Active tuberculosis diagnosis method based on proteomics and metabonomics

Abstract

The invention discloses an active tuberculosis diagnosis method based on proteomics and metabonomics. The method comprises the following steps: collecting a serum sample of a suspected active tuberculosis patient; detecting specific proteins in the serum sample by using a proteomics technology; detecting specific amino acids in the serum sample by using a metabonomics technology; carrying out standardization treatment on the obtained protein, amino acid and bile acid data; establishing a diagnosis model by using a machine learning algorithm; inputting data of a to-be-diagnosed sample into the model, and calculating a probability value of the to-be-diagnosed sample suffering from active tuberculosis; according to a preset threshold value, whether the sample provider suffers from the active tuberculosis or not is judged. By comprehensively analyzing multiple omics data, the sensitivity and specificity of diagnosis are greatly improved; by utilizing a machine learning algorithm, the disease risk can be objectively and quantitatively evaluated, and human misjudgment is reduced; the detection process is simple and fast, the sample demand is small, and clinical popularization and application are facilitated.

Description

Technical Field

[0001] The present invention relates to the field of medical diagnosis technology, in particular to a method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics. Background Art

[0002] Tuberculosis is a chronic infectious disease caused by Mycobacterium tuberculosis, which mainly affects the lungs and can also affect other organs and systems. According to the World Health Organization, about one-third of the world's population is infected with Mycobacterium tuberculosis, with more than 10 million new cases and more than 1 million deaths each year. It is one of the major infectious diseases that endanger human health.

[0003] Early diagnosis is the key to controlling the spread of tuberculosis and improving treatment outcomes. However, existing tuberculosis diagnostic methods have some limitations: 1) sputum smear microscopy has low sensitivity and is prone to missed diagnosis; 2) culture methods have a long cycle and limited clinical significance; 3) molecular biological methods lack specificity and are prone to misdiagnosis; 4) tuberculin tests and interferon-γ release tests cannot distinguish between latent infection and active tuberculosis; 5) imaging manifestations lack specificity and are difficult to identify from other lung diseases. Therefore, it is urgent to develop new diagnostic techniques to improve the sensitivity and specificity of tuberculosis diagnosis.

[0004] In recent years, multi-omics technologies represented by proteomics and metabolomics have shown great potential in the discovery and diagnosis of disease markers. Proteomics can analyze disease-related biomarkers at the protein level and reveal the molecular mechanisms of disease occurrence and development. Metabolomics focuses on endogenous small molecule metabolites of the body, reflecting the overall changes in the metabolic network under disease conditions. The combination of the two can describe the biological characteristics of the disease from multiple levels and angles, providing a more comprehensive and reliable basis for diagnosis.

[0005] The present invention is proposed based on the above background, and aims to establish a new method for diagnosing active pulmonary tuberculosis by utilizing proteomics and metabolomics technologies, detecting specific protein, amino acid and bile acid markers, and combining with machine learning algorithms, in order to improve the sensitivity and specificity of diagnosis and provide strong support for the prevention and control of pulmonary tuberculosis. Summary of the invention

[0006] The object of the present invention is to provide a method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics to solve the problems raised in the above background technology.

[0007] In order to achieve the above object, the present invention adopts the following technical scheme: a method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics, comprising the following steps:

[0008] 1) Collect serum samples from patients suspected of active pulmonary tuberculosis;

[0009] 2) Use proteomics technology to detect specific proteins in serum samples;

[0010] 3) Use metabolomics techniques to detect specific amino acids in serum samples;

[0011] 4) Standardize the obtained protein, amino acid and bile acid data;

[0012] 5) Use machine learning algorithms to build diagnostic models, use data from healthy controls and confirmed active TB patients to train the models, and evaluate model performance through cross-validation;

[0013] 6) Input the standardized data of the sample to be diagnosed into the trained model to calculate the probability value of the sample suffering from active pulmonary tuberculosis;

[0014] 7) Determine whether the sample provider suffers from active pulmonary tuberculosis based on a preset threshold.

[0015] Preferably, the proteomics technique in step 2) is immunoturbidimetry technique.

[0016] Preferably, the specific proteins in step 2) include apolipoprotein A1 (APOA1), apolipoprotein B (APOB), complement C3 (C3), fibronectin (FN), apolipoprotein A2 (APOA2) and apolipoprotein C3 (APOC3).

[0017] Preferably, the metabolomics technique in step 3) is liquid chromatography-triple quadrupole mass spectrometry (LC / MS) technique.

[0018] Preferably, the specific amino acids in step 3) include alanine, glutamic acid, citrulline, arginine, lysine, tyrosine, ornithine, valine and histidine, and specific bile acids include glycine-conjugated bile acid (GCA), glycine-conjugated deoxycholic acid (GCDCA), glycine-conjugated linoleic acid (GLCA), glycine-conjugated cholic acid (GDCA), taurine-conjugated bile acid (TCA), taurine-conjugated deoxycholic acid (TCDCA), taurine-conjugated linoleic acid (TLCA), taurine-conjugated cholic acid (TDCA), taurine-conjugated ursodeoxycholic acid (TUDCA), linoleic acid (LCA), deoxycholic acid (DCA) and ursodeoxycholic acid (UDCA).

[0019] Preferably, the standardization process in step 4) includes logarithmic transformation, centering and scaling, etc., and specifically includes the following sub-steps: logarithmic transformation of the data to make it close to normal distribution; calculating the mean and standard deviation of each variable, subtracting the mean from the original data and dividing it by the standard deviation to achieve centering and scaling.

[0020] Preferably, the step 5) uses a machine learning algorithm including logistic regression, random forest or support vector machine, the logistic regression model uses L1 regularization, and the penalty term coefficient is determined by 10-fold cross validation; the number of trees in the random forest model is set to 500, and the square root of the number of features is used as the maximum number of features for each tree; the support vector machine model uses a linear kernel function, and the penalty term coefficient is determined by 10-fold cross validation.

[0021] Preferably, the threshold in step 7) is determined by receiver operating characteristic curve (ROC curve) analysis, and the probability value corresponding to the maximum Youden index is used as the optimal threshold. If the probability value is greater than or equal to the threshold, it is judged as positive, otherwise it is judged as negative.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] 1) Through comprehensive analysis of proteomics and metabolomics data, the biological characteristics of active pulmonary tuberculosis are characterized from multiple levels, greatly improving the sensitivity and specificity of diagnosis;

[0024] 2) Using machine learning algorithms, we can mine key information with diagnostic value from high-dimensional data, objectively and quantitatively assess the risk of disease, and reduce human misjudgment;

[0025] 3) The amount of serum sample required for the detection process is small, sampling is convenient, and the detection cycle is short, which is conducive to clinical promotion and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 The schematic diagram of the immunoturbidimetric method of the present invention for determining specific proteins in the serum of 100 cases of active pulmonary tuberculosis and 100 cases of healthy controls;

[0027] Figure 2 Schematic diagram of the liquid chromatography / mass spectrometry (LC / MS) method of the present invention for determining specific bile acids in the serum of 100 cases of active pulmonary tuberculosis and 100 healthy controls Figure 1 ;

[0028] Figure 3 Schematic diagram of the liquid chromatography / mass spectrometry (LC / MS) method of the present invention for determining specific bile acids in the serum of 100 cases of active pulmonary tuberculosis and 100 healthy controls Figure 2 ;

[0029] Figure 4 Schematic diagram of the liquid chromatography / mass spectrometry (LC / MS) method of the present invention for determining specific bile acids in the serum of 100 cases of active pulmonary tuberculosis and 100 healthy controls Figure 3 ;

[0030] Figure 5Schematic diagram of the liquid chromatography / mass spectrometry (LC / MS) method of the present invention for determining specific amino acids in the serum of 100 cases of active pulmonary tuberculosis and 100 healthy controls;

[0031] Figure 6 The figure is a schematic diagram of ROC curve analysis for distinguishing tuberculosis from non-tuberculosis by the active tuberculosis detection and diagnosis model composed of a specific protein, bile acid and amino acid combination of the present invention. DETAILED DESCRIPTION

[0032] In order to further explain the technical solution of the present invention, it is described in detail below through specific embodiments.

[0033] See also Figure 1-Figure 6 The present invention provides a method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics, comprising the following steps:

[0034] 1. Collect serum samples from patients suspected of active pulmonary tuberculosis:

[0035] A total of 100 patients with active pulmonary tuberculosis and 100 healthy controls were included, including 100 males and 100 females, aged 18-65 years old, with an average age of (42.5±13.7) years old. All subjects signed informed consent. 5 mL of venous blood was drawn in the morning on an empty stomach, centrifuged at 3000 rpm for 10 minutes, and serum was separated and stored in a -80℃ refrigerator for later use.

[0036] 2. Use proteomics technology to detect specific proteins in serum samples. The proteomics technology is immunoturbidimetry technology. The specific proteins include apolipoprotein A1 (APOA1), apolipoprotein B (APOB), complement C3 (C3), fibronectin (FN), apolipoprotein A2 (APOA2) and apolipoprotein C3 (APOC3): The specific detection steps are as follows:

[0037] 1) Sample preparation: Dilute the serum sample to be tested appropriately so that its concentration is within the detection range.

[0038] 2) Add reagents: Mix the diluted serum sample with a specific antibody reagent, which is usually a polyclonal or monoclonal antibody.

[0039] 3) Antigen-antibody reaction: The target protein (antigen) in the sample binds to the specific antibody to form an antigen-antibody complex.

[0040] 4) Turbidimetric Assay: The formation of antigen-antibody complexes will lead to an increase in the turbidity of the solution. A turbidimeter is used to measure the turbidity change of the solution at a specific wavelength.

[0041] 5) Data analysis: The measured turbidity values ​​are converted to the concentration of the target protein according to the standard curve. The standard curve is drawn using standards of known concentrations.

[0042] 6) Data test: SPSS22.0 and MedCalc 15.0 were used for statistical analysis. The normal distribution test method was KS test. Normally distributed data were expressed as x±s, and the independent sample t test was used for inter-group comparison; non-normally distributed quantitative data were expressed as M(Q1,Q3), and the Anna-Whitney U test was used for inter-group comparison.

[0043] 3) Metabolomics technology was used to detect specific amino acids in serum samples. The metabolomics technology was liquid chromatography-triple quadrupole mass spectrometry (LC / MS) technology. Specific amino acids included alanine, glutamic acid, citrulline, arginine, lysine, tyrosine, ornithine, valine and histidine, and specific bile acids included glycine-conjugated bile acid (GCA), glycine-conjugated deoxycholic acid (GCDCA), glycine-conjugated linoleic acid (GLCA), glycine-conjugated cholic acid (GDCA), taurine-conjugated bile acid (TCA), taurine-conjugated deoxycholic acid (TCDCA), taurine-conjugated linoleic acid (TLCA), taurine-conjugated cholic acid (TDCA), taurine-conjugated ursodeoxycholic acid (TUDCA), linoleic acid (LCA), deoxycholic acid (DCA) and ursodeoxycholic acid (UDCA). The specific detection steps were as follows:

[0044] Take 50 μL of serum sample, add 200 μL of pre-cooled methanol, vortex shake, centrifuge at 4°C, 12000 rpm for 10 min, and filter the supernatant through a 0.22 μm filter for LC-MS / MS analysis. The chromatographic system is Agilent 1290 Infinity II UHPLC, the chromatographic column is Waters ACQUITY UPLC BEH Amide column (2.1 mm × 100 mm, 1.7 μm), the mobile phase A is an aqueous solution of 25 mM ammonium acetate and 25 mM ammonia, the mobile phase B is acetonitrile, gradient elution, flow rate 0.5 mL / min, column temperature 25°C, injection volume 2 μL. The mass spectrometry system was Agilent 6460Triple Quad LC / MS, using an AJS electrospray ion source (AJS-ESI), multiple reaction monitoring (MRM) in positive and negative ion modes, capillary voltage 4000 V for positive ion, 3500 V for negative ion, vaporization temperature 250 °C, air flow 5 L / min, nebulizer gas pressure 45 psi, and fragmentation voltage 380 V. Agilent MassHunter workstation software was used for data acquisition and processing.

[0045] 4. The obtained protein, amino acid and bile acid data were standardized, including logarithmic transformation, centering and scaling. The standardization process includes the following sub-steps:

[0046] The data were logarithmically transformed to make them close to normal distribution; the mean and standard deviation of each variable were calculated, and the original data were subtracted from the mean and divided by the standard deviation to achieve centering and scaling;

[0047] The data were logarithmically transformed, centered, and scaled using R language. The logarithmic transformation was performed using the log2 function, and the centering and scaling were performed using the scale function.

[0048] 5. Use machine learning algorithms, such as logistic regression, random forest, or support vector machine, to build a diagnostic model, use data from healthy controls and confirmed active TB patients to train the model, and evaluate model performance through cross-validation;

[0049] The logistic regression model uses L1 regularization, and the penalty coefficient is determined by 10-fold cross validation. The number of trees in the random forest model is set to 500, and the square root of the number of features is used as the maximum number of features for each tree. The support vector machine model uses a linear kernel function, and the penalty coefficient is determined by 10-fold cross validation.

[0050] The specific steps were as follows: the data of 100 healthy controls and 100 confirmed active pulmonary tuberculosis patients were randomly divided into a training set (80%) and a test set (20%), and the logistic regression, random forest, and support vector machine models were constructed using the glmnet, randomForest, and e1071 packages in the R language, respectively. The logistic regression model used L1 regularization, and the optimal penalty coefficient was determined by the cv.glmnet function; the number of trees in the random forest model was set to 500, and the square root of the number of features was used as the maximum number of features for each tree; the support vector machine model used a linear kernel function, and the optimal penalty coefficient was determined by the tune.svm function. The model performance was evaluated by 10-fold cross validation, and the sensitivity, specificity, positive predictive value, negative predictive value, and AUC values ​​were calculated.

[0051] 6. Input the standardized data of the sample to be diagnosed into the trained model and calculate the probability value of active pulmonary tuberculosis.

[0052] 7. According to the preset threshold, determine whether the sample provider has active pulmonary tuberculosis. If the probability value is greater than or equal to the threshold, it is judged as positive, otherwise it is judged as negative. The threshold is determined by receiver operating characteristic curve (ROC curve) analysis, and the probability value corresponding to the maximum Youden index is taken as the optimal threshold.

[0053] The ROC curve was drawn using the pROC package, and the probability value corresponding to the maximum Youden index was used as the optimal threshold. If the probability value of the sample is greater than or equal to the threshold, it is judged as positive, otherwise it is judged as negative.

[0054] The results showed that the sensitivity of the logistic regression model was 92%, the specificity was 90%, the positive predictive value was 90.2%, the negative predictive value was 91.8%, and the AUC value was 0.96; the sensitivity of the random forest model was 94%, the specificity was 92%, the positive predictive value was 92.2%, the negative predictive value was 93.9%, and the AUC value was 0.98; the sensitivity of the support vector machine model was 90%, the specificity was 88%, the positive predictive value was 88.2%, the negative predictive value was 89.8%, and the AUC value was 0.95.

[0055] The present invention provides a method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics, which is expected to become a powerful supplementary means. Combined with existing imaging, bacteriological and immunological examination methods, a more complete and accurate strategy for diagnosing pulmonary tuberculosis can be formed, providing strong support for the early detection and timely treatment of pulmonary tuberculosis, and has broad application prospects. The method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics established by the present invention has high diagnostic efficiency, can effectively make up for the shortcomings of existing diagnostic means, and provide an important basis for clinical diagnosis and treatment decisions.

[0056] The above description is only a preferred example of the present invention and is not intended to limit the present invention. Although the present invention is described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics, characterized in that: The following steps are involved: 1) Collect serum samples from patients suspected of active pulmonary tuberculosis; 2) Use proteomics technology to detect specific proteins in serum samples; 3) Use metabolomics technology to detect specific amino acids in serum samples; 4) Standardize the obtained protein, amino acid and bile acid data; 5) Use machine learning algorithms to build diagnostic models, use data from healthy controls and confirmed active TB patients to train the models, and evaluate model performance through cross-validation; 6) Input the standardized data of the sample to be diagnosed into the trained model to calculate the probability value of the sample suffering from active pulmonary tuberculosis; 7) Determine whether the sample provider suffers from active pulmonary tuberculosis based on a preset threshold.

2. A method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics according to claim 1, characterized in that: The proteomics technique in step 2) is immunoturbidimetry technique.

3. The method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics according to claim 1, characterized in that: The specific proteins in step 2) include apolipoprotein A1 (APOA1), apolipoprotein B (APOB), complement C3 (C3), fibronectin (FN), apolipoprotein A2 (APOA2) and apolipoprotein C3 (APOC3).

4. The method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics according to claim 1, characterized in that: The metabolomics technology in step 3) is liquid chromatography-triple quadrupole mass spectrometry (LC / MS) technology.

5. The method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics according to claim 1, characterized in that: The specific amino acids in step 3) include alanine, glutamic acid, citrulline, arginine, lysine, tyrosine, ornithine, valine and histidine, and specific bile acids include glycine-conjugated bile acid (GCA), glycine-conjugated deoxycholic acid (GCDCA), glycine-conjugated linoleic acid (GLCA), glycine-conjugated cholic acid (GDCA), taurine-conjugated bile acid (TCA), taurine-conjugated deoxycholic acid (TCDCA), taurine-conjugated linoleic acid (TLCA), taurine-conjugated cholic acid (TDCA), taurine-conjugated ursodeoxycholic acid (TUDCA), linoleic acid (LCA), deoxycholic acid (DCA) and ursodeoxycholic acid (UDCA).

6. The method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics according to claim 1, characterized in that: The standardization process in step 4) includes logarithmic transformation, centering and scaling, etc., and specifically includes the following sub-steps: logarithmic transformation of the data to make it close to normal distribution; calculating the mean and standard deviation of each variable, subtracting the mean from the original data and dividing it by the standard deviation to achieve centering and scaling.

7. The method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics according to claim 1, characterized in that: The step 5) uses a machine learning algorithm including logistic regression, random forest or support vector machine, etc. The logistic regression model uses L1 regularization, and the penalty term coefficient is determined by 10-fold cross validation; the number of trees in the random forest model is set to 500, and the square root of the number of features is used as the maximum number of features for each tree; the support vector machine model uses a linear kernel function, and the penalty term coefficient is determined by 10-fold cross validation.

8. The method for diagnosing active pulmonary tuberculosis based on proteomics and metabolomics according to claim 1, characterized in that: The threshold in step 7) is determined by receiver operating characteristic curve (ROC curve) analysis, and the probability value corresponding to the maximum Youden index is used as the optimal threshold. If the probability value is greater than or equal to the threshold, it is judged as positive, otherwise it is judged as negative.

Images