PLS-DA model, and when screening the non-targeted differential markers, multivariate statistical analysis is performed using MetaboAnalyst 5.0 online platform, Excel, or SIMCA 14.1 software; Preferably, before the metabolomics analysis of the sample to be tested is performed, the sample to be tested is further pretreated, and the pretreatment comprises the following steps: taking a mango pulp sample, extracting it with a 0.05%-0.20% acetic acid aqueous solution or with a 30-70% (v / v) methanol aqueous solution containing 0.05%-0.20% acetic acid, taking the supernatant, and filtering; Preferably, the candidate targeted differential markers are qualitatively and quantitatively validated using at least one of the following methods: LC-MS, GC-MS, NMR, NIR, DART-MS, matrix-assisted laser desorption ionization (MALDI)-MS and ion mobility mass spectrometry (IMS-MS).

3. A method for identifying a mango pulp processing method, characterized in that: Obtain at least one of the following mango pulp processing mode markers according to the screening method for mango pulp processing mode markers according to any one of claims 1 to 2: a non-targeted differential marker, a pseudo-targeted differential marker, and a targeted differential marker; Then, a third discriminant model was established and verified: Based on the quantitative results of mango pulp processing method markers, multiple classification models were used to independently predict the processing type of the samples. The prediction results of multiple classification models were integrated to construct a third discriminant model for mango pulp processing methods. Then, a discriminant analysis is performed on the mango pulp samples based on the established third discriminant model.

4. The method for identifying a mango pulp processing method according to claim 3, wherein: The prediction results of integrating multiple classification models include: obtaining the final classification results of mango pulp processing methods through voting or weighting; Preferably, the results of different classification models for processing methods are counted, the results of the processing methods are voted or weighted, and the processing method with the largest number of votes or the largest weight is selected as the final classification result of the mango pulp processing method.

5. The method for identifying a mango pulp processing method according to claim 4, wherein: The classification model is selected from at least three of the following: At least one of random forest (RF), support vector machine (SVM), k-nearest neighbor, partial least squares (PLS), least squares support vector machine (LSSVM), BP neural network (BPNN), convolutional neural network (CNN), extreme learning machine (ELM), particle swarm optimization BP neural network (PSO-BP), recurrent neural network, Bayesian classifier, K-nearest neighbor algorithm, K-means algorithm, linear regression, logistic regression, multivariate nonlinear regression fitting method, Adaboost algorithm, hidden Markov model, decision tree algorithm, clustering algorithm, generative adversarial network, stacked autoencoder, fully connected network, unsupervised pre-trained network, deep belief network, deep Boltzmann machine and neural tensor network.

6. The method for identifying a mango pulp processing method according to claim 3, wherein: By comparing the differences in accuracy, stability, sensitivity, linear range and coverage of non-targeted differential markers, quasi-targeted differential markers and targeted differential markers, mango pulp processing method markers were selected according to the comparison results to construct a third discriminant model for mango pulp processing methods.

7. The method for identifying a mango pulp processing method according to claim 3, wherein: The discriminant analysis of mango pulp samples based on the established third discriminant model includes the following steps: obtaining the content of the mango pulp processing method marker in the sample to be tested, and discriminating the mango pulp processing method of the sample to be tested based on the third discriminant model.

8. An electronic device, characterized in that: It includes: processor and memory; The processor is connected to the memory, wherein: The memory is used to store a computer program, and the processor is used to call the computer program to execute the method for identifying a mango pulp processing method according to any one of claims 3 to 7.

9. A computer-readable storage medium, characterized in that The method comprises instructions, which, when executed on a computer, enable the computer to execute the method for identifying the mango pulp processing method according to any one of claims 3 to 7.

10. The use of a reagent for detecting a marker of mango pulp processing mode in distinguishing mango pulp processing mode, characterized in that: The mango pulp processing method marker is selected from at least one of the following: isoleucine valine, valine, arginine, glycyl-L-leucine, paeonol, (-)-catechin, azelaic acid, gulonic acid, epigallocatechin, ferulic acid and methyl gallate; Preferably, the marker for determining whether the mango pulp has been sterilized is selected from at least one of the following: isoleucine valine, valine, arginine, glycyl-L-leucine, paeonol, (-)-catechin and azelaic acid; Preferably, the marker for distinguishing mango pulp that has been ultra-high pressure sterilized or mango pulp that has been sterilized at high temperature for a short time is selected from at least one of the following: (-)-catechin, methyl gallate, ferulic acid and epigallocatechin; Preferably, the application includes: obtaining the content of the mango pulp processing method marker of the sample to be tested, and discriminating the mango pulp processing method of the sample to be tested based on the constructed third discriminant model of the mango pulp processing method.

Images