Screening method of smoke disease differential metabolite biomarker combination based on serum metabolism fingerprint spectrogram and marker combination

Abstract

The invention provides a screening method of a smoke disease differential metabolite biomarker combination based on a serum metabolism fingerprint spectrogram and the marker combination, and relates to the field of molecular detection, the method comprises the following steps: 1, using a nanoparticle enhanced laser desorption ionization time-of-flight mass spectrometry technology to carry out serum metabolism fingerprint spectrogram imaging; 2, preprocessing the metabolic fingerprint spectrograms of the serum samples of the smog disease patient and the healthy control group to obtain m / z signals, and performing machine learning to obtain a model; and 3, constructing a smoke disease difference metabolite combination by using the model obtained in the step 2. According to the present invention, the nanoparticle enhanced laser desorption ionization time-of-flight mass spectrometry is adopted, the trace serum sample is used, and the high-flux, high-speed, high-sensitivity and high-repeatability identification of the metabolism biomarker combination of the serum metabolism fingerprint spectrum is achieved in the non-invasive manner;

Description

Technical Field

[0001] The present invention relates to the field of molecular detection, and in particular to a screening method for a combination of biomarkers of moyamoya disease differential metabolites based on serum metabolic fingerprint spectra and a marker combination. Background Art

[0002] Moyamoya disease (MMD) is a chronic progressive cerebrovascular disease characterized by bilateral occlusion or stenosis of the internal carotid arteries at the entrance of the circle of Willis. The main manifestations of the disease are intracranial ischemia and hemorrhage.

[0003] Digital subtraction angiography (DSA) remains the gold standard for the diagnosis of moyamoya disease, but it is invasive and some patients are allergic to contrast agents; magnetic resonance imaging / angiography (MRI / MRA) is also increasingly used to identify the disease, but it is time-consuming and has low spatial resolution. Both methods are expensive and require experienced experts to annotate the images, so they are not suitable for the universal diagnosis of moyamoya disease. Timely detection, diagnosis and surgical intervention of moyamoya disease can not only reduce the long-term risk of stroke, but also reverse hypoperfusion and reduce brain microstructural damage. Therefore, there is an urgent need to construct a simple and effective detection method to achieve timely diagnosis of moyamoya disease, which will be of great significance to patients with moyamoya disease.

[0004] In the design of detection methods, the selection of biomarkers and detection methods are the two main aspects that determine the diagnostic performance. For biomarker selection, molecular biomarkers at the genomic / proteomic / metabolism level in biological fluids are effective indicators for identifying changes in biological systems associated with pathological processes, and the acquisition of biological fluids is minimally invasive. For the detection of brain diseases such as moyamoya disease, blood testing has shown excellent safety, simplicity and cost-effectiveness due to its non-invasiveness, convenience and wide applicability. Genomic and proteomic biomarkers are located upstream of metabolic pathways and can usually reflect indirect information of biological systems. In contrast, metabolic biomarkers, as the end products of the pathway, can directly characterize disease phenotypes and progression. At the same time, since small molecule metabolites can cross the blood-brain barrier and can reflect the status of brain diseases more timely, metabolic biomarkers have more potential as biomarkers for diagnosing brain diseases than other omics. It is worth noting that considering the complexity of biological fluids, such as millions of molecules and biomarkers with a wide dynamic range spanning orders of magnitude, it is crucial to construct a set of biomarker combinations for accurate diagnosis. In this context, machine learning methods, capable of encoding data and performing feature selection, are ideal for extracting key information from complex metabolic datasets.

[0005] Among the detection methods, mass spectrometry is a basic tool for selecting metabolic biomarkers. Unlike traditional spectroscopic methods (e.g., nuclear magnetic resonance, NMR) and biochemical methods, mass spectrometry can detect the precise mass-to-charge ratio of molecules and has the advantage of high resolution (ppm). To date, most methods using gas chromatography / liquid chromatography (GC / LC) to detect biological samples require processing procedures for metabolite enrichment and purification to address molecular abundance and sample complexity. However, laser desorption / ionization (LDI MS) relies on nanostructures on the chip surface to selectively identify metabolites, which can achieve high-performance detection and obtain metabolic fingerprints. At present, metabolic fingerprints based on LDI MS have covered a variety of diseases (stroke, cerebral aneurysm, breast cancer, gastric cancer, etc.). However, there is no report on the metabolic fingerprint of moyamoya disease serum. Therefore, the development of the metabolic fingerprint of moyamoya disease serum will provide a high-performance diagnostic method for moyamoya disease based on biological metabolites, promote the development of serum metabolic analysis of moyamoya disease, and lay the foundation for non-invasive blood testing of moyamoya disease in the future.

[0006] Therefore, technicians in this field are committed to developing a combination of bio-metabolite markers that can diagnose moyamoya disease quickly, accurately, conveniently and at a low cost. Summary of the invention

[0007] In view of the above-mentioned defects of the prior art, the technical problem to be solved by the present invention is to provide a screening method for a combination of biological metabolic markers that can predict moyamoya disease quickly, accurately, conveniently and at a low cost.

[0008] To achieve the above object, the present invention provides a method for screening a combination of differential metabolite biomarkers of moyamoya disease based on serum metabolic fingerprints, characterized in that the method comprises the following steps:

[0009] Step 1: Use nanoparticle enhanced laser desorption ionization time-of-flight mass spectrometry to perform serum metabolic fingerprint imaging;

[0010] Step 2: Preprocess the metabolic fingerprint spectra of serum samples from patients with moyamoya disease and healthy controls to obtain the mass-to-nuclear ratio, i.e., m / z signal, and perform machine learning to obtain a model;

[0011] Step 3: Use the model obtained in step 2 to construct a combination of differential metabolites of moyamoya disease.

[0012] In a preferred embodiment of the present invention, step 1 comprises:

[0013] Step 1.1. Preparation of instruments and reagents: matrix-assisted laser desorption ionization time-of-flight mass spectrometry, serum sample, deionized water, matrix;

[0014] Step 1.2, dilute the serum sample 10 times with deionized water;

[0015] Step 1.3, preparing a matrix solution;

[0016] Step 1.4, prepare the sample on the mass spectrometry target plate, spot 1 μL of each diluted serum sample and dry at room temperature;

[0017] Step 1.5, prepare the matrix on the mass spectrometry target plate, spot 1 μL of each matrix solution, and dry at room temperature;

[0018] Step 1.6, collecting serum metabolic fingerprints in a matrix-assisted laser desorption ionization Fourier transform ion cyclotron resonance mass spectrometer;

[0019] Step 1.7, obtain the serum metabolic fingerprint spectrum.

[0020] In another preferred embodiment of the present invention, the matrix in step 1.1 is inorganic nanoparticles, and step 1.3 specifically comprises preparing the inorganic nanoparticles into a 1 mg / mL matrix solution with deionized water.

[0021] In another preferred embodiment of the present invention, the step 2 comprises:

[0022] Step 2.1, Preparation of instruments and reagents: python (3.9.0), Orange (3.25.0);

[0023] Step 2.2, performing metabolic fingerprint imaging on a large number of serum samples through step 1;

[0024] Step 2.3: Preprocess the serum metabolic fingerprints of all samples on Python to obtain m / z signals;

[0025] Step 2.4, divide all serum samples into training set and test set;

[0026] Step 2.5: Use the algorithm in Python and Orange to train on the training set, obtain the prediction performance of the model on the training set, and use the test set to verify the model performance.

[0027] In another preferred embodiment of the present invention, the serum samples include moyamoya disease serum samples and healthy control serum samples, and the training set and the test set include moyamoya disease serum sample data and healthy control serum sample data, respectively.

[0028] In another preferred embodiment of the present invention, the preprocessing in step 2.3 includes spectrum line smoothing, baseline correction and spectrum peak matching.

[0029] In another preferred embodiment of the present invention, the algorithm includes one of a neural network algorithm, a k-nearest neighbor algorithm, an adaptive boosting algorithm, a ridge regression algorithm and a naive Bayes algorithm.

[0030] In another preferred embodiment of the present invention, the step 3 comprises:

[0031] Step 3.1, preparation of instruments and reagents: python (3.9.0), SPSS (24.0);

[0032] Step 3.2, using SPSS to perform a t-test on the intensity of the m / z signal obtained in step 2 in samples of moyamoya disease patients and healthy controls, and screen out signals with differences;

[0033] Step 3.3, using SPSS to perform fold change statistics on the intensity of the m / z signal obtained in step 2 in the moyamoya disease patients and healthy control samples, and screen out signals with differences;

[0034] Step 3.4, using a global interpretative algorithm in Python to rank the contribution of the m / z signals obtained in step 2 in the model obtained in step 2, and obtain the top 20% of the signals;

[0035] Step 3.5, intersecting the signals obtained in steps 3.2, 3.3, and 3.4 to obtain potential metabolic biomarkers;

[0036] Step 3.6: Use the newly collected serum sample data to perform performance verification on the potential metabolic biomarkers in step 3.5 and determine the metabolic biomarker combination.

[0037] In another preferred embodiment of the present invention, the signal of difference in step 3.2 is specifically P<0.05, and the signal of difference in step 3.3 is specifically |Log2(FC)|>1.8.

[0038] The present invention also provides a metabolite biomarker combination screened by the above method, wherein the biomarker combination comprises 6 biomarkers, and the mass-to-nuclear ratios, i.e., m / z values ​​of all biomarkers are 145.967, 147.975, 151.019, 166.991, 208.062, and 225.046, respectively.

[0039] Technical Effects

[0040] 1. Traditional moyamoya disease requires digital subtraction angiography (DSA) or magnetic resonance imaging, which are time-consuming, expensive, and require experienced doctors compared to blood tests. The present invention is based on LDI-MS technology, uses trace serum samples, and conducts machine learning on serum metabolic fingerprint spectra of moyamoya disease patients and healthy controls to construct an accurate characterization model (screening of moyamoya disease metabolic markers), screens out moyamoya disease blood biomarkers, and constructs a moyamoya disease biomarker screening panel, creating conditions for the screening of moyamoya disease blood metabolic biomarkers, and solving the problems of invasiveness, high cost, slow analysis speed, and physician dependence in traditional moyamoya disease detection technology.

[0041] 2. Traditional metabolic analysis techniques (such as liquid / gas chromatography-mass spectrometry, nuclear magnetic resonance) require a strict pretreatment process (~0.5-3 hours, including complex pretreatment operations such as isotope treatment, derivatization, protein precipitation and chromatographic separation) and a large sample size (~10-500 μL / sample) to overcome the complexity of samples in biological fluids and the low abundance of metabolites. The present invention detects the serum of patients with moyamoya disease by nanoparticle-enhanced laser desorption ionization time-of-flight mass spectrometry to obtain a metabolic fingerprint spectrum of moyamoya disease serum. Nanoparticle-enhanced laser desorption ionization time-of-flight mass spectrometry: the measured polymer sample or biological macromolecule is mixed with a small molecule matrix compound solution in a certain proportion, so that the measured sample is dispersed in the matrix in a single molecule state. After drying, a pulsed laser is irradiated onto the matrix and the sample. Since the sample has no absorption at this laser wavelength, and the matrix has strong absorption, the matrix absorbs energy and excites desorption. Due to the high energy, the excitation process causes the matrix to explode, so that the sample chain is expelled and ionized. This process does not cause the polymer chain to break. Usually Only molecular ions and polymers of molecular ions are generated. The ionized molecules are accelerated under the action of a strong electric field, gain kinetic energy (KE), and then fly along the field-free area of ​​the flight tube. Different substances have different molecular weights and different flight times, so different substances are detected. Nanoparticle-enhanced laser desorption ionization time-of-flight mass spectrometry is used, and only trace serum samples (100nL) are used to achieve high-throughput (384 samples / chip) serum metabolic fingerprint spectra in a non-invasive manner. High detection speed (about 30s / sample), high sensitivity (as low as 2.9pmol) and high repeatability (RSD<15%) detection, effectively solving the above-mentioned traditional technologies' defects such as large sample size and complex pretreatment.

[0042] In summary, the present invention adopts nanoparticle enhanced laser desorption ionization time-of-flight mass spectrometry, uses only trace serum samples (100nL), and achieves high-throughput (384 samples / chip), high detection speed (about 30s / sample), high sensitivity (as low as 2.9pmol) and high repeatability (RSD<15%) detection of serum metabolic fingerprint spectra in a non-invasive manner. It effectively solves the defects of the above-mentioned traditional technology such as large sample size and complex pretreatment. In addition, the present invention constructs a model for accurate characterization (screening of moyamoya disease metabolic markers) by machine learning of serum metabolic fingerprint spectra of moyamoya disease patients and healthy controls, and identifies 6 metabolic biomarker combinations, which are expected to be applied to the detection of moyamoya disease in clinical practice. It solves the problems of low throughput, complex operation, high cost, slow analysis speed, and physician dependence in traditional moyamoya disease detection technology.

[0043] This technology is very simple to implement, uses a small sample size, has a fast detection speed, high throughput, and good repeatability, and is expected to be applied on a large scale in different laboratories and industrial production. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a representative metabolic fingerprint spectrum of serum of a moyamoya disease patient and a healthy control in a preferred embodiment of the present invention;

[0045] Figure 2 It is a heat map corresponding to 134 m / z signals after preprocessing of serum metabolic fingerprint spectra of 288 moyamoya disease patients and healthy controls in a preferred embodiment of the present invention;

[0046] Figure 3 It is a working characteristic curve of a preferred embodiment of the present invention for predicting moyamoya disease using an algorithm in a training set;

[0047] Figure 4 It is a working characteristic curve of a preferred embodiment of the present invention for predicting moyamoya disease using an algorithm in a test set;

[0048] Figure 5 It is a schematic diagram of the performance of a combination of biomarkers and a single biomarker in predicting moyamoya disease according to a preferred embodiment of the present invention;

[0049] Figure 6 It is a schematic diagram of the performance of using a biomarker combination to predict moyamoya disease in an external validation set according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following describes several preferred embodiments of the present invention with reference to the drawings in the specification, so that the technical content is clearer and easier to understand. The present invention can be embodied in many different forms of embodiments, and the protection scope of the present invention is not limited to the embodiments mentioned in the text.

[0051] The present invention provides a method for screening a combination of differential metabolite biomarkers of moyamoya disease based on serum metabolic fingerprint spectra, characterized in that the method comprises the following steps:

[0052] Step 1: Use nanoparticle enhanced laser desorption ionization time-of-flight mass spectrometry to perform serum metabolic fingerprint imaging;

[0053] Step 2: Preprocess the metabolic fingerprint spectra of serum samples of moyamoya disease patients and healthy controls to obtain m / z signals, and perform machine learning to obtain a model;

[0054] Step 3: Use the model obtained in step 2 to construct a combination of differential metabolites of moyamoya disease.

[0055] In a preferred embodiment of the present invention, step 1 comprises:

[0056] Step 1.1. Preparation of instruments and reagents: matrix-assisted laser desorption ionization time-of-flight mass spectrometry, serum sample, deionized water, matrix;

[0057] Step 1.2, dilute the serum sample 10 times with deionized water;

[0058] Step 1.3, preparing a matrix solution;

[0059] Step 1.4, prepare the sample on the mass spectrometry target plate, spot 1 μL of each diluted serum sample and dry at room temperature;

[0060] Step 1.5, prepare the matrix on the mass spectrometry target plate, spot 1 μL of each matrix solution, and dry at room temperature;

[0061] Step 1.6, collecting serum metabolic fingerprints in a matrix-assisted laser desorption ionization Fourier transform ion cyclotron resonance mass spectrometer;

[0062] Step 1.7, obtain the serum metabolic fingerprint spectrum.

[0063] In another preferred embodiment of the present invention, the matrix in step 1.1 is inorganic nanoparticles.

[0064] In another preferred embodiment of the present invention, the step 1.3 is specifically to prepare the inorganic nanoparticles into a 1 mg / mL matrix solution with deionized water.

[0065] In another preferred embodiment of the present invention, the step 2 comprises:

[0066] Step 2.1, Preparation of instruments and reagents: python (3.9.0), Orange (3.25.0);

[0067] Step 2.2, performing metabolic fingerprint imaging on a large number of serum samples through step 1;

[0068] Step 2.3: Preprocess the serum metabolic fingerprints of all samples on Python to obtain m / z signals;

[0069] Step 2.4, divide all serum samples into training set and test set;

[0070] Step 2.5: Use the algorithm in Python and Orange to train on the training set, obtain the prediction performance of the model on the training set, and use the test set to verify the model performance.

[0071] In another preferred embodiment of the present invention, the serum samples include moyamoya disease serum samples and healthy control serum samples, and the training set and the test set include moyamoya disease serum sample data and healthy control serum sample data, respectively.

[0072] In another preferred embodiment of the present invention, the preprocessing in step 2.3 includes spectrum line smoothing, baseline correction and spectrum peak matching.

[0073] In another preferred embodiment of the present invention, the algorithm includes one of a neural network algorithm, a k-nearest neighbor algorithm, an adaptive boosting algorithm, a ridge regression algorithm and a naive Bayes algorithm.

[0074] In another preferred embodiment of the present invention, the step 3 comprises:

[0075] Step 3.1, preparation of instruments and reagents: python (3.9.0), SPSS (24.0);

[0076] Step 3.2, using SPSS to perform a t-test on the intensity of the m / z signal obtained in step 2 in samples of moyamoya disease patients and healthy controls, and screen out signals with differences;

[0077] Step 3.3, using SPSS to perform fold change statistics on the intensity of the m / z signal obtained in step 2 in the moyamoya disease patients and healthy control samples, and screen out signals with differences;

[0078] Step 3.4, using a global interpretative algorithm in Python to rank the contribution of the m / z signals obtained in step 2 in the model obtained in step 2, and obtain the top 20% of the signals;

[0079] Step 3.5, intersect the signals obtained in steps 3.2, 3.3, and 3.4 to obtain potential metabolic biomarkers (markers corresponding to mass to charge ratio (m / z): 145.967, 147.975, 151.019, 166.991, 208.062, 225.046);

[0080] Step 3.6: Use the newly collected serum sample data to perform performance verification on the potential metabolic biomarkers in step 3.5 and determine the metabolic biomarker combination.

[0081] In another preferred embodiment of the present invention, the signal of difference in step 3.2 is specifically P<0.05, and the signal of difference in step 3.3 is specifically |Log2(FC)|>1.8.

[0082] Example 1: Using nanoparticle enhanced laser desorption ionization time-of-flight mass spectrometry to perform serum metabolic fingerprint imaging:

[0083] Step 1: Preparation of instruments and reagents: matrix-assisted laser desorption ionization time-of-flight mass spectrometry, serum samples, deionized water, matrix (inorganic nanoparticles);

[0084] Step 2: Dilute the serum sample 10 times with deionized water.

[0085] Step 3: Prepare the inorganic nanoparticles into a 1 mg / mL matrix solution with deionized water.

[0086] Step 4: Prepare the sample on the mass spectrometry target plate, spot 1 μL of each diluted serum sample and dry at room temperature;

[0087] Step 5: Prepare the matrix on the mass spectrometry target plate, spot 1 μL of each matrix solution and dry at room temperature;

[0088] Step 6: Serum metabolic fingerprinting in matrix-assisted laser desorption ionization Fourier transform ion cyclotron resonance mass spectrometry;

[0089] Step 7: Obtain serum metabolic fingerprint spectrum for subsequent analysis.

[0090] Example 2: Machine learning of metabolic fingerprint spectra of serum samples from patients with moyamoya disease and healthy controls

[0091] Step 1: Preparation of instruments and reagents: python (3.9.0), Orange (3.25.0);

[0092] Step 2: Based on step 1, metabolic fingerprint imaging was performed on 144 serum samples of moyamoya disease and 144 serum samples of healthy controls ( Figure 1 );

[0093] Step 3: The serum metabolic fingerprints of 288 samples were preprocessed on Python (3.9), including spectral smoothing, baseline correction and peak matching, and 134 m / z signals were obtained ( Figure 2 );

[0094] Step 4: Divide the 288 serum samples into a training set (230 samples, including 115 moyamoya disease patients and 115 healthy controls) and a test set (58 samples, including 29 moyamoya disease patients and 29 healthy controls);

[0095] Step 5: Use algorithms (neural network algorithm, k-nearest neighbor algorithm, adaptive boosting algorithm, ridge regression algorithm and naive Bayes algorithm) on Python (2022.2.3) and Orange (3.25.0) to train the model on the training set and get the performance of the model on the training set ( Figure 3 ,Table 1), the neural network algorithm performed best;

[0096] Step 6: Use the model trained in step 5 to make predictions on the test set on Python (2022.2.3) and Orange (3.25.0) to get the performance of the five algorithms on the test set ( Figure 4 ,Table 1), the neural network algorithm performed best;

[0097] Example 3: Construction of a metabolic biomarker combination for diagnosis of moyamoya disease:

[0098] Step 1: Preparation of instruments and reagents: python (3.9), SPSS (24.0);

[0099] Step 2: SPSS (24.0) was used to perform a t-test on the intensities of the 134 m / z signals obtained in step 2 in moyamoya disease patients and healthy control samples, and 100 signals with differences (p<0.05) were screened.

[0100] Step 3: SPSS (24.0) was used to perform fold change statistics on the intensities of the 134 m / z signals obtained in step 2 in moyamoya disease patients and healthy control samples, and 27 signals with differences (|Log2(FC)|>1.8) were screened.

[0101] Step 4: Use the global interpretative algorithm (Integrated Gradient) in python (3.9) to sort the contribution of the 134 m / z signals obtained in step 2 in the neural network with the best model effect, and get the top 20%, a total of 18 signals.

[0102] Step 5: Intersect the signals obtained in steps 2, 3, and 4 to obtain 6 potential metabolic biomarkers;

[0103] Step 6: Compared with single markers, the combination of 6 markers showed the best performance in distinguishing patients with moyamoya disease specifically ( Figure 5 ); Preferably, the mass-to-nuclear ratios, i.e., m / z values, of the 6 biomarkers screened out are 145.967, 147.975, 151.019, 166.991, 208.062, and 225.046, respectively.

[0104] Step 7: Perform metabolic fingerprint imaging on an external validation set from multiple centers (including 49 serum samples of moyamoya disease and 37 healthy control samples) through step 1;

[0105] Step 8: The best neural network algorithm in step 2 is obtained by combining the six markers obtained in step 7 of step 3 with the data of patients with moyamoya disease in the external validation set to distinguish between moyamoya disease and the control group ( Figure 6 ),.

[0106] The preferred specific embodiments of the present invention are described in detail above. It should be understood that ordinary technicians in the field can make many modifications and changes based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by technicians in the technical field based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for screening a combination of differential metabolite biomarkers of moyamoya disease based on serum metabolic fingerprints, characterized in that: The method comprises the following steps: Step 1: Use nanoparticle enhanced laser desorption ionization time-of-flight mass spectrometry to perform serum metabolic fingerprint imaging; Step 2: Preprocess the metabolic fingerprint spectra of serum samples from patients with moyamoya disease and healthy controls to obtain the mass-to-nuclear ratio, i.e., m / z signal, and perform machine learning to obtain a model; Step 3: Use the model obtained in step 2 to construct a combination of differential metabolites of moyamoya disease.

2. The method according to claim 1, characterized in that The step 1 comprises: Step 1.

1. Preparation of instruments and reagents: matrix-assisted laser desorption ionization time-of-flight mass spectrometry, serum sample, deionized water, matrix; Step 1.2, dilute the serum sample 10 times with deionized water; Step 1.3, preparing a matrix solution; Step 1.4, prepare the sample on the mass spectrometry target plate, spot 1 μL of each diluted serum sample and dry at room temperature; Step 1.5, prepare the matrix on the mass spectrometry target plate, spot 1 μL of each matrix solution, and dry at room temperature; Step 1.6, collecting serum metabolic fingerprints in a matrix-assisted laser desorption ionization Fourier transform ion cyclotron resonance mass spectrometer; Step 1.7, obtain the serum metabolic fingerprint spectrum.

3. The method according to claim 2, characterized in that The matrix in step 1.1 is inorganic nanoparticles, and step 1.3 specifically comprises preparing the inorganic nanoparticles into a 1 mg / mL matrix solution with deionized water.

4. The method according to claim 1, characterized in that The step 2 comprises: Step 2.1, Preparation of instruments and reagents: python, Orange; Step 2.2, performing metabolic fingerprint imaging on a large number of serum samples through step 1; Step 2.3: Preprocess the serum metabolic fingerprints of all samples on Python to obtain m / z signals; Step 2.4, divide all serum samples into training set and test set; Step 2.5: Use the algorithm in Python and Orange to train on the training set, obtain the prediction performance of the model on the training set, and use the test set to verify the model performance.

5. The method according to claim 4, characterized in that The serum samples include serum samples of moyamoya disease and serum samples of healthy controls, and the training set and the test set include serum sample data of moyamoya disease and serum sample data of healthy controls, respectively.

6. The method according to claim 4, characterized in that The preprocessing in step 2.3 includes spectrum smoothing, baseline correction and spectrum peak matching.

7. The method according to claim 4, characterized in that The algorithm includes one of a neural network algorithm, a k-nearest neighbor algorithm, an adaptive boosting algorithm, a ridge regression algorithm and a naive Bayes algorithm.

8. The method according to claim 1, characterized in that The step 3 comprises: Step 3.1, Preparation of instruments and reagents: python, SPSS; Step 3.2, using SPSS to perform a t-test on the intensity of the m / z signal obtained in step 2 in samples of moyamoya disease patients and healthy controls, and screen out signals with differences; Step 3.3, using SPSS to perform fold change statistics on the intensity of the m / z signal obtained in step 2 in the moyamoya disease patients and healthy control samples, and screen out signals with differences; Step 3.4, using a global interpretative algorithm in Python to rank the contribution of the m / z signals obtained in step 2 in the model obtained in step 2, and obtain the top 20% of the signals; Step 3.5, intersecting the signals obtained in steps 3.2, 3.3, and 3.4 to obtain potential metabolic biomarkers; Step 3.6: Use the newly collected serum sample data to perform performance verification on the potential metabolic biomarkers in step 3.5 and determine the metabolic biomarker combination.

9. The method according to claim 8, characterized in that The signal of the presence of difference in step 3.2 is specifically P<0.05, and the signal of the presence of difference in step 3.3 is specifically |Log2(FC)|>1.

8.

10. The metabolite biomarker combination screened out using the screening method for the metabolite biomarker combination according to any one of claims 1 to 9, characterized in that: The biomarker combination includes 6 biomarkers, and the mass-to-nuclear ratios, i.e., m / z values, of all biomarkers are 145.967, 147.975, 151.019, 166.991, 208.062, and 225.046, respectively.

Images