Citrus huanglongbing marker combination based on metabonomics and screening method thereof

Abstract

The invention discloses a metabonomics-based Citrus Huanglongbing marker combination and a screening method thereof, and relates to a method for screening biomarkers of an asymptomatic Citrus Huanglongbing sample, metabolite screened by the method is adopted to further construct a prediction model, and the prediction model is used for detecting Citrus Huanglongbing in asymptomatic leaves. The biomarker of the asymptomatic sample can be screened through the method, and the method is of great significance to improvement of diagnosis accuracy.

Description

Technical Field

[0001] This invention belongs to the field of citrus disease detection technology, specifically relating to a combination of biomarkers for citrus Huanglongbing (HLB) based on metabolomics and its screening method. Background Technology

[0003] Huanglongbing (HLB) is an extremely serious and devastating disease in the citrus industry. Its primary cause is infection by *Bacillus phloemellatus*, which is transmitted by the citrus psyllid. After infection, citrus plants exhibit a range of symptoms, including yellowing shoots, weakened tree vigor, mottled yellowing of leaves, slow growth, and uneven fruit coloring, ultimately leading to plant death.

[0004] Currently, there is no effective treatment for Huanglongbing (HLB) of citrus, nor are there any completely resistant citrus varieties. Therefore, early detection of infected citrus trees is crucial for reducing the source of infection in the field and improving control effectiveness. However, due to the low concentration and uneven distribution of HLB pathogens in citrus plants, traditional molecular detection techniques such as PCR may produce false negative results. This false negative situation is especially common in citrus trees that have not yet shown symptoms in the early stages of infection.

[0005] To address this issue, early diagnostic techniques are used to identify and remove diseased trees as early as possible, thereby reducing the source of infection in the field and improving control effectiveness. Meanwhile, the Huanglongbing pathogen is unevenly distributed within the plant, especially on asymptomatic leaves, making accurate detection via PCR difficult.

[0006] Therefore, developing more sensitive and accurate detection technologies is crucial for controlling citrus Huanglongbing (HLB). After citrus plants are infected by HLB, they exhibit a series of responses, leading to changes in various primary and secondary metabolites within the plant. Based on these characteristics, metabolites with significant differential changes, high specificity, and high sensitivity within infected citrus plants can be screened as biomarkers, providing a theoretical basis and practical evidence for the detection of HLB. Summary of the Invention

[0007] To address the difficulty of diagnosing asymptomatic citrus Huanglongbing (HLB) samples using existing PCR detection methods, this invention provides a metabolomics-based combination of HLB biomarkers and a screening method for them. This method is a way to screen biomarkers for asymptomatic HLB samples. The metabolites screened using this method are further used to construct a predictive model and then used to detect HLB in asymptomatic leaves. This method can screen for biomarkers in asymptomatic samples, which is of great significance for improving diagnostic accuracy.

[0008] This invention provides the following solution: This invention provides a method for screening biomarkers suitable for asymptomatic samples of citrus Huanglongbing (HLB), comprising: (1) Collect leaf samples from infected and healthy plants as analytical samples; (2) Non-targeted metabolomics analysis was used to screen for differential metabolites between healthy controls and citrus Huanglongbing positive asymptomatic leaf samples; (3) Based on the differentially expressed metabolites screened by non-targeted metabolomics analysis, the random forest algorithm is used to screen metabolites with high contribution to the differential expression and to calculate the minimum number of biomarkers. (4) The area under the curve (AUC) of each biomarker and biomarker combination was calculated using ROC curves. When 0.7 > AUC > 0.5, the biomarker or combination was considered effective; when 0.9 > AUC > 0.7, the biomarker or combination was considered to have good performance; and when AUC > 0.9, it was considered to have excellent performance. The following metabolites were screened as biomarkers for the asymptomatic stage of Huanglongbing: glutamate, L-Thr-L-Leu, uridine, guanosine, uracil, adenosine, thiamine, linolenic acid, maleic acid, malic acid, phenylalanine, pyroglutamic acid, citric acid, glucose, sucrose, arginine, proline, piperidine acid, and tryptophan. (5) Quantitatively detect the screened markers to obtain the absolute content of the marker in the leaves; (6) Use Lasso regression to construct a prediction model.

[0009] The objective of this invention is achieved through the following technical solution: A method for screening biomarkers for citrus Huanglongbing (HLB) based on metabolomics includes the following steps: S1. Collect and extract plant leaf samples: Accurately weigh citrus leaves from the healthy group and the asymptomatic group, add methanol or acetonitrile solution, homogenize and grind, then extract with ultrasonic assistance at 4-50℃. After extraction, centrifuge at 4℃, filter the supernatant through a 0.22μm filter membrane, and test it. At the same time, take equal amounts of all sample filtrates and combine them as QC (quality control) samples. S2. On-machine testing and acquisition of primary mass spectrometry data: Primary mass spectrometry data of the sample metabolome were acquired using the full scan function of positive and negative ion modes of a liquid chromatography-mass spectrometry (LC-MS) instrument, including: S2-1, Chromatographic conditions are as follows: Chromatographic column: Waters HSS T3 column or other reversed-phase column; Mobile phase: Positive ion mode, phase A (0.05%-0.2% formic acid aqueous solution), phase B (acetonitrile); Negative ion mode, phase A (10-20 mmol / L ammonium acetate + 10-20 mmol / L NH3·H2O aqueous solution), phase B (acetonitrile); Elution conditions: 0-20 min, phase B ratio 5%-95%; Flow rate: 0.1-0.5 mL / min; Column temperature: 40℃; Injection volume: 2 μL; S2-2, Mass spectrometry conditions are as follows: Positive ion mode mass spectrometry conditions: Drying gas temperature: 300℃; Sheath gas temperature: 375℃; Drying gas flow rate: 9L / min; Sheath gas flow rate: 11L / min; Fragmentation voltage: 125V; Nebulizer gas pressure: 45psi; MS scan range: m / z 80-1000; Capillary voltage: 4000V; Nozzle voltage: 500V; All the above parameters fluctuate within ±20%. S2-3, Negative Ion Mode Mass Spectrometry Conditions: Drying Gas Temperature: 300℃; Sheath Gas Temperature: 375℃; Drying Gas Flow Rate: 9L / min; Sheath Gas Flow Rate: 11L / min; Fragmentation Voltage: 125V; Nebulizing Gas Pressure: 45psi; MS Scan Range: m / z 80-1040; Capillary Voltage: 3000V; Nozzle Voltage: 1500V; S3. Data preprocessing after offline processing, including: Peak alignment, chromatogram extraction, and deconvolution operations were performed on the raw data to obtain a data matrix containing three sets of data: mass-to-charge ratio, sample name, and peak area. The data matrix is ​​processed, including calculating the coefficient of variation (CV) of the QC samples, deleting metabolites with CV values ​​greater than 0.3; deleting metabolites with missing values ​​greater than 50% in the QC samples and greater than 80% in the test samples; and finally filling in the remaining missing values. S4. Screening and identifying differentially expressed metabolites: Differential metabolites were identified using the secondary mass spectrometry function of a liquid chromatography-mass spectrometry (LC-MS) system, under the same chromatographic conditions as S2. The secondary mass spectra of the differential metabolites were matched, and the precursor ion, daughter ion, and retention time were examined for identification. The requirements were that the precursor ion error should be less than ±25 ppm, the daughter ion error should be less than ±50 ppm, and the retention time error should be less than ±0.5 min. S5. Screening biomarkers, calculating the minimum number of biomarker combinations, and constructing a prediction model, specifically including: The random forest algorithm was used to sort the contribution of all differentially expressed metabolites, and the number of biomarkers required for the minimum biomarker combination was calculated using the rfcv function. The area under the curve (AUC) for each differential metabolite and biomarker combination was calculated using ROC curves. A prediction model was established using Lasso regression, and the classification threshold was determined using the cutoff value of the ROC curve.

[0010] Furthermore, the collection and extraction of plant leaf samples described in S1 involves accurately weighing 10-500 mg of citrus leaves from the healthy group and the asymptomatic group, adding 1-50 mL of methanol or acetonitrile solution with a volume concentration of 10%-80%, homogenizing at 30-60 Hz for 90 s, followed by ultrasonic-assisted extraction at 4-50℃ for 10-50 min, centrifuging at 13000 rpm at 4℃ for 15 min after extraction, filtering the supernatant through a 0.22 μm filter membrane, and testing it on an instrument; at the same time, equal amounts of all sample filtrates are taken and combined as QC (quality control) samples.

[0011] Furthermore, the full-scan function of the positive and negative ion modes of the liquid chromatography-mass spectrometry (LC-MS) instrument described in S2 for acquiring primary mass spectrometry data of the sample metabolomics is either non-targeted metabolomics using a high-resolution LC-MS instrument or targeted metabolomics using a low-resolution LC-MS instrument.

[0012] Furthermore, in the chromatographic and mass spectrometric conditions described in S2, each numerical parameter fluctuates within a range of ±20%.

[0013] This invention also relates to a combination of biomarkers for citrus Huanglongbing (HLB), obtained according to the above-mentioned method for screening HLB biomarkers based on metabolomics. The biomarkers are one or more selected from L-Thr-L-Leu, glutamic acid, uridine, guanosine, uracil, thiamine, adenosine, phenylalanine, citric acid, maleic acid, malic acid, linolenic acid, sucrose, proline, glucose, and arginine. The biomarkers and their combination have high sensitivity and specificity, with an area under the ROC curve greater than 0.9, which is of great significance for the diagnosis of HLB.

[0014] Furthermore, the aforementioned citrus Huanglongbing is caused by Bacillus phloem. Candidatus Liberibacter Diseases caused by asiaticus.

[0015] Furthermore, the aforementioned markers are small organic molecules with a relative molecular mass of less than 1000.

[0016] In this invention, normal citrus samples were obtained from virus-free Satsuma mandarin seedlings isolated and cultivated in net greenhouses at the Guangxi Academy of Special Crops in early January 2024, while citrus Huanglongbing (HLB) samples were obtained from an abandoned Satsuma mandarin orchard in Tanxia Town, Lingchuan County, Guilin City at the end of November 2023. The citrus HLB sample collection method was as follows: leaves with HLB symptoms and asymptomatic leaves were collected from each plant. PCR was used to detect HLB-symptomatic leaves, and asymptomatic leaf samples from positive plants were included in subsequent metabolomics analysis.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention describes a method for screening biomarkers for citrus Huanglongbing based on metabolomics. This method uses high performance liquid chromatography-mass spectrometry to perform non-targeted metabolomics analysis to identify differential metabolites between the diseased group and the normal group. Further, it uses targeted metabolomics for absolute quantification, and employs a random forest algorithm to screen potential biomarkers and calculate the number of biomarker combinations. Finally, it uses Lasso regression to establish a predictive model.

[0018] 2. This invention discloses a method for screening biomarkers for citrus Huanglongbing (HLB) based on metabolomics. Fifty healthy plant samples and 50 asymptomatic HLB samples were analyzed. High-performance liquid chromatography-mass spectrometry (HPLC-MS / MS) was used to obtain fingerprint spectra of small molecule metabolites in both positive and negative ion modes. Through differential metabolite screening and machine learning analysis and feature selection, diagnostic biomarkers suitable for citrus HLB were obtained. Machine learning was used to model the monitoring data, resulting in a diagnostic model with advantages of accuracy and high sensitivity. This invention also provides a diagnostic model for citrus HLB and a method for constructing such a model. The model construction method is simple and exhibits high sensitivity and specificity for citrus HLB, providing effective technical support for early diagnosis and treatment of HLB.

[0019] 3. The present invention provides a combination of biomarkers for citrus Huanglongbing (HLB), which are screened by the method of the present invention. These biomarkers include one or more combinations of glutamic acid, L-Thr-L-Leu, uridine, guanosine, uracil, adenosine, thiamine, linolenic acid, maleic acid, malic acid, phenylalanine, pyroglutamic acid, citric acid, glucose, sucrose, arginine, proline, and piperidine acid. The metabolic biomarkers provided by the present invention can accurately detect HLB, with high sensitivity and specificity, and an area under the ROC curve greater than 0.9. Attached Figure Description

[0020] Figure 1 This is a sorting chart of Mean decrease accuracy and Meanecrease gini values ​​of candidate markers in embodiments of the present invention; Figure 2 This is the error CV graph of 10-fold cross-validation in an embodiment of the present invention; Figure 3 This is a box plot of the content of each marker in the embodiments of the present invention; Figure 4 These are ROC curves of a single marker and a combination of markers in embodiments of the present invention (A: ROC curve of a combination of markers; B: ROC curve of a single marker). Figure 5This is the ROC curve of the Lasso regression prediction results in this embodiment of the invention. Detailed Implementation

[0021] The present invention will now be described in detail through specific embodiments. The description of preferred embodiments is merely illustrative and is not intended to limit the scope, application, or use of the invention. Special note: Conditions not explicitly stated in the embodiments should be performed under conventional conditions or conditions recommended by the manufacturer, and unless otherwise specified, all raw materials used may be commercially available products. Example

[0022] A method for screening biomarkers for citrus Huanglongbing (HLB) based on metabolomics includes the following steps: 1. Sample collection and extraction: Leaf samples (n=50) from the HLB-Asylum positive group were collected at the end of November 2023 from a neglected orchard of Satsuma mandarin oranges in Tanxia Town, Lingchuan County, Guangxi Zhuang Autonomous Region. Leaves showing HLB symptoms were also collected from the same plants. All samples from the HLB-Asylum positive group were tested by PCR. Control group samples (CK) were collected in early December 2023 from virus-free and disease-free seedlings grown in a net greenhouse at the Guangxi Academy of Special Crops (n=50). All samples were pre-cooled at 4℃ after collection and then stored at -80℃. Take 100 mg of leaf sample, add 1 mL of 70% methanol solution, homogenize at 45 Hz for 90 s, then extract by sonication in an ice-water bath for 30 min, and finally centrifuge at 4 ℃ and 13000 rpm. Take the supernatant for testing. At the same time, take equal amounts of the filtrate of all samples and combine them as QC (quality control) samples. 2. Instruments and reagents: Agilent 6530 UPLC-QTof-MS (equipped with Dual AJS electrospray ionization source), deionized water, acetonitrile (mass spectrometry level), formic acid (Maclean, mass spectrometry level), ammonium acetate (Maclean, mass spectrometry level); 3. Experimental Procedure 3.1 UPLC chromatographic conditions: Chromatograph: Agilent 1290 ultra-high performance liquid chromatograph; Chromatographic column: Waters HSS T3 column (id 2.1×100mm, 2.1μm), injection volume 2 μL, flow rate 0.3mL / min, column temperature 40℃; positive ion mode: mobile phase A: water (0.1% formic acid), mobile phase B: acetonitrile; gradient ratio: 5%-95%, 0-20min; negative ion mode: mobile phase A: water (20mM ammonium acetate + 10mM NH3·H2O), mobile phase B: acetonitrile; gradient ratio: 5%-95%, 0-20min. 3.2 Ion source and mass spectrometry conditions: Mass spectrometer: Agilent 6530 ESI QToF electrospray ionization quadrupole time-of-flight mass spectrometer; Positive ion mode ion source conditions: Drying gas temperature: 300℃; Sheath gas temperature: 375℃; Drying gas flow rate: 9L / min; Sheath gas flow rate: 11L / min; Fragmentation voltage: 125V; Nebulizing gas pressure: 45psi; Capillary voltage: 4000V; Nozzle voltage: 500V; Cone voltage: 60V; Octodex voltage: 750V; Correction ions: m / z 118.0863, 922.0098, real-time correction enabled; Positive ion mode mass spectrometry conditions: Primary mass spectrometry acquisition rate: 1 spectral / s; MS scan range: m / z 80-1000; Secondary mass spectrometry uses auto MS / MS combined with target MS / MS for data acquisition; MS 1 Scan speed 1 spectral / s, MS 2 Scan speed: 3 spectral / s; m / z acquisition range: MS 1 80-1000, MS 2 50-1000; Collision energy: 10, 20, 40 eV; Negative ion mode ion source conditions: Dry gas temperature: 300℃; Sheath gas temperature: 375℃; Dry gas flow rate: 9L / min; Sheath gas flow rate: 11L / min; Fragmentation voltage: 125V; Nebulizing gas pressure: 45psi; Capillary voltage: 3000V; Nozzle voltage: 1500V; Cone voltage: 60V; Octodex voltage: 750V; Correction ions: m / z 112.9885, 1033.9881, real-time correction enabled; Negative ion mode mass spectrometry conditions: Primary mass spectrometry acquisition rate: 1 spectral / s; MS scan range: m / z 80-1040; Secondary mass spectrometry uses auto MS / MS combined with target MS / MS for data acquisition; MS 1 Scan speed 1 spectral / s, MS 2 Scan speed: 3 spectral / s; m / z acquisition range: MS 1 80-1040, MS 2 50-1040; Impact energy: 10, 20, 40 eV; 3.3 Data Preprocessing: UPLC-QTof-MS was used to analyze asymptomatic positive leaves (n=50) and healthy leaves (n=50). The data were processed by MSdial for peak alignment and feature extraction. Then, the data were preprocessed, including removing features with too many missing values, removing features with too high coefficient of variation (greater than 0.3) in QC samples, and filling the remaining missing values ​​with half of the minimum value. 3.4 Multivariate analysis, screening and identification of differentially metabolites: Multivariate analysis was performed on the CK and HLB-Asy groups using PCA and OPLS-DA. To avoid overfitting in the OPLS-DA analysis, 200 permutation tests were conducted. Differential metabolites were screened according to the following criteria: significance P < 0.05, fold difference FC > 2.0 or < 0.5, and variable projection importance VIP > 2.0. The screened metabolites were identified using the Metlin database in conjunction with standards, and retention time RT and MS were also examined. 1 MS 2 Three indicators, if all three indicators simultaneously meet the matching requirements, constitute a reliable marker; 3.5 Biomarker selection based on machine learning techniques: The Random Forest algorithm is used to score the weights of candidate markers. To ensure robustness, the model is run 10 times, and the final importance ranking of the selected candidate markers is represented by the average of their importance evaluation indices. Data is randomly sampled, with 70% used as the training set and 30% as the test set. The Random Forest model is trained on the training set, and the importance of features is evaluated based on MeanDecreaseGini or MeanDecreaseAccuracy from the Random Forest. To ensure data robustness, this is repeated 10 times, and the average of MeanDecreaseGini and MeanDecreaseAccuracy is calculated. K-fold cross-validation is performed using the RFCV function, repeated 10 times, and the average error is calculated from the 10 results. The CV value was repeated N times, and a line graph was plotted based on the average of the N values. The number of biomarkers corresponding to the minimum error CV was selected as the optimal number of biomarkers. Metabolic biomarkers were analyzed using ROC curves, and the performance was judged by the area under the curve (AUC): when 0.7 > AUC > 0.5, the biomarker was effective; when 0.9 > AUC > 0.7, the biomarker was good; and when AUC > 0.9, the performance was excellent. 3.6 Targeted Quantitative Validation: The screened biomarkers were absolutely quantified and box plots were plotted using the LCMS external standard method. The mass spectrometry conditions were the same as in 3.2. The biomarker content in the leaf samples was expressed as mg / g. 3.7 Lasso Regression for Predictive Model Construction: The levels of each biomarker were subjected to Lasso regression to construct a predictive model.

[0023] result: Based on the above method, endogenous metabolites that could potentially become diagnostic markers for Huanglongbing (HLB) were screened, as shown in Table 1:

[0024] The candidate markers listed in Table 1 were ranked using the random forest algorithm, and the results are shown below. Figure 2 The top seven candidate biomarkers were uridine, uracil, thiamine, glutamate, adenosine, L-Thr-L-Leu, and guanosine.

[0025] Figure 1 This is a ranking plot of the Mean Decrease Accuracy and Mean Decrease Gini values ​​of candidate biomarkers. The RFCV function was used to calculate the number of biomarkers, and 10 cross-validations were performed. The results are as follows: Figure 2 As shown, the error CV value is lowest when the number of markers n=6, therefore the number of marker combinations is 6.

[0026] Figure 2 This is a 10-fold cross-validation error CV plot. Considering the importance ranking of candidate biomarkers provided by random forest and the limitation on the number of biomarker combinations, uridine, uracil, thiamine, glutamate, adenosine, and L-Thr-L-Leu were ultimately selected as the biomarker combination. To verify the content of these biomarkers in leaves, LCMS external standard method was used to quantify the biomarkers individually, and the results were plotted as box plots. (See figure) Figure 3 .

[0027] Figure 3 The results are box plots of the content of each biomarker. The area under the ROC curve was used to evaluate the performance of the biomarker combination. Figure 4 As shown in Figure A, the area under the ROC curve for the marker combination is 1, indicating excellent classification performance. Figure 4 As shown in B, the area under the ROC curve for all markers is 1, indicating excellent classification performance.

[0028] Figure 4 These are ROC curves for single biomarkers and biomarker combinations. A: ROC curve for biomarker combinations; B: ROC curve for single biomarkers. A predictive model is built using Lasso regression, yielding the following regression equation: y=0.093+2.58 × L-Thr-L-Leu-0.909 × Glutamic Acid+0.022 × Uridine+2.743 × Uracil+6.226 × Adenosine+536.886 × Thiamine The predicted values ​​obtained by substituting the levels of each biomarker into the Lasso regression equation were used to plot an ROC curve, such as... Figure 5 As shown; Figure 5 This is the ROC curve of the Lasso regression prediction results, with an AUC value of 1.0, a Youden index of 1, and a cutoff value of 0.2252.

[0029] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, any improvements and changes made without departing from the inventive concept of the present invention are within the protection scope of the present invention.

Claims

1. A method for screening biomarkers for citrus Huanglongbing (HLB) based on metabolomics, characterized in that: Includes the following steps: S1. Collect and extract plant leaf samples: Accurately weigh citrus leaves from the healthy group and the asymptomatic group, add methanol or acetonitrile solution, homogenize and grind, then extract with ultrasonic assistance at 4-50℃. After extraction, centrifuge at 4℃, filter the supernatant through a 0.22μm filter membrane, and test it. At the same time, take equal amounts of all sample filtrates and combine them as QC samples. S2. On-machine testing and acquisition of primary mass spectrometry data: Primary mass spectrometry data of the sample metabolome were acquired using the full scan function of positive and negative ion modes of a liquid chromatography-mass spectrometry (LC-MS) instrument, including: S2-1, Chromatographic conditions are as follows: Chromatographic column: Waters HSS T3 column or other reversed-phase column; Mobile phase: Positive ion mode, phase A: 0.05%-0.2% formic acid aqueous solution, phase B: acetonitrile; Negative ion mode, phase A: 10-20 mmol / L ammonium acetate + 10-20 mmol / L NH3·H2O aqueous solution, phase B: acetonitrile; Elution conditions: 0-20 min, phase B ratio 5%-95%; Flow rate: 0.1-0.5 mL / min; Column temperature: 40℃; Injection volume: 2 μL; S2-2, Mass spectrometry conditions are as follows: Positive ion mode mass spectrometry conditions: Drying gas temperature: 300℃; Sheath gas temperature: 375℃; Drying gas flow rate: 9L / min; Sheath gas flow rate: 11L / min; Fragmentation voltage: 125V; Nebulizer gas pressure: 45psi; MS scan range: m / z 80-1000; Capillary voltage: 4000V; Nozzle voltage: 500V; All the above parameters fluctuate within ±20%. S2-3, Negative Ion Mode Mass Spectrometry Conditions: Drying Gas Temperature: 300℃; Sheath Gas Temperature: 375℃; Drying Gas Flow Rate: 9L / min; Sheath Gas Flow Rate: 11L / min; Fragmentation Voltage: 125V; Nebulizing Gas Pressure: 45psi; MS Scan Range: m / z 80-1040; Capillary Voltage: 3000V; Nozzle Voltage: 1500V; S3. Data preprocessing after offline processing, including: Peak alignment, chromatogram extraction, and deconvolution operations were performed on the raw data to obtain a data matrix containing three sets of data: mass-to-charge ratio, sample name, and peak area. The data matrix is ​​processed, including calculating the coefficient of variation (CV) of the QC samples, deleting metabolites with CV values ​​greater than 0.3; deleting metabolites with missing values ​​greater than 50% in the QC samples and greater than 80% in the test samples; and finally filling in the remaining missing values. S4. Screening and identifying differentially expressed metabolites: Differential metabolites were identified using the secondary mass spectrometry function of a liquid chromatography-mass spectrometry (LC-MS) system, under the same chromatographic conditions as S2. The secondary mass spectra of the differential metabolites were matched, and the precursor ion, daughter ion, and retention time were examined for identification. The requirements were that the precursor ion error should be less than ±25 ppm, the daughter ion error should be less than ±50 ppm, and the retention time error should be less than ±0.5 min. S5. Screening biomarkers, calculating the minimum number of biomarker combinations, and constructing a prediction model, specifically including: The random forest algorithm was used to sort the contribution of all differentially expressed metabolites, and the number of biomarkers required for the minimum biomarker combination was calculated using the rfcv function. The area under the curve (AUC) for each differential metabolite and biomarker combination was calculated using ROC curves. A prediction model was established using Lasso regression, and the classification threshold was determined using the cutoff value of the ROC curve.

2. The method for screening biomarkers for citrus Huanglongbing based on metabolomics according to claim 1, characterized in that: The collection and extraction of plant leaf samples described in S1 involves accurately weighing 10-500 mg of citrus leaves from the healthy group and the asymptomatic group, adding 1-50 mL of methanol or acetonitrile solution with a volume concentration of 10%-80%, homogenizing at 30-60 Hz for 90 seconds, followed by ultrasonic-assisted extraction at 4-50℃ for 10-50 minutes, centrifuging at 13000 rpm at 4℃ for 15 minutes after extraction, filtering the supernatant through a 0.22 μm filter membrane, and then testing it. At the same time, equal amounts of all sample filtrates are taken and combined as QC samples.

3. The method for screening biomarkers for citrus Huanglongbing based on metabolomics according to claim 1, characterized in that: The full-scan function of positive and negative ion mode of liquid chromatography-mass spectrometry (LC-MS) described in S2 is used to acquire primary mass spectrometry data of sample metabolomics. This is non-targeted metabolomics using high-resolution LC-MS or targeted metabolomics using low-resolution LC-MS.

4. The method for screening biomarkers for citrus Huanglongbing based on metabolomics according to claim 1, characterized in that: In the chromatographic and mass spectrometric conditions described in S2, each numerical parameter fluctuates within a range of ±20%.

5. A biomarker combination for citrus Huanglongbing (HLB), characterized in that: The method for screening biomarkers for citrus Huanglongbing based on metabolomics according to any one of claims 1-4 is used to obtain the biomarkers, which are one or more of L-Thr-L-Leu, glutamic acid, uridine, guanosine, uracil, thiamine, adenosine, phenylalanine, citric acid, maleic acid, malic acid, linolenic acid, sucrose, proline, glucose, and arginine. The biomarkers or combinations have high sensitivity and specificity, and the area under the ROC curve is greater than 0.

9.

6. The combination of biomarkers for citrus Huanglongbing according to claim 5, characterized in that: The aforementioned citrus Huanglongbing is caused by Bacillus phloem. Candidatus Liberibacter Diseases caused by asiaticus.

7. The combination of citrus Huanglongbing biomarkers according to claim 5, characterized in that: The aforementioned markers are small organic molecules with a relative molecular mass of less than 1000.

Images