Method for screening metabolic biomarkers based on saliva metabolic fingerprint and application thereof

Abstract

The application discloses a method for screening metabolic biomarkers based on saliva metabolic fingerprint and application, and relates to the technical field of biological sample metabolic analysis. The method comprises the following steps: step 1, inorganic nanoparticles are used to enhance laser desorption ionization time-of-flight mass spectrometry technology, saliva samples of Parkinson's disease patients and healthy volunteers are subjected to metabolic detection, and saliva metabolic fingerprints are obtained; step 2, the saliva metabolic fingerprints obtained in step 1 are subjected to a machine learning algorithm, saliva samples are subjected to a variety of model training, a variety of models are evaluated through a cross-validation method, and a final prediction model is screened according to a verification result; step 3, saliva samples of Parkinson's disease patients and healthy volunteers are predicted according to the prediction model of step 2, a prediction result is obtained, and metabolic biomarkers in the saliva samples are screened according to the prediction result. The application has high screening efficiency, the result is accurate and reliable, and the screened metabolic biomarkers can be used for prediction of Parkinson's disease.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biological sample metabolic analysis, and particularly relates to a method for screening metabolic biomarkers based on saliva metabolic fingerprint spectrum and application. BACKGROUND

[0002] Parkinson's disease is a common degenerative disease, with 6.1 million patients worldwide in 2016, and the disease course can last for decades. Global disease burden studies show that the incidence and prevalence of Parkinson's disease are rapidly increasing, and it is estimated that the number of patients worldwide will reach 100 million by 2030. As the development of Parkinson's disease gradually intensifies, effective management requires timely and accurate diagnosis before seeking medical advice. There is currently no diagnostic test for confirmed Parkinson's disease, so correct judgment of whether a patient has Parkinson's disease depends on one or more combinations of four main symptoms, such as rhythmic tremor, motor slowing, rigidity, or postural instability. However, since these clinical symptoms can overlap with other neurodegenerative diseases, diagnosis remains challenging. For example, the dopamine transporter single-photon emission computed tomography imaging technique approved by the U.S. Food and Drug Administration in 2011 also cannot distinguish between Parkinson's disease and other Parkinson's syndromes (such as progressive supranuclear palsy and multiple system atrophy). Therefore, it is of great significance to construct a simple and effective method for the management of Parkinson's disease, which has great social and economic benefits.

[0003] Detecting the earliest stage of the disease before the appearance of clinical symptoms of Parkinson's disease can enhance the effectiveness of drug treatment, thereby slowing the progression of the disease. The choice of biomarkers and detection methods is crucial for the design of diagnostic detection methods. Molecular biomarkers at the genomic / proteomic / metabolic level in biological fluids are effective indicators of changes in biological systems related to pathological processes, and the acquisition of biological fluids has less invasiveness. Genomic and proteomic biomarkers are located upstream of metabolic pathways, and usually reflect indirect information about the biological system. In contrast, metabolic biomarkers, as end products of pathways, can directly characterize disease phenotypes and progression.

[0004] Mass spectrometry has become a major tool for screening metabolic biomarkers. However, most methods for detecting biological samples using gas chromatography / liquid chromatography require enrichment and strict purification pretreatment procedures, including isotopic treatment, derivatization, protein precipitation, and complex pretreatment operations such as chromatographic separation, which generally require 0.5-3 hours to address the problem of molecular abundance and sample complexity; meanwhile, a large sample volume is required for detection, generally 10-500 μL / sample. These processing procedures require a large amount of time and resources, and can result in the loss or contamination of metabolites. In recent years, the development of matrix-assisted laser desorption / ionization mass spectrometry has made it possible to directly detect metabolites in ultra-low volume biological fluids without enrichment or purification. This method has the advantage of high-throughput screening and can be used for accurate non-targeted metabolic analysis, thereby allowing the identification of specific disease biomarkers to reflect the true condition of patients.

[0005] Saliva biomarkers have attracted attention for the diagnosis of Parkinson's disease as a non-invasive and cost-effective method. Recent studies have shown that certain metabolites in saliva may be related to the development of Parkinson's disease, however, there is no report on methods for screening metabolic biomarkers related to Parkinson's disease.

[0006] Therefore, those skilled in the art are committed to developing a matrix-assisted laser desorption / ionization mass spectrometry to obtain metabolic fingerprints in saliva, which can be an important tool for screening saliva biomarkers for the diagnosis of Parkinson's disease. SUMMARY

[0007] In view of the above-mentioned defects of the prior art, the technical problem to be solved by the present application is to provide a matrix-assisted laser desorption / ionization mass spectrometry to obtain metabolic fingerprints in saliva for the diagnosis of Parkinson's disease.

[0008] To achieve the above-mentioned object, the present application provides a method for screening metabolic biomarkers based on saliva metabolic fingerprinting, comprising the following steps:

[0009] Step 1, using inorganic nanoparticle-enhanced laser desorption ionization time-of-flight mass spectrometry, metabolic detection of saliva samples of Parkinson's patients and healthy volunteers is performed to obtain saliva metabolic fingerprinting;

[0010] Step 2, the saliva metabolic fingerprinting obtained in step 1 is subjected to a variety of model training using a machine learning algorithm, and the variety of models are evaluated by a cross-validation method, and the final prediction model is selected according to the validation results;

[0011] Step 3, predicting the saliva samples of the Parkinson's disease patients and the healthy volunteers according to the prediction model of Step 2 to obtain a prediction result, and screening the metabolic biomarkers in the saliva samples according to the prediction result.

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

[0013] Step 1, 1: diluting the saliva samples of the Parkinson's disease patients and the healthy volunteers with deionized water;

[0014] Step 1, 2: preparing inorganic nanoparticles into a matrix solution with deionized water;

[0015] Step 1, 3: spotting the diluted saliva samples on a mass spectrometry target plate and naturally drying at room temperature to complete the preparation of the saliva samples on the mass spectrometry target plate;

[0016] Step 1, 4: spotting the matrix solution on the mass spectrometry target plate and naturally drying at room temperature to complete the preparation of the matrix on the mass spectrometry target plate;

[0017] Step 1, 5: collecting the saliva metabolic fingerprints of the Parkinson's disease patients and the healthy volunteers by using a laser desorption ionization time-of-flight mass spectrometry technology to obtain a saliva metabolic fingerprint spectrum.

[0018] Further, in the step 1, 1, the saliva samples of the Parkinson's disease patients and the healthy volunteers are diluted 10 times with deionized water; and in the step 1, 3, 1 μL of the diluted saliva samples is spotted on the mass spectrometry target plate.

[0019] Further, in the step 1, 2, the inorganic nanoparticles are prepared into a 1 mg / mL matrix solution with deionized water; and in the step 1, 4, 1 μL of the 1 mg / mL matrix solution is spotted on the mass spectrometry target plate.

[0020] In another preferred embodiment of the present application, the step 2 comprises:

[0021] Step 2, 1: preprocessing the saliva metabolic fingerprint spectra of the Parkinson's disease patients and the healthy volunteers on MATLAB R2020a to collect the m / z signals of each saliva sample;

[0022] Step 2, 2: dividing the collected saliva samples of the Parkinson's disease patients and the healthy volunteers into corresponding training sets and test sets;

[0023] Step 2, 3: using a machine learning algorithm on the training sets to train multiple models for the saliva samples and screen a final prediction model.

[0024] Further, in the step 2, 1, the preprocessing comprises spectral line smoothing, baseline correction and spectral peak matching.

[0025] Further, the machine learning algorithm in steps 2 and 3 includes lasso regression, extreme gradient boosting, support vector machine, random forest, adaptive boosting and deep learning, a total of 6 machine learning algorithms.

[0026] In another preferred embodiment of the present application, step 3 comprises:

[0027] Step 3, 1: Fold-change analysis is performed on the saliva fingerprint of healthy volunteers and Parkinson's patients, and the differences between the two groups of metabolic data are compared to screen metabolites with a Fold-change value greater than 1.20 or less than 0.83;

[0028] Step 3, 2: Student's t-test analysis is performed on the saliva fingerprint of healthy volunteers and Parkinson's patients, and the significant differences between the metabolite levels of the Parkinson's group and the healthy control group are compared to screen metabolites with P less than 0.05;

[0029] Step 3, 3: Integrated Gradient analysis is performed on the saliva fingerprint of healthy volunteers and Parkinson's patients, which is used to explain the prediction results of the prediction model, calculate the contribution degree of each metabolite to the prediction results of the model, and screen metabolites with a contribution degree greater than 0.2;

[0030] Step 3, 4: Cross-matching is performed on the metabolites screened by Fold-change, Student's t-test and Integrated Gradient, and finally the differential metabolites are screened and the metabolic biomarkers are identified.

[0031] Further, the metabolic biomarkers C4H7N3O, C4H9N3O2, C6H 12 O6, C 21 H 30 O3, C 24 H 38 O3, C 23 H 42 NO4, C 10 H 15 N5O 10 P2.

[0032] The present application also provides a method for screening metabolic biomarkers based on the above saliva metabolic fingerprint spectrum, and the use of the metabolic biomarkers C4H7N3O, C4H9N3O2, C6H 12 O6, C 21 H 30 O3, C24 H 38 O3、C 23 H 42 NO4、C 10 H 15 N5O 10 P2。

[0033] Compared with the prior art, the present application has the following beneficial effects:

[0034] The present application uses inorganic nanoparticles to enhance the laser desorption ionization time-of-flight mass spectrometry technology, which can process 300 samples per hour, and the detection time of each sample is about 30 seconds; the detection lower limit can reach 0.3 pmol; the processing efficiency is high; at the same time, the processed results have high repeatability, and the coefficient of variation is less than 10%, and the results are accurate;

[0035] The present application uses six machine learning algorithms (algorithm 1 lasso regression, algorithm 2 extreme gradient boosting, algorithm 3 support vector machine, algorithm 4 random forest, algorithm 5 adaptive boosting, and algorithm 6 deep learning) on the training set to evaluate the model, and the final prediction model result is more accurate;

[0036] The present application screens 7 differential metabolites through Fold-change (fold change) and Student's t-test (Student's T test) and Integrated Gradient (integrated gradient), and explains the contribution of each metabolite in the prediction model by introducing Integrated Gradient (integrated gradient), so as to ensure that the screened metabolites are reliable and can be used for prediction or diagnosis of Parkinson's disease.

[0037] The concept, specific structure and generated technical effects of the present application will be further described below in combination with the drawings, so as to fully understand the purpose, features and effects of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is the metabolic fingerprint spectrum of 125 healthy volunteers and 187 Parkinson's patients collected by the method of preferred embodiment 1 of the present application;

[0039] Figure 2 is the 405 m / z signal corresponding heat map collected after the saliva metabolic fingerprint spectrum of 312 saliva samples is pretreated by MATLAB R2020a in embodiment 2 of the present application;

[0040] Figure 3is the prediction efficiency chart of Parkinson's disease using 6 machine learning algorithms (algorithm 1 lasso regression, algorithm 2 extreme gradient boosting, algorithm 3 support vector machine, algorithm 4 random forest, algorithm 5 adaptive boosting, algorithm 6 deep learning) in embodiment 2 of the present application;

[0041] Figure 4 is the receiver operating characteristic curve of the subjects using deep learning (algorithm 6) for Parkinson's disease and healthy volunteers (test set);

[0042] Figure 5 is the Integrated Gradient analysis used to evaluate the contribution value of each metabolite in the prediction model in embodiment 3 of the present application. DETAILED DESCRIPTION

[0043] The preferred embodiments of the present application are described below with reference to the accompanying drawings, so that the technical content of the present application is more clear and easy to understand. The present application can be embodied in many different forms, and the scope of protection of the present application is not limited to the embodiments described herein.

[0044] In the drawings, components of the same structure are denoted by the same reference numerals, and components having similar structures or functions are denoted by similar reference numerals. The size and thickness of each component shown in the drawings are arbitrarily shown, and the present application is not limited to the size and thickness of each component. In order to make the drawing clearer, the thickness of some components is appropriately exaggerated in some places in the drawing.

[0045] Embodiment 1

[0046] Inorganic nanoparticles are used to enhance the laser desorption ionization time-of-flight mass spectrometry technology to detect metabolites in saliva samples of Parkinson's disease patients and healthy volunteers, and obtain saliva metabolic fingerprint. Laser desorption ionization time-of-flight mass spectrometry has the advantages of fast detection speed and high throughput, and comprises the following steps:

[0047] Step 1, 1: dilute the saliva samples of Parkinson's disease patients and healthy volunteers 10 times with deionized water;

[0048] Step 1, 2: prepare a 1 mg / mL matrix solution of inorganic nanoparticles with deionized water;

[0049] Step 1, 3: spot the diluted saliva sample (1 μL) on the mass spectrometry target plate, and naturally dry at room temperature to complete the preparation of the saliva sample on the mass spectrometry target plate;

[0050] Step 1, 4: spot 1 mg / mL matrix solution (1 μL) on the mass spectrometry target plate, and naturally dry at room temperature to complete the preparation of the matrix on the mass spectrometry target plate;

[0051] Step 1, 5: Collect saliva metabolite fingerprint of Parkinson's disease patients and healthy volunteers by using laser desorption ionization time-of-flight mass spectrometry technology, and obtain saliva metabolite fingerprint spectrum.

[0052] Example 2

[0053] The saliva sample of the healthy volunteer control group and Parkinson's disease patients is analyzed by the method of Example 1 to obtain the saliva metabolite fingerprint spectrum, and the saliva samples of the Parkinson's disease patients and the healthy control group are trained by using machine learning algorithm; various models are evaluated by using cross-validation method, and the final prediction model is selected according to the verification result; the established model is used to predict new saliva samples to determine whether the sample belongs to Parkinson's disease patients or healthy volunteer control group. This method can be used to screen and predict the saliva metabolite fingerprint spectrum of Parkinson's population, which specifically includes the following steps:

[0054] Step 1: The saliva samples of 125 healthy volunteer control group and 187 Parkinson's patients are analyzed by the method of Example 1 to obtain the saliva metabolite fingerprint spectrum, and the spectrum is shown in Figure 1 .

[0055] Step 2: The saliva metabolite fingerprint spectrum of 312 samples is pretreated on MATLAB (R2020a), including spectrum smoothing, baseline correction and spectrum peak matching, etc., and finally 405 m / z signals are collected for each sample, as shown in Figure 2 .

[0056] Step 3: The 312 saliva samples are divided into training set and test set, wherein the training set includes 85 healthy control saliva samples and 127 Parkinson's saliva samples, and the test set includes 40 healthy control saliva samples and 60 Parkinson's saliva samples.

[0057] Step 4: The 6 machine learning algorithms (algorithm 1 lasso regression, algorithm 2 extreme gradient boosting, algorithm 3 support vector machine, algorithm 4 random forest, algorithm 5 adaptive boosting, and algorithm 6 deep learning) are used to evaluate the model on the training set, and the final prediction model is selected; the test set is tested in the final prediction model to obtain the prediction performance, wherein the diagnostic performance of the training set and the test set on the training model is shown in Figure 3 and Table 1.

[0058] Table 1. Diagnostic performance of 405 m / z signals using deep learning (algorithm 6) in the prediction model

[0059]

[0060]

[0061] ByFigure 3 As shown in Table 1, algorithm 6 (deep learning) has the highest confidence interval and sensitivity, and therefore, algorithm 6 is selected as the final prediction model. The receiver operating characteristic curve of algorithm 6 for the Parkinson's and healthy volunteer control (test set) is shown in Figure 4 .

[0062] Example 3

[0063] According to the prediction results of the prediction model, the saliva samples of the Parkinson's patients and the healthy volunteers are predicted to obtain prediction results, and the metabolic biomarkers in the saliva samples are screened according to the prediction results. Specifically, the following steps are included:

[0064] Step 1: Fold-change analysis is performed on the saliva fingerprint of the healthy volunteer control group and the Parkinson's patients to compare the differences in metabolic data between the two groups, and to screen metabolites with a Fold-change value greater than 1.20 or less than 0.83;

[0065] Step 2: Student's t-test analysis is performed on the saliva fingerprint of the healthy control group and the Parkinson's patients to compare whether there is a significant difference in the levels of metabolites between the Parkinson's group and the healthy control group, and to screen metabolites with a P value less than 0.05;

[0066] Step 3: Integrated Gradient analysis is performed on the saliva fingerprint of the healthy control group and the Parkinson's patients to explain the prediction results of the prediction model, calculate the contribution of each metabolite to the prediction results of the model, and screen metabolites with a contribution greater than 0.2, as shown in Figure 5 .

[0067] Step 4: Cross-match the metabolites screened by Fold-change, Student's t-test, and Integrated Gradient, and finally screen 7 differential metabolites, and identify 7 metabolic biomarkers, as shown in Table 2.

[0068] Table 2. Information of 7 metabolic biomarkers

[0069]

[0070]

[0071] Therefore, the above-mentioned 7 metabolic biomarkers can be used for predicting or diagnosing Parkinson's disease.

[0072] The preferred embodiments of the present application have been described above in detail. It should be understood that modifications and variations to the preferred embodiments could be made by those skilled in the art without departing from the spirit and scope of the application. Accordingly, it is intended that there be included within the scope of the application, all such modifications and variations as would be apparent to those skilled in the art upon reading this disclosure. It is intended to obtain for the inventors such patent rights as are available for any patent granted on the present application.

Claims

1. The use of a metabolic biomarker in predicting Parkinson's disease, characterized in that, The metabolic biomarkers are selected from C4H7N3O, C4H9N3O2, and C6H 12 O6, C 21 H 30 O3, C 24 H 38 O3, C 23 H 42 NO4, C 10 H 15 N5O 10 P2.

Images