Metabolites of Chenopodium album honey and untargeted metabolomics combined with supervised learning for identification of Chenopodium album honey

Abstract

The present invention provides a method for identifying the metabolites of Veratrum honey and non-target metabolomics combined with supervised learning, and relates to the technical field of identification of Chinese bee honey varieties. Cuminaldehyde, apigenin, p-isobutylbenzoic acid, acetyl dopamine, oligomaltose, and abscisic acid are provided as combined metabolites to accurately distinguish Veratrum honey from other honeys, and non-target metabolomics technology is used to obtain metabolites of Veratrum honey harvested at different times. At the same time, metabolites of other Chinese bee honeys are identified; multiple supervised learning is used to distinguish and optimize Chinese bee honey and Veratrum honey. The present invention overcomes the shortcomings of the prior art, can accurately and quickly identify Veratrum honey, and can locally identify the variety and authenticity of Veratrum honey.

Description

Technical Field

[0001] The invention relates to the technical field of identification of Chinese bee honey varieties, and in particular to a method for identifying metabolites of quinoa cone honey and identifying quinoa cone honey by combining non-target metabolomics with supervised learning. Background Art

[0002] Castanopsis fissa (Champ.ex Benth.) is a tree of the Fagaceae family and the genus Castanopsis. It is mainly distributed in the southern part of Jiangxi, Hunan, Yunnan, and Guizhou provinces. Castanopsis fissa is an auxiliary nectar plant with certain medicinal effects. It is known as "clearing away heat and dampness, clearing the liver and improving eyesight." Castanopsis fissa honey is a natural sweetener brewed from the nectar, secretions or honeydew collected by Chinese honey bees. It is a special economic Chinese honey derived from the Castanopsis fissa industry.

[0003] At present, there are no relevant standards or testing methods for the identification of quinoa honey varieties, which makes it difficult to ensure the quality of quinoa honey and market supervision. At the same time, driven by profit, unscrupulous vendors began to sell quinoa honey under false labels, which triggered a crisis of trust among consumers and was not conducive to the healthy development of special high-quality honey brands and the high-value utilization of quinoa honey.

[0004] In terms of the identification of special Chinese bee honey varieties, invention patented methods such as PCR testing, volatile hydrocarbons, kynurenic acid, 3-amino-2-naphthoic acid, indole-3-acetic acid methyl ester, quercetin and 3,10-dihydroxydecanoic acid content have been reported to be used to distinguish between Chinese bee honey and Italian bee honey (CN202111079089.1, CN202210768906.2, CN202011120746.8, CN202010197039.2, CN202010197054.7, CN202010197053.2, CN202210646486.0).

[0005] However, there are many varieties of single nectar of Chinese bee honey, including longan, lychee, nut, Schefflera, pomelo, ginseng, Chinese tallow tree, chestnut, gallnut, etc., and these public methods cannot distinguish the varieties of different Chinese bee honey. Invention patent CN202210812726.X discloses a method for identifying Chinese bee honey varieties based on a visualized mass spectrometry molecular network diagram, which distinguishes Chinese bee honey varieties based on whether the mass spectrometry molecular network diagram is the same. The method is to treat each mass spectrum as a vector and compare it with all other mass spectra using cosine similarity. When the similarity between two mass spectra is higher than a threshold, they will be connected together in the molecular network. Considering that the composition of Chinese bee honey collected in different seasons or Chinese bee honey of different maturity is significantly different, and there are highly similar but not identical components between different Chinese bee honeys, therefore, judging only by whether they are the same or not, the results have obvious limitations and high false positives. In contrast, non-targeted metabolomics combined with supervised learning methods can better determine the quality of Chinese bee honey from the overall metabolite types and changes. At the same time, this method can also obtain key differential metabolites, and the Chinese bee honey varieties can be further identified based on the types and contents of these compounds. At present, there are no research and patent reports on the identification method of the variety of Veratrum honey. The development of the identification method of the variety of Veratrum honey can provide a strong guarantee for the quality and authenticity identification of Veratrum honey. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides a method for identifying metabolites of Veratrum honey and non-target metabolomics combined with supervised learning to identify Veratrum honey. The method is based on a high-throughput non-target metabolomics method and multiple supervised learning to optimize and screen the best model, which can achieve the overall distinction between the Veratrum honey variety and the other 9 types of Chinese honey varieties. At the same time, multiple key differential metabolites are obtained according to the SHAP value to further identify the Veratrum honey variety locally.

[0007] To achieve the above objectives, the technical solution of the present invention is implemented through the following technical solutions:

[0008] The metabolites used for identifying the honey of Veratrum cones are a combination of cuminaldehyde, 1-Stearoylglycerol, thulyl acid, apigenin, p-isobutylbenzoic acid, acetyl dopamine, maltooligosaccharide, abscisic acid and L-carvone.

[0009] Preferably, the metabolites used to identify the honey of Veratrum cones are a combination of cuminaldehyde, apigenin, p-isobutylbenzoic acid, acetyl dopamine, maltooligosaccharides, abscisic acid and L-carvone.

[0010] Metabolites are used to accurately distinguish Quinoa honey from other Chinese bee honeys, and the metabolites are used to identify the variety, quality and authenticity of Quinoa honey.

[0011] A method for identifying Chenopodium album honey by combining non-target metabolomics with supervised learning, characterized in that the method for identifying Chenopodium album honey comprises the following steps:

[0012] S1. Optimization of non-target metabolomics identification technology;

[0013] S2. Identify the metabolites of Chenopodium album honey collected at different times and the metabolites of other Chinese honey;

[0014] S3, common marker rapid matching;

[0015] S4. Use multiple supervised learning methods for training, including eight classification algorithms: xgboost, lightgbm, Random Forest, ExtraTrees, GaussianNB, KNeighbors, DecisionTree and SVC;

[0016] S5. Use SHAP value to obtain key differential metabolites;

[0017] S6. Accurately distinguish Quinoa honey from other Chinese bee honeys based on key differential metabolites, and use metabolites to identify the variety, quality and authenticity of Quinoa honey.

[0018] The non-target metabolomics identification technology in S1 is optimized to optimize the separation conditions of the compounds, and specifically includes the following steps:

[0019] S1-1, set the mobile phase to 0.1-0.5% formic acid and 1-5mmol / L ammonium formate aqueous solution as C, and 0.1-0.5% formic acid and 1-5mmol / L ammonium formate methanol solution as B;

[0020] S1-2, set the gradient elution program as follows: 0-1 min, 99% C; 1-5 min, 99-15% C; 5-25 min, 15-1% C; 25-30 min, 1% C; 30-31 min, 1-99% C; 31-35 min, 99% C, flow rate 0.25 mL / min, injection volume 0.2 μL;

[0021] S1-3. Use ultra-high performance liquid chromatography-mass spectrometry to identify non-target metabolomics components and obtain raw data.

[0022] The other Chinese bee honeys in S2 include longan, litchi, nut, Schefflera australis, pomelo, ginseng, Chinese tallow tree, chestnut, and gallnut honey.

[0023] The metabolite identification method is to first obtain the extract of each honey component, then use the optimized separation conditions combined with mass spectrometry equipment to identify each honey to obtain the original data, further align the original data, RT correction, standardization and peak area processing, and save the processed data as Excel or store it as a database in the form of a dictionary according to the following formula:

[0024] database = {"type":"data"}

[0025] {"data":{"metabolite type":X i , "Relative content": Y i , "Retention time": Z i}}

[0026] The species include longan, litchi, nut, Schefflera chinensis, pomelo, ginseng, Chinese tallow tree, chestnut, gallnut and quinoa honey; metabolite types refer to the metabolite types identified by co-identification, i means there are i metabolites in total, where X i is a list of common i metabolites, Y i is the relative content of the common metabolites, Z i is the retention time RT value of the i-th metabolite.

[0027] The training using multiple supervised learning in S4 includes the following steps:

[0028] S4-1. Select Random Forest as the supervised learning method;

[0029] S4-2. Optimize the Random Forest parameters to n_estimators: 50-200, max_depth: 15-18, min_samples_leaf: 1-8, min_samples_split: 2-5, max_features: 0.1-1.

[0030] The method for obtaining key differential metabolites using SHAP values ​​in S5 includes the following steps:

[0031] S5-1. Import the Random Forest model into the SHapleyAdditive exPlanations package to perform SHAP calculation and obtain the SHAP value of each feature.

[0032] S5-2, defined as key differential metabolites based on SHAP values ​​greater than 0.1.

[0033] The present invention provides a method for identifying metabolites of Chenopodium album honey and identifying Chenopodium album honey by combining non-target metabolomics with supervised learning, which has the advantages of:

[0034] (1) The present invention is the first to develop a method for identifying the variety of Chenopodium album honey;

[0035] (2) The present invention uses a non-targeted metabolomics combined with supervised learning to identify the varieties of Chenopodium album honey with high efficiency and accuracy;

[0036] (3) The present invention can distinguish the Chenopodium album honey variety from the other nine kinds of Chinese honey varieties as a whole;

[0037] (4) The present invention provides a combination of six key differential metabolites: cuminaldehyde, apigenin, p-isobutylbenzoic acid, acetyl dopamine, oligomaltose and abscisic acid, which can be used to locally identify the variety and authenticity of Chenopodium album honey. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic diagram of the principal component analysis used in Example 3 to distinguish Chenopodium album honey from Apis cerana honey;

[0039] Figure 2 This is a schematic diagram of the prediction results of the Random Forest model in Example 3 of the present invention for distinguishing Chenopodium album honey from Apis cerana honey;

[0040] Figure 3 Schematic diagram of the structures of 9 key differential metabolites screened based on SHAP values ​​in Example 3 of the present invention;

[0041] Figure 4 This is a box plot of the relative contents of the six key metabolite compounds with significant differences in Example 3 of the present invention in Chenopodium album honey and Apis cerana honey. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention is clearly and completely described below in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Embodiment 1:

[0044] A total of 32 mature Chenopodium album honey samples (mature when the honeycomb capping rate was greater than 95% as observed by visual observation) were obtained from Jiangxi, Hunan, Yunnan and Guizhou regions and named LS, with numbers ranging from LS1 to LS32.

[0045] Accurately weigh 5 g of the above-mentioned Chenopodium album honey into a centrifuge tube, add 10 mL of deionized water and vortex until mixed; then, add 10 mL of methanol for extraction, centrifuge at 9000 rpm for 30 min, and then filter the supernatant with a 0.22 μm filter membrane to obtain a Chenopodium album honey component extract.

[0046] Take 5 mL of methanol and 5 mL of pure water to activate the Agilent SPE column, load the above-mentioned Veratrum cone honey component extract into the Agilent SPE column, and rinse the column with 3 mL of water after the loading is completed; further elute with methanol twice, 4 mL each time, collect the eluate for nitrogen blowing and concentration, and re-dissolve with 1 mL of methanol; use ultra performance liquid chromatography-mass spectrometry equipment to extract the Veratrum cone honey component extract for non-targeted component identification, and obtain the metabolite data of the Veratrum cone honey component extract.

[0047] The detection method of the ultra-high performance liquid chromatography-mass spectrometry equipment is as follows: the chromatographic column is EclipsePlus C18RRHD (3.0mm×150mm, 1.8μm). The mobile phase is 0.1% formic acid-5mmol / L ammonium formate aqueous solution (C) and 0.1% formic acid-5mmol / L ammonium formate methanol solution (B). The gradient elution program is: 1min, 99% C; 5min, 15% C; 5min, 15% C; 25min, 1% C; 30min, 99% C; 35min, 99% C, flow rate 0.25mL / min. The injection volume is 0.2μL.

[0048] Full scan MS and secondary MS modes were performed. The main mass spectrometry parameters were: sheath gas flow rate was set to 50Arb; aux gas flow rate was set to 12Arb; capillary temperature: 330℃; full MS resolution, 70000; AGC target for full MS was 3e6; MS / MS resolution, 17500; AGC target forMS / MS, 1e5; MS / MS isolation window, 4.0m / z; NCE, 30; spray voltage, 3.5kV (negative) or 4.0kV (positive).

[0049] Embodiment 2:

[0050] A total of 9 types of Chinese bee honey were obtained from Jiangxi and Hunan, namely longan honey, litchi honey, nut honey, Schefflera honey, pomelo honey, ginseng honey, Chinese tallow honey, chestnut honey and gallnut honey. The 9 types of reported Chinese bee honey were named ZF. Accurately weigh 5g of the above ZF honey in a centrifuge tube, add 10mL of deionized water and vortex until mixed; then, add 10mL of methanol for extraction, centrifuge at 9000rpm for 30min, and then filter the supernatant with a 0.22μm filter membrane to obtain the ZF honey component extract. Take 5mL of methanol and 5mL of pure water to activate the Agilent SPE column, load the above ZF honey component extract into the Agilent SPE column, and rinse the column with 3mL of water after loading; further elute with methanol twice, 4mL each time, collect the eluate for nitrogen concentration, and re-dissolve with 1mL of methanol; use ultra-performance liquid chromatography-mass spectrometry to extract the ZF honey component extract for non-targeted component identification, and obtain the metabolite data of the ZF honey component extract.

[0051] The detection method of the ultra-high performance liquid chromatography-mass spectrometry equipment is as follows: the chromatographic column is EclipsePlus C18RRHD (3.0mm×150mm, 1.8μm). The mobile phase is 0.1% formic acid-5mmol / L ammonium formate aqueous solution (C) and 0.1% formic acid-5mmol / L ammonium formate methanol solution (B). The gradient elution program is: 1min, 99% C; 3 minutes, 99% C; 15min, 10% C; 25min, 1% C; 30min, 99% C; 35min, 99% C, flow rate 0.25mL / min, injection volume 0.2μL.

[0052] Full scan MS and secondary MS modes were performed. The main mass spectrometry parameters were: sheath gas flow rate was set to 50Arb; aux gas flow rate was set to 12Arb; capillary temperature: 330℃; full MS resolution, 70000; AGC target for full MS was 3e6; MS / MS resolution, 17500; AGC target forMS / MS, 1e5; MS / MS isolation window, 4.0m / z; NCE, 30; spray voltage, 3.5kV (negative) or 4.0kV (positive).

[0053] Embodiment 3:

[0054] The raw data obtained from implementation 1 and implementation 2 were aligned, RT corrected, standardized and processed for peak area, and the processed data were saved as Excel or stored as a database in dictionary form according to the following formula:

[0055] database = {"type":"data"}

[0056] {"data":{"metabolite type":X i , "Relative content": Y i , "Retention time": Z i}}

[0057] The species include longan, litchi, nut, Schefflera chinensis, pomelo, ginseng, Chinese tallow tree, chestnut, gallnut and quinoa honey; metabolite types refer to the metabolite types identified by co-identification, i means there are 129 metabolites, among which X 129 is a list of 129 metabolites, Y 129 is the relative content of 129 metabolites, Z 129 RT values ​​of the retention times of 129 metabolites.

[0058] The metabolite types and relative contents in ZL and LS data were extracted and principal component analysis was performed. Figure 1 shown.

[0059] It can be found that the first two PCA1 and PCA2 can explain the difference between ZL and LS by 34.21% and 16.45% respectively, and the sum of the two is greater than 50%, indicating that the principal component analysis method can distinguish ZL and LS varieties, among which PCA1 and PCA2 can explain the global changes.

[0060] The supervised learning Random Forest algorithm was further used to construct a binary classification model to distinguish between ZL and LS varieties, where the Random Forest algorithm parameters were as follows: n_estimators: 100, max_depth: 15, in_samples_leaf: 2, min_samples_split: 2, and max_features: 0.4.

[0061] The results are as follows Figure 2 As shown, Figure 2 This is the confusion matrix result of distinguishing ZL and LS varieties under 5-fold cross validation. The results show that Random Forest can accurately distinguish ZL and LS varieties.

[0062] The best Random Forest model was imported into the SHapleyAdditive exPlanations package for SHAP calculation, and nine key metabolite compounds were obtained according to mean SHAP>0.1: cuminaldehyde, 1-Stearoylglycerol, sulinemic acid, apigenin, p-isobutylbenzoic acid, acetyl dopamine, maltooligosaccharide, abscisic acid and L-carvone, such as Figure 3shown.

[0063] The relative contents of 9 key metabolites in ZL and LS were further analyzed. Figure 4 As shown. Cumaldehyde, apigenin, p-isobutylbenzoic acid, acetyl dopamine, maltooligosaccharides and abscisic acid have significant differences in ZL and LS. In summary, the method provided by the present invention can predict the variety of Veratrum honey based on the overall metabolites according to the best model Random Forest. At the same time, a combined metabolite of cumaldehyde, apigenin, p-isobutylbenzoic acid, acetyl dopamine, maltooligosaccharides and abscisic acid can be used to accurately distinguish Veratrum honey from the other 9 types of Chinese honey, and can be used for the identification of the variety, quality and authenticity of Veratrum honey.

[0064] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying metabolites of Chenopodium album honey, characterized in that: The metabolites used for identifying the honey of Chenopodium album are a combination of cuminaldehyde, 1-Stearoylglycerol, thulyl acid, apigenin, p-isobutylbenzoic acid, acetyl dopamine, maltooligosaccharide, abscisic acid and L-carvone.

2. A method for identifying metabolites of Chenopodium album honey, characterized in that: The metabolites used for identifying the honey of Chenopodium album are a combination of cuminaldehyde, apigenin, p-isobutylbenzoic acid, acetyl dopamine, oligomaltose, abscisic acid and L-carvone.

Images