Plasma metabolism marker screening method, equipment, medium and product

Screening plasma metabolic markers through liquid chromatography and machine learning algorithms has solved the problem of early diagnosis of renal cancer, achieved efficient screening and diagnosis of renal cancer, simplified the diagnosis process, improved the treatment effect and patient survival rate .

CN119068993APending Publication Date: 2024-12-03PEKING UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411029230.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-30
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing renal cancer diagnosis methods mainly rely on imaging examinations and invasive surgery, resulting in a low early diagnosis rate and a lack of effective liquid biopsy methods, making it difficult to achieve early screening and diagnosis.

Method used

The combined liquid chromatography and mass spectrometry technology combined with support vector machine model and greedy algorithm were used to screen out renal cancer-related metabolic markers through plasma metabolomic analysis, and a biomarker combination for kidney cancer diagnosis was constructed.

Benefits of technology

It has realized the early diagnosis of renal cancer, simplified the diagnosis process, reduced patient discomfort and medical costs, improved diagnostic efficiency, and has significant clinical application value.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068993A_ABST
    Figure CN119068993A_ABST
Patent Text Reader

Abstract

The invention discloses a plasma metabolism marker screening method, equipment, a medium and a product, and relates to the technical field of plasma metabolism marker screening. The method comprises the following steps: treating each plasma sample in a plasma sample set by using a liquid chromatography-mass spectrometry technology, obtaining an original metabolic fingerprint spectrum corresponding to each plasma sample, and performing first feature extraction processing to obtain a first feature matrix corresponding to each plasma sample; training a support vector machine model by taking the first feature matrix as input and the tag as output to obtain a metabolism feature sequence; according to the metabolism characteristic sequence, a plasma metabolism marker set is determined through a greedy algorithm and a support vector machine model, the rationality of plasma metabolism marker screening can be improved through the liquid chromatography-mass spectrometry technology and the greedy algorithm, and then a basis is provided for manufacturing of kidney cancer early-stage detection equipment; and the automation level of early screening of the kidney cancer is improved while the professional property of kidney cancer screening personnel is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of plasma metabolic marker screening, and in particular to a plasma metabolic marker screening method, equipment, medium and product. Background Art

[0002] Currently, the diagnosis of kidney cancer mainly relies on imaging examinations and invasive surgery, which results in most patients being diagnosed in the late stage. Early screening is crucial to reducing the mortality rate of kidney cancer. Treatment outcomes are closely related to early diagnosis. The surgical cure rate of early kidney cancer can exceed 80%, while the 5-year survival rate of advanced kidney cancer is less than 30%. Therefore, early screening, diagnosis and treatment are the best ways to improve the management of kidney cancer. However, there is currently no effective liquid biopsy method for the diagnosis of kidney cancer.

[0003] Renal cancer cells show a variety of metabolic abnormalities, including but not limited to enhanced glycolysis, tryptophan metabolism, glutamine carboxylation, lactic acid fermentation triggered by hypoxia-inducible factor α (HIF-α), and upregulation of fatty acid synthesis. In addition, carnitine β-oxidation is significantly downregulated. These metabolic disorders in tumor cells may be reflected in plasma, providing an opportunity to identify potential biomarkers for renal cancer. Metabolomics refers to a high-throughput detection technology that detects small molecules less than 1200 Da related to metabolism in body fluids or tissues, which can accurately identify and quantify sugars, nucleotides, amino acids, and lipids. Compared with other omics, the metabolome is considered to be the closest to the phenotype and can sensitively reflect the state of the body. Especially in the field of cancer research, metabolic disorders are regarded as a key feature of malignant tumor cells. In recent years, metabolomics has been widely used in the discovery of biomarkers and pathological mechanism studies of various diseases including cancer, and has broad prospects for clinical application. With the successful application of artificial intelligence algorithms in the medical field, the combination of artificial intelligence and plasma metabolomics is expected to achieve early and accurate diagnosis of diseases such as tumors. Summary of the invention

[0004] The purpose of this application is to provide a plasma metabolic marker screening method, equipment, medium and product, which can improve the rationality of plasma metabolic marker screening, and further provide a basis for the manufacture of early detection equipment for renal cancer, while reducing the professionalism of renal cancer screening personnel and improving the level of automation of early screening for renal cancer.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for screening plasma metabolic markers, comprising:

[0007] Acquire a plasma sample set; any plasma sample in the plasma sample set is labeled as a renal cancer patient or a healthy control;

[0008] Each plasma sample in the plasma sample set is processed separately by using liquid chromatography-mass spectrometry technology to obtain the original metabolic fingerprint corresponding to each plasma sample;

[0009] Performing first feature extraction processing on the plurality of original metabolic fingerprints respectively to obtain a first feature matrix corresponding to each plasma sample; the first feature matrix is ​​used to describe the metabolic features of the corresponding plasma sample;

[0010] The support vector machine model is trained with the first feature matrix as input and the label as output to obtain a metabolic feature sequence; the metabolic feature sequence is obtained by arranging all metabolic features in the feature matrix in descending order according to the weights output by the support vector machine model, and intercepting the first preset number of metabolic features;

[0011] According to the metabolic feature sequence, a plasma metabolic marker set is determined using a greedy algorithm and a support vector machine model.

[0012] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned plasma metabolic marker screening method.

[0013] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned plasma metabolic marker screening method when executed by a processor.

[0014] In a fourth aspect, the present application provides a computer program product, including a computer program, which implements the above-mentioned plasma metabolic marker screening method when executed by a processor.

[0015] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0016] The present application provides a method, device, medium and product for screening plasma metabolite markers, using liquid chromatography-mass spectrometry technology to process each plasma sample in a plasma sample set, and obtain the original metabolic fingerprint map corresponding to each plasma sample; perform first feature extraction processing on multiple original metabolic fingerprint maps, and obtain the first feature matrix corresponding to each plasma sample; use the first feature matrix as input and the label as output to train a support vector machine model to obtain a metabolic feature sequence; according to the metabolic feature sequence, use a greedy algorithm and a support vector machine model to determine a plasma metabolite marker set. This method combines plasma metabolomics and machine learning data analysis methods to develop an efficient biomarker combination and an accurate diagnostic model suitable for the diagnosis of renal cancer. Compared with traditional diagnostic methods, the diagnostic marker screening method proposed in the present invention is simple and convenient. The obtained model shows excellent diagnostic performance in practical applications and is suitable for kidney cancer, especially for the diagnosis of early asymptomatic kidney cancer. Early screening of kidney cancer can be achieved through simple blood sample testing, avoiding radiation or puncture damage to the human body caused by imaging or puncture sampling methods, which can reduce patient discomfort, reduce medical costs, improve diagnostic efficiency, and has significant clinical application advantages.

[0017] The early detection device for renal cancer constructed based on the method of the present invention has a simple and fast diagnostic process, which greatly improves the sampling efficiency of early screening of renal cancer patients, helps promote the implementation of early diagnosis and treatment of renal cancer, thereby improving the treatment effect and patient survival rate. Therefore, the present invention not only has important value in clinical use, but also has broad promotion potential in the field of public health. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 This is a flow chart of a method for screening plasma metabolic markers in one embodiment of the present application;

[0020] Figure 2 This is a schematic diagram of a plasma metabolic marker screening method in one embodiment of the present application;

[0021] Figure 3 This is a model accuracy score graph for positive ion mode feature screening in one embodiment of the present application;

[0022] Figure 4 This is a model accuracy score graph for negative ion mode feature screening in one embodiment of the present application;

[0023] Figure 5 This is a non-targeted metabolomics classification performance diagram based on thirty-two renal cancer diagnostic markers in one embodiment of the present application;

[0024] Figure 6 This is a non-targeted metabolomics classification performance diagram based on sixteen renal cancer diagnostic markers in one embodiment of the present application;

[0025] Figure 7 This is a non-targeted metabolomics classification performance diagram based on seven renal cancer diagnostic markers in one embodiment of the present application;

[0026] Figure 8 A typical selected ion chromatogram (EIC) diagram of the PRM targeted metabolome profile for constructing a diagnostic model for 7 plasma metabolite markers in one embodiment of the present application;

[0027] Fig. 9 The standard curve of the diagnostic marker azelaic acid in one embodiment of the present application and its linear equation and quantitative range diagram;

[0028] Fig.10 The standard curve of the diagnostic marker N,N-dimethylglycine in one embodiment of the present application and its linear equation and quantitative range diagram;

[0029] Fig.11 The standard curve and linear equation and quantitative range diagram of the diagnostic marker N-acetyl-L-aspartic acid in one embodiment of the present application;

[0030] Fig.12 The standard curve and linear equation and quantitative range diagram of the diagnostic marker N-ε-acetyl-L-lysine in one embodiment of the present application;

[0031] Fig.13 The standard curve of the diagnostic marker proline and its linear equation and quantitative range diagram in one embodiment of the present application;

[0032] Fig.14 This is a diagram showing the statistical difference of the diagnostic marker 2-hydroxyphenylacetic acid between renal cancer and normal control groups in one embodiment of the present application;

[0033] Fig.15 This is a diagram showing the statistical difference of the diagnostic marker azelaic acid between the renal cancer group and the normal control group in one embodiment of the present application;

[0034] Fig.16 This is a diagram showing the statistical difference of the diagnostic marker (Z,Z)-4-oxo-2,5-heptadienedioic acid between renal cancer and normal control groups in one embodiment of the present application;

[0035] Fig.17This is a diagram showing the statistical difference of the diagnostic marker N,N-dimethylglycine between renal cancer and normal control groups in one embodiment of the present application;

[0036] Fig.18 This is a diagram showing the statistical difference of the diagnostic marker N-acetyl-L-aspartic acid between renal cancer and normal control groups in one embodiment of the present application;

[0037] Fig.19 This is a diagram showing the statistical difference of the diagnostic marker N-ε-acetyl-L-lysine between renal cancer and normal control groups in one embodiment of the present application;

[0038] Fig. 20 This is a diagram showing the statistical difference of the diagnostic marker proline between renal cancer and normal control groups in an embodiment of the present application;

[0039] Fig.21 The ROC curve diagram of the renal cancer diagnosis model training cohort constructed using 7 plasma metabolic markers in an embodiment of the present application;

[0040] Fig. 22 This is a ROC curve diagram of the internal validation cohort of the renal cancer diagnostic model constructed using 7 plasma metabolic markers in an embodiment of the present application;

[0041] Fig.23 The figure is a ROC curve diagram of the external test cohort of the renal cancer diagnosis model constructed using 7 plasma metabolic markers in an embodiment of the present application;

[0042] Fig.24 This is a ROC curve diagram of a multi-center external validation cohort of a renal cancer diagnostic model constructed using 7 plasma metabolic markers in an embodiment of the present application;

[0043] Fig.25 The ROC curve of the prospective cohort of the renal cancer diagnosis model constructed using 7 plasma metabolic markers in an embodiment of the present application;

[0044] Fig.26 The ROC curve diagram of the renal cancer diagnosis model training cohort constructed using three plasma metabolic markers in an embodiment of the present application;

[0045] Fig. 27 This is a ROC curve diagram of the internal validation cohort of the renal cancer diagnostic model constructed using three plasma metabolic markers in an embodiment of the present application;

[0046] Fig.28 The figure is a ROC curve diagram of the external test cohort of the renal cancer diagnosis model constructed using three plasma metabolic markers in an embodiment of the present application;

[0047] Fig.29This is a ROC curve diagram of a multi-center external validation cohort of a renal cancer diagnostic model constructed using three plasma metabolic markers in an embodiment of the present application;

[0048] Fig.30 The ROC curve of the prospective cohort of the renal cancer diagnosis model constructed using three plasma metabolic markers in an embodiment of the present application;

[0049] Fig.31 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0051] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0052] In an exemplary embodiment, Figure 1 As shown, a method for screening plasma metabolic markers comprises:

[0053] Step 101: Obtain a plasma sample set. The label of any plasma sample in the plasma sample set is a renal cancer patient or a healthy control.

[0054] Step 102: Use liquid chromatography-mass spectrometry to process each plasma sample in the plasma sample set to obtain an original metabolic fingerprint corresponding to each plasma sample.

[0055] Step 103: Perform first feature extraction processing on the multiple original metabolic fingerprints to obtain a first feature matrix corresponding to each plasma sample. The first feature matrix is ​​used to describe the metabolic features of the corresponding plasma sample.

[0056] Step 104: Using the first feature matrix as input and the label as output, the support vector machine model is trained to obtain a metabolic feature sequence. The metabolic feature sequence is obtained by arranging all metabolic features in the feature matrix in descending order according to the weights output by the support vector machine model, and intercepting the first preset number of metabolic features.

[0057] Step 105: According to the metabolic feature sequence, a set of plasma metabolic markers is determined using a greedy algorithm and a support vector machine model.

[0058] After step 105, the method further includes:

[0059] Step 106: Obtain a second preset quantity.

[0060] Step 107: Selecting a second preset number of plasma metabolite markers from the plasma metabolite marker set as target metabolic features.

[0061] Step 108: preparing an extraction solution; the extraction solution contains multiple isotope internal standards; the isotope internal standards correspond to the target metabolic characteristics one by one; the isotope internal standards are used to absolutely quantify the target metabolic characteristics.

[0062] Step 109: using the extraction liquid to process each plasma sample in the plasma sample set separately to obtain multiple extracted samples.

[0063] Step 1010: using a parallel reaction monitoring targeted metabolomics quantitative method to process multiple extraction samples respectively to obtain multiple targeted metabolic fingerprints.

[0064] Step 1011: performing second feature extraction processing on multiple targeted original metabolic fingerprints respectively to obtain a second feature matrix corresponding to each plasma sample; the second feature matrix is ​​used to describe the target metabolic features of the corresponding plasma sample.

[0065] Step 1012: using the second feature matrix as input and the label as output, training the support vector machine model to obtain a kidney cancer screening model;

[0066] Step 1013: embed the computer program product that implements the renal cancer screening model into the processor to construct an early detection device for renal cancer. The input of the early detection device for renal cancer is the second characteristic matrix to be tested. The second characteristic matrix to be tested is obtained by processing the plasma sample of the subject to be tested using liquid chromatography-mass spectrometry technology to obtain the original metabolic fingerprint map to be tested, and performing second feature extraction processing on the original metabolic fingerprint map to be tested.

[0067] Step 1014: The sample to be tested is processed using a parallel reaction monitoring targeted metabolomics quantitative method to obtain a targeted metabolic fingerprint map to be tested; the sample to be tested is obtained by extracting the plasma sample of the subject to be tested using an extraction solution.

[0068] Step 1015: Perform feature extraction processing on the targeted metabolic fingerprint to be tested, and construct a second feature matrix to be tested; the second feature matrix to be tested is used to describe the target metabolic features of the corresponding plasma sample.

[0069] Step 1016: input the second characteristic matrix to be tested into the early detection device for kidney cancer to obtain the early detection result of kidney cancer of the subject to be tested.

[0070] Step 105 includes:

[0071] Step 105 - 1 : construct an empty set as the initial plasma metabolic marker set.

[0072] Step 105-2: Set the number of iterations i=1.

[0073] Step 105 - 3 : Let the number of target metabolic features to be determined n=1, and determine the first n metabolic features in the metabolic feature sequence as the target metabolic features to be determined.

[0074] Step 105-4: Perform feature extraction processing on multiple original metabolic fingerprints respectively, and select parameters corresponding to the target undetermined metabolic features to construct multiple feature matrices at the i-th iteration.

[0075] Step 105-5: construct a training set at the i-th iteration and a validation set at the i-th iteration according to the multiple feature matrices at the i-th iteration.

[0076] Step 105-6: Using multiple feature matrices in the training set at the i-th iteration as input and labels as output, the support vector machine model is trained to obtain the training model at the i-th iteration.

[0077] Step 105-7: Use the validation set at the i-th iteration to validate the training model at the i-th iteration, and obtain the accuracy of the training model at the i-th iteration.

[0078] Step 105 - 8 : Determine whether the number of iterations i is less than 2, and obtain a first determination result.

[0079] Step 105 - 9 : If the first judgment result is yes, then add the i-th metabolic feature in the metabolic feature sequence to the initial plasma metabolic marker set, increase the value of i by 1, and return to step 105 - 3 .

[0080] Step 105-10: If the first judgment result is no, determine whether the accuracy of the training model at the i-th iteration is greater than the accuracy of the training model at the i-1-th iteration to obtain a second judgment result.

[0081] Step 105 - 11: If the second judgment result is yes, then add the i-th metabolic feature in the metabolic feature sequence to the initial plasma metabolic marker set, increase the value of i by 1, increase the value of n by 1, and return to step 105 - 3.

[0082] Step 105 - 12: If the second judgment result is no, determine that the initial plasma metabolite marker set at the i-1th iteration is the plasma metabolite marker set, increase the value of i by 1, and return to step 105 - 3.

[0083] Below, the present invention is specifically described by taking 302 plasma samples of renal cancer patients (of which 84.4% are patients with T1 renal cancer) and 202 plasma samples of healthy controls as examples. Figure 2 , this implementation includes:

[0084] Screening of diagnostic markers for renal cancer:

[0085] 1. Obtain the training set.

[0086] The discovery cohort included plasma samples from 302 patients with renal cancer (84.4% of whom were patients with stage T1 renal cancer) and 202 plasma samples from healthy controls. The plasma from patients with renal cancer refers to the preoperative plasma from patients with renal cell carcinoma confirmed by postoperative pathology. Healthy controls refer to healthy people who were free of kidney disease after physical examination. Detailed information on the discovery cohort is shown in Table 1.

[0087] Table 1 Information of renal cancer and healthy controls in the discovery cohort

[0088]

[0089] 2. Non-targeted metabolomics analysis of plasma using liquid chromatography-mass spectrometry.

[0090] All plasma samples were centrifuged and stored in a -80°C refrigerator. Before non-targeted metabolomics analysis, plasma samples were taken out and a series of sample pretreatment steps were performed. The pretreated samples were subjected to metabolomics analysis using a high performance liquid chromatography-mass spectrometer to obtain the original metabolic fingerprint containing chromatographic and mass spectrometric data. The specific steps are as follows:

[0091] 2.1 Apparatus and reagents.

[0092] The experimental instruments include: a high performance liquid chromatography-mass spectrometry system (model U3000 / QEaxctive, ThermoFisher Scientific), an ultracentrifuge (Beckman Coulter), a constant temperature horizontal shaking mixer, a vacuum evaporator and an ultrapure water system (Millipore).

[0093] Experimental consumables include: Waters XBridge BEHAmide chromatographic column (specifications: 100×2.1mm, 2.5μm),

[0094] 1.5ml EP tube, 1.5ml injection bottle, 300μl inner tube, pipette, 1000μl and 200μl pipette tips, marker pen, disposable gloves and dust mask.

[0095] The experimental reagents include: methanol (Thermo Fisher, mass spectrometry grade), acetonitrile (Thermo Fisher, mass spectrometry grade), ammonium acetate (Merck), and pure water (TOC < 10 ppb).

[0096] 2.2 Pretreatment of plasma samples.

[0097] Before pretreatment of plasma samples, quality control samples (QC) were prepared by mixing 10 μl from each renal cancer plasma sample and healthy plasma sample and aliquoting. Subsequently, all renal cancer plasma samples, healthy plasma samples and quality control samples were subjected to the following pretreatment steps:

[0098] a. Use a pipette to take 50 μl of the analysis sample or quality control sample and place it in a 1.5 ml centrifuge tube.

[0099] b. Add 200 μl of methanol precooled to -20°C, shake horizontally to mix thoroughly for 30 minutes, centrifuge at 12,000 rpm at 4°C for 20 minutes, take the supernatant, and evaporate to dryness in a vacuum concentrator.

[0100] c. The dry powder sample was reconstituted with acetonitrile-water reconstitution solution (1:1, V / V) and centrifuged at 12,000 rpm for 20 minutes at 4°C.

[0101] d. Transfer the supernatant to a dedicated LC-MS injection bottle and prepare for liquid chromatography-mass spectrometry (LC-MS) analysis.

[0102] 2.3 Non-targeted metabolomics detection of plasma.

[0103] The processed renal cancer plasma samples and healthy plasma samples were used as the analysis objects. In order to avoid the deviation caused by the injection order, the injection order of the samples was randomized. After every 15 samples were analyzed, a quality control sample was added. The liquid chromatography and mass spectrometry conditions used in the analysis were as follows:

[0104] Mobile phase: Phase A was an acetonitrile aqueous solution containing 5 mM ammonium acetate (19:1, V / V); Phase B was pure acetonitrile.

[0105] Flow rate: 0.35 ml / min; column temperature: 30°C; injection volume: 5 μl.

[0106] Chromatographic gradient elution conditions: maintain 95% phase B for 0-2 minutes, linearly decrease phase B from 95% to 50% for 2-15 minutes, maintain 50% phase B for 15-18 minutes, linearly increase phase B from 50% to 95% for 18-19 minutes, and then maintain 95% phase B for 4 minutes.

[0107] Mass spectrometry conditions: electrospray ion source in positive ion mode (ESI+) and negative ion mode (ESI-), ion source temperature of 320℃, backflush gas set to 2, desolvation temperature of 300℃, sheath gas and auxiliary gas set to 40 and 10 respectively; capillary voltage in positive and negative ion modes were +3.3kV and -3kV, respectively, and cone voltage was set to 0V; data acquisition mode was data dependent acquisition (DDA); the mass-to-charge ratio (m / z) acquisition range of the primary mass spectrometer was 60-800, the resolution was 35000, the number of target ions was 5x106, and the maximum ion injection time was 80ms; the secondary mass spectrometer acquisition resolution was 17500, the number of target ions was 1x10 5 , the maximum ion injection time is 70ms, the number of cycles is 10 times, and the collision energies are 15, 30, and 45eV.

[0108] 3. Screening and identification of plasma metabolic markers.

[0109] The samples were subjected to non-targeted metabolome analysis according to the above chromatographic mass spectrometry conditions to obtain the original metabolic fingerprints. These profiles were comprehensively analyzed by applying machine learning algorithms to screen out potential biomarkers that can effectively distinguish renal cancer patients from healthy controls. The detailed operation steps are as follows:

[0110] 3.1 Metabolic data preprocessing.

[0111] Plasma samples were analyzed in positive ion mode (ESI+) and negative ion mode (ESI-) under HPLC-MS system to obtain the original metabolic fingerprint. After the raw mass spectrometry data were converted into ABF format using Reifycs file converter software, data preprocessing was performed using MS-Dial software. The preprocessing steps included retention time calibration, metabolite peak detection, peak matching, peak alignment, noise filtering, and data standardization. During the preprocessing process, the mass deviation of the primary mass spectrometry was set to 0.01Da, the mass deviation of the secondary mass spectrometry was set to 0.05Da, and other parameters were kept at the default settings. After preprocessing, a two-dimensional data matrix with metabolites as rows and samples as columns was obtained, and the median represented the content of the corresponding metabolites. The peak of each metabolite was qualitatively analyzed by its retention time and mass-to-charge ratio, and the peak area including isotope peaks, adducts and fragment ions was calculated. After preprocessing, 1257 metabolite peaks were identified in the positive ion mode and 781 metabolite peaks were identified in the negative ion mode, which will be used for subsequent machine learning analysis.

[0112] 3.2LC-MS experiment quality control.

[0113] During the LC-MS plasma non-targeted metabolomics analysis, the prepared QC samples were evenly inserted into the samples to be tested in the order of one QC sample after every 15 samples to be tested, which was used to continuously monitor the stability and repeatability of the entire process from sample injection pretreatment to final analysis and detection. After preprocessing the original metabolic fingerprint map with MS-Dial software, the coefficient of variation (%RSD) of each metabolite in the QC sample was further calculated. The results showed that the coefficient of variation of most metabolites was maintained below 30%, which demonstrated the excellent quality control in the experimental process and ensured the reliability and validity of the obtained metabolomics data.

[0114] 3.3 Apply the support vector machine (SVM) algorithm in machine learning to build a classification model.

[0115] The support vector machine (SVM) algorithm was used to perform pattern recognition learning on the preprocessed two-dimensional metabolite data matrix. In the model training stage, the plasma sample data of renal cancer patients and healthy controls were randomly divided, with 3 / 4 as the training set and 1 / 4 as the test set, and the randomization process was repeated 5000 times to enhance the generalization ability of the model. The performance parameters of the obtained model, including sensitivity, specificity, and accuracy, are shown in Table 2. In both positive and negative ion modes, the average accuracy of the model exceeded 0.97, which confirmed the high efficiency and accuracy of the SVM model in distinguishing renal cancer patients from healthy controls.

[0116] Table 2 Performance of the classification model based on the support vector machine (SVM) algorithm

[0117]

[0118]

[0119] 3.4 Screening and identification of plasma metabolic markers.

[0120] According to the support vector machine model established above, the weight score of the metabolic characteristics is obtained, the top 100 features ranked by weight are selected, and the greedy algorithm and the machine learning algorithm are combined for feature screening, and important features are gradually accumulated to build a test model. Among them, when the greedy algorithm gradually adds features, it will only be included in the feature set when the new features can improve the prediction efficiency, so as to ensure that each step of selection is moving in the direction of improving model performance. The machine learning algorithm does not perform screening when gradually accumulating features, and the top N features with average weight are used in the Nth cycle. The model classification accuracy is evaluated to show the classification efficiency of different models, and the optimal number of features and combination methods are determined; the standard for screening the optimal number of features and combination methods is: the model accuracy no longer increases when the number of features is increased. The feature selection score of the SVM model is as follows: Figure 3 and Figure 4 shown.

[0121] The selected target differential metabolites were optimized and screened based on mass spectrometry, and potential metabolite markers were screened and obtained based on the chromatographic peak shape and the quality of the secondary mass spectrometry spectrum using MS-Dial software. Based on the primary and secondary mass spectrometry information of potential metabolite markers, their molecular mass and molecular formula were inferred and compared with the information in the metabolite spectrum database to identify metabolites and ultimately determine plasma metabolite markers suitable for the diagnosis of renal cancer.

[0122] Through the above identification method, the present invention successfully identified 32 plasma metabolic markers as diagnostic markers suitable for the diagnosis of renal cancer, which can realize the diagnosis of renal cancer as early as possible. See Table 3. These markers are 1-allyl-2-hydroxy-7,8,9,10-tetrahydroazepino[2,1-b]quinazolin-12(6H)-one, 4-aminovaleric acid betaine, vinyl carbamate, 5'-S-methylthioadenosine, nicotinamide, L-alanine, L-ethionine, kobusone, chenodeoxycholic acid, columbianetin, phenylalanine betaine, betaine), (3S)-5-[(1S,8aR)-2,5,8a-tetramethyl-4-oxo-4a,6,7,8-tetrahydro-1H-naphthalen-1-yl]-3-methylpentanoic acid, Terbinafine, adenosine monophosphate (AMP), N,N-dimethylglycine, proline, L-alanyl-L-alanine (Ala-Ala), azelaic acid, (Z,Z)-4-oxo-2,5-hetpadienedioic acid acid), L-aspartic acid, mesaconic acid, 1H-benzotriazole, threonine, 2,4-dinitrophenol (2,4-Dinitrophenol), 2-Oxobutyricacid, D-Gluconic acid, N-Acetyl-L-aspartic acid, 6-Hydroxynicotinic acid, N-Methyllysine, 2-Aminoadipic acid, 2-Hydroxyphenylacetic acid, N-Epsilon-Acetyl-L-Lysine.

[0123] The 32 renal cancer-related metabolic markers identified in the present invention have an accuracy of 98.57% (95% confidence interval is 98.53% to 98.61%) in the non-targeted test set, which shows its high efficiency and reliability in renal cancer diagnosis. The detailed classification performance is shown in Figure 5 .

[0124] Furthermore, the present invention selects 16 renal cancer-related metabolic markers from the above 32 renal cancer metabolic markers, including: 1-allyl-2-hydroxy-7,8,9,10-tetrahydroazepino[2,1-b]quinazolin-12(6H)-one, vinyl carbamate, 5'-S-methylthioadenosine, columbianetin, L-alanyl-L-alanine (Ala-Ala), azelaic acid, (Z,Z)-4-oxo-2,5-hetpadienedioic acid, acid, Mesaconic acid, 1H-Benzotriazole, D-Gluconic acid, 6-Hydroxynicotinic acid, 2-Hydroxyphenylacetic acid, N-Epsilon-Acetyl-L-Lysine, N,N-Dimethylglycine, Proline, N-Acetyl-L-aspartic acid. These metabolite markers can achieve an accuracy of 98.41% (95% confidence interval of 98.37% to 98.46%) in the non-targeted test set, with similar diagnostic performance to the 32 diagnostic markers. For detailed classification performance, see Figure 6 .

[0125] At the same time, the present invention selects 7 renal cancer related metabolic markers from the above 32 renal cancer metabolic markers, which have an accuracy of 95.88% (95% confidence interval is 95.81% to 95.95%) in the non-targeted test set and also have efficient diagnostic performance. The detailed classification performance is shown in Figure 7 The application of these plasma metabolic markers provides a new strategy for the early detection and diagnosis of renal cancer. In the construction of diagnostic models based on plasma metabolic markers and in clinical diagnostic practice, these markers can be used alone or in specific combinations as needed.

[0126] Table 3 Thirty-two plasma metabolic markers

[0127]

[0128]

[0129] Next, the method described in this embodiment is specifically described by taking the construction of a kidney cancer diagnostic model using 16 plasma metabolic markers as an example.

[0130] 1. Research subjects.

[0131] The discovery cohort in Example 1 was targeted for retesting (one of which was not retested due to insufficient sample size) as a training cohort (comprising 302 plasma samples of renal cancer patients and 201 plasma samples of healthy controls), and a renal cancer diagnostic model was constructed using a support vector machine algorithm. The plasma of renal cancer patients refers to the preoperative plasma of patients with renal cell carcinoma confirmed by postoperative pathology. Healthy controls refer to healthy people who have no kidney disease after physical examination. Detailed information on the discovery cohort is shown in Table 1.

[0132] 2. Parallel reaction monitoring (PRM) targeted metabolomics analysis based on 16 renal cancer diagnostic markers.

[0133] All plasma samples were centrifuged and stored in a -80°C refrigerator. Before PRM targeted metabolomics analysis, plasma samples were taken out and a series of sample pretreatment steps were performed. The pretreated samples were subjected to metabolomics analysis using a high performance liquid chromatography-mass spectrometer to obtain the original metabolic fingerprint map containing chromatographic and mass spectrometric data. The specific steps are as follows:

[0134] 2.1 Apparatus and reagents.

[0135] The experimental instruments include: a high performance liquid chromatography-mass spectrometry system (model U3000 / QEaxctive, ThermoFisher Scientific), an ultracentrifuge (Beckman Coulter), a constant temperature horizontal shaking mixer, a vacuum evaporator and an ultrapure water system (Millipore).

[0136] Experimental consumables include: Waters XBridge BEHAmide chromatographic column (specifications: 100×2.1mm, 2.5μm),

[0137] 1.5ml EP tube, 1.5ml injection bottle, 300μl inner tube, pipette, 1000μl and 200μl pipette tips, marker pen, disposable gloves and dust mask.

[0138] The experimental reagents include: methanol (Thermo Fisher, mass spectrometry grade), acetonitrile (Thermo Fisher, mass spectrometry grade), ammonium acetate (Merck), and pure water (TOC < 10 ppb).

[0139] 2.2 Parallel reaction monitoring (PRM) targeted metabolomics quantification.

[0140] All renal cancer plasma samples, healthy plasma samples and quality control samples were pre-processed together. The specific operations are as follows:

[0141] (1) Sample preparation.

[0142] (a) Preparation of quality control samples: Before pretreatment of plasma samples, quality control samples (QC) were prepared by taking 10 μl from each renal cancer plasma sample and healthy plasma sample, mixing them and then aliquoting them.

[0143] (b) Prepare the extract containing the isotopic internal standard (IS): Use pre-cooled methanol solution to prepare [ 2 H8]-N-ε-acetyl-L-lysine ([ 2 H8]-N-Epsilon-Acetyl-L-Lysine) 10ng / ml.

[0144] (c) Extraction samples: Use a pipette to draw 50 μl of analytical sample / quality control sample, add 200 μl of the extraction solution prepared in step (a) to each tube of sample, shake for 20 minutes, centrifuge at 12,000 rpm at 4°C for 20 minutes, draw the supernatant, and evaporate to dryness in a vacuum concentrator.

[0145] (d) Reconstituted samples: The dry powder was reconstituted with acetonitrile-water reconstitution solution (1:1, V / V), and then centrifuged at 12,000 rpm for 20 min at 4°C in a high-speed centrifuge. The supernatant was transferred into an LC-MS injection vial.

[0146] (2). Parallel reaction monitoring (PRM) targeted metabolomics analysis.

[0147] The processed renal cancer plasma samples and healthy plasma samples were used as the analysis objects. In order to avoid the deviation caused by the injection order, the injection order of the samples was randomized. After every 15 samples were analyzed, a quality control sample was added. The liquid chromatography and mass spectrometry conditions used in the analysis were as follows:

[0148] Mobile phase: Phase A was an acetonitrile aqueous solution containing 5 mM ammonium acetate (19:1, V / V); Phase B was pure acetonitrile.

[0149] Flow rate: 0.35 ml / min; column temperature: 30°C; injection volume: 5 μl.

[0150] Chromatographic gradient elution conditions: 95% B from 0 to 2 minutes, gradually decreasing to 50% B from 2 to 15 minutes, maintaining at 50% B from 15 to 17.5 minutes, gradually increasing to 95% B from 17.5 to 17.8 minutes, and maintaining at 95% B from 17.8 to 19 minutes.

[0151] Mass spectrometry conditions: the electrospray ion source was used in positive ion mode (ESI+) and negative ion mode (ESI-), the ion source temperature was 320℃, the backflush gas was set to 2, the desolvation temperature was 300℃, the sheath gas and auxiliary gas were set to 40 and 10 respectively; the capillary voltage was +3.3kV and -3kV in positive and negative ion modes respectively, and the cone voltage was set to 0V; the acquisition mode was parallel reaction monitoring mode (PRM); the mass-to-charge ratio (m / z) acquisition range of the primary mass spectrometer was 60-800, the resolution was 35000, the number of target ions was 5x106, and the maximum ion injection time was 80ms; the secondary mass spectrometer acquisition resolution was 17500, the number of target ions was 1x105, the maximum ion injection time was 50ms, the isolation window was 4.0m / z, and the collision energy was 30.

[0152] (3). PRM targeted map data processing.

[0153] The samples were analyzed according to the above chromatographic mass spectrometry conditions to obtain the original metabolic fingerprints of PRM targeting 16 metabolite markers for all samples. The typical selected ion chromatograms (EIC) are shown in Figure 8 . Figure 8 Among them, 2-Hydroxyphenylacetic acid, N,N-dimethylglycine, N-Acetyl-L-aspartic acid, N-Epsilon-Acetyl-L-Lysine, Azelaic acid, (Z,Z)-4-oxo-2,5-hetpadienedioic acid, and Proline were detected. The targeted metabolome profiles of plasma samples from patients with renal cancer and healthy subjects were processed using Skyline software to generate a two-dimensional matrix of metabolite markers, in which rows represent metabolite markers and columns represent analytical samples. At the same time, the internal standard data of each sample were processed using MS-Dial software to generate a two-dimensional matrix containing internal standard information, in which rows represent internal standards and columns represent analytical samples. The two-dimensional matrix data of the metabolite markers are normalized by the corresponding internal standard content to obtain a two-dimensional matrix representing the quantitative data of the metabolite markers for subsequent machine learning.

[0154] (4).LC-MS experiment quality control.

[0155] During the LC-MS plasma PRM targeted metabolomics analysis, the prepared QC samples were evenly inserted into the samples to be tested in the order of one QC sample after every 15 samples to be tested, to continuously monitor the stability and repeatability of the entire process from sample injection pretreatment to final analysis and detection. The coefficient of variation (%RSD) of each metabolite in the QC sample was further calculated through PRM targeted spectral data preprocessing. The results showed that the coefficient of variation of all plasma metabolite markers was maintained below 30%, which demonstrated excellent quality control in the experimental process and ensured the reliability and validity of the obtained metabolomics data.

[0156] (5). Quantitative validation of parallel reaction monitoring (PRM) targeted metabolomics.

[0157] In order to determine the quantitative effect of parallel reaction monitoring (PRM) targeted metabolomics, the present invention verifies five metabolites (N,N-dimethylglycine, N-acetyl-L-aspartic acid, N-epsilon-acetyl-L-lysine, azelaic acid, and proline) among the 16 metabolites, and the steps are as follows:

[0158] (a). Preparation of standard stock solution and isotope internal standard (IS) stock solution: Dissolve the standard substances N,N-dimethylglycine, N-acetyl-L-aspartic acid, N-ε-acetyl-L-lysine, azelaic acid, proline and their corresponding isotope internal standards (IS) including N,N-dimethylglycine, N-acetyl-L-aspartic acid, N-ε-acetyl-L-lysine, azelaic acid and proline in methanol, respectively. The above concentrations are all 1 mg / ml.

[0159] (b) Prepare mixed standards of corresponding concentration gradients: using the extract containing isotope internal standard (IS) prepared in step (1b), serially dilute the combined standard stock solution to obtain the standard working calibration solution, with the dilution ratio being 1-1 / 64 (1, 1 / 2, 1 / 4, 1 / 8, 3 / 32, 1 / 16, 3 / 64, 1 / 32, 1 / 64).

[0160] (c) Parallel reaction monitoring (PRM) targeted metabolomics analysis: the steps are the same as above (2).

[0161] (d) PRM target map data processing: the steps are the same as above (3).

[0162] (e) PRM quantitative verification: A two-dimensional matrix of markers obtained in step (3) is constructed, in which each row is metabolite information, each column is the analysis sample, and the median is the ratio of the peak area of ​​the corresponding metabolite to the peak area of ​​the internal standard. The standard curve is established using the peak area ratio of the marker to the internal standard in each standard sample in the gradient dilution standard working calibration sample (internal standard method). After fitting, the linear equation and linear correlation coefficient R are obtained. 2 The linear correlation coefficients were all greater than 0.99, which proved that the quantitative effect of this parallel reaction monitoring (PRM) targeted method was excellent. The quantitative concentrations of the corresponding metabolites in all samples could be calculated by the established standard curve. The linear correlation equations and linear ranges of the five diagnostic markers are shown in Figure 2. Figures 9 to 13 shown.

[0163] 2.3 Construction of renal cancer diagnostic model.

[0164] The two-dimensional matrix data of 16 metabolite markers obtained by graph preprocessing were learned using the machine learning support vector machine (SVM) algorithm. 302 patients with renal cancer and 201 healthy controls were used as training sets to build the model, and the modeling parameter C = 5. The training set classification accuracy (Accuracy) of the obtained renal cancer diagnosis model reached 97.48%, the specificity (Specificity) was 95.12% (95% CI, 91.06% to 99.19%), the sensitivity (Sensitivity) was 98.08% (95% CI, 93.41% to 100.00%), the recall rate (Recall) was 0.9808, and the precision (Precision) was 0.9781. It shows that the developed diagnostic model shows high sensitivity, specificity, and accuracy in distinguishing renal cancer patients, confirming the effectiveness of the model in the early diagnosis of renal cancer.

[0165] Next, a renal cancer diagnostic model constructed based on 16 plasma metabolic markers was validated from multiple sources.

[0166] 1. Research subjects.

[0167] This embodiment includes an internal validation cohort, an external validation cohort, a multicenter external validation cohort, and a prospective cohort for testing and validating the established diagnostic model. The internal validation cohort contains 202 participants (101 renal cancer patients and 101 healthy controls), while the independent external validation cohort contains 158 participants (100 renal cancer patients and 58 healthy controls). The multicenter external validation cohort included 346 participants from four hospitals, including 179 renal cancer patients and 167 healthy controls. The prospective validation cohort included 123 participants, 54 of whom were renal cancer patients and 69 were healthy controls. The plasma of patients with renal cancer refers to the preoperative plasma of patients with renal cell carcinoma confirmed by postoperative pathology. Healthy controls refer to healthy people who have no kidney disease after physical examination.

[0168] 2. Experimental method.

[0169] The detection and data preprocessing methods of this embodiment and embodiment 2 are the same.

[0170] 3. Multi-source validation of renal cancer diagnostic model.

[0171] After collecting PRM targeted metabolomics data of four validation cohorts and performing data preprocessing, four quantitative two-dimensional matrices were obtained. These four validation data were further substituted into the renal cancer diagnostic model constructed in Example 2, and the predictive performance is shown in Table 4. The renal cancer diagnostic model constructed by the present invention demonstrated excellent diagnostic efficacy in the four validation cohorts, and showed a high degree of consistency and reliability, with an accuracy greater than 90%, a specificity greater than 89%, a sensitivity greater than 94%, and a recall rate and precision greater than 92%. This study confirmed that the developed renal cancer diagnostic model has excellent diagnostic efficacy in clinical application scenarios including large-scale sample sets and multiple sources, can adapt to different clinical environments, and has the potential to be used in a wide range of clinical fields.

[0172] Table 4 Multi-source validation of the renal cancer diagnostic model established by sixteen metabolic markers

[0173]

[0174] Next, a renal cancer diagnostic model was constructed using 7 plasma metabolic markers and validated from multiple sources.

[0175] 1. Construct a renal cancer diagnostic model using 7 plasma metabolic markers.

[0176] The research objects and experimental analysis methods of Example 2 are the same. Only in the machine learning SVM modeling, the two-dimensional matrix data of seven plasma metabolic markers (2-hydroxyphenylacetic acid, N, N-dimethylglycine, N-acetyl-L-aspartic acid, N-ε-acetyl-L-lysine, Azelaic acid, (Z, Z)-4-oxo-2,5-hetpadienedioic acid, and Proline) are used for machine learning and modeling. The training set classification accuracy, specificity, sensitivity, area under the ROC curve AUC, recall, and precision of the obtained renal cancer diagnosis model are shown in Table 6.

[0177] The scatter plots of the statistical differences of the seven diagnostic markers between renal cancer and normal controls are shown in Figure 2. Figures 14 to 20 As shown, * represents P≤0.05, ** represents P≤0.01, *** represents P≤0.001, **** represents P≤0.0001, and the P value shows that the 7 diagnostic markers are all statistically different. The detailed difference multiples, model VIP weights and P value data are shown in Table 5.

[0178] This confirms that the renal cancer model constructed using seven plasma metabolic markers is still highly effective in the diagnosis of renal cancer, with an AUC of up to 0.9877.

[0179] 2. Multi-source validation of renal cancer diagnostic model.

[0180] This embodiment has the same research subjects and experimental analysis methods as those when constructing a renal cancer diagnostic model using 16 plasma metabolite markers, except that the renal cancer model used is constructed using 7 metabolite markers. The prediction performance of the four validation cohorts is shown in Table 6. The renal cancer diagnostic model constructed by the present invention demonstrated relatively excellent diagnostic efficacy in the four validation cohorts, and showed a high degree of consistency and reliability, with accuracy and AUC both greater than 90%, and sensitivity greater than 89%.

[0181] The ROC curves of the four test sets of the renal cancer diagnosis model constructed using 7 plasma metabolic markers are shown in Figure 2. Fig.21This example demonstrates that the renal cancer diagnostic model built with 7 metabolite markers still has excellent diagnostic efficacy in large-scale sample sets and multi-source clinical application scenarios, can adapt to different clinical sources, and has the potential for wide application in the clinical field in the future.

[0182] Table 5 Differences of seven metabolic markers in renal cancer

[0183]

[0184] Table 6 Multi-source validation of the renal cancer diagnostic model established by seven metabolic markers

[0185]

[0186] Next, a renal cancer diagnostic model was constructed using three plasma metabolic markers and validated from multiple sources.

[0187] 1. Construct a renal cancer diagnostic model using three plasma metabolic markers.

[0188] This embodiment has the same research objects and experimental analysis methods as those in Embodiment 2. Only in step 2.3, when machine learning SVM modeling is performed, two-dimensional matrix data of three plasma metabolic markers (N,N-dimethylglycine, azelaic acid, (Z,Z)-4-oxo-2,5-hetpadienedioic acid) are used for machine learning and modeling. The training set classification accuracy, specificity, sensitivity, area under the ROC curve AUC, recall rate, and precision of the obtained renal cancer diagnosis model are shown in Table 7. This confirms that the renal cancer model constructed using three plasma metabolic markers is still highly effective in renal cancer diagnosis, with an AUC of 0.97092, as shown in Table 7. Figures 26 to 30 .

[0189] 2. Multi-source validation of renal cancer diagnostic model.

[0190] The renal cancer model used in this example is built with only three metabolic markers. The prediction performance of the four validation cohorts is shown in Table 7. The renal cancer diagnostic model constructed by the present invention showed excellent diagnostic efficacy in the four validation cohorts, and showed high consistency and reliability, with AUC greater than 82%, sensitivity greater than 80%, and ROC curves shown in Table 7. Figure 22 to Figure 25This study confirmed that the renal cancer diagnostic model built with three metabolic markers still has excellent diagnostic efficacy in large-scale sample sets and multi-source clinical application scenarios, can adapt to different clinical sources, and has the potential for widespread application in the clinical field in the future.

[0191] Table 7 Multi-source validation of the renal cancer diagnostic model established by three metabolic markers

[0192]

[0193] Use any three or more of the seven plasma metabolic markers to construct a renal cancer diagnostic model and verify it from multiple sources.

[0194] 1. Use any three or more of the seven plasma metabolic markers to construct a renal cancer diagnostic model.

[0195] The research object and experimental analysis method of this embodiment are the same as those of Example 2. Only in step 2.3 of machine learning SVM modeling, two-dimensional matrix data of any three or more of the seven plasma metabolic markers (2-hydroxyphenylacetic acid, N, N-dimethylglycine, N-acetyl-L-aspartic acid, N-epsilon-acetyl-L-lysine, Azelaic acid, (Z, Z)-4-oxo-2,5-heptadienedioic acid, and Proline) are used for machine learning and modeling. The area under the ROC curve AUC of the training set classification of the obtained renal cancer diagnosis model is shown in Table 8. This confirms that the renal cancer model constructed using any three or more of the seven plasma metabolic markers can be used for renal cancer diagnosis.

[0196] 2. Multi-source validation of renal cancer diagnostic model.

[0197] This example has the same research subjects and experimental analysis methods as Example 3, except that the renal cancer model used is constructed by any three or more of the seven plasma metabolite markers. Prediction performance of the four validation cohorts: The area under the receiver operating curve (AUC) is shown in Table 8. It is proved that any three or more of the renal cancer diagnostic metabolite marker combinations constructed by the present invention have excellent diagnostic efficacy in the four validation cohorts.

[0198] Table 8 Area under the receiver operating curve (AUC) of the multi-source validation of the renal cancer diagnostic model established by any three or more of the seven plasma metabolic markers

[0199]

[0200] The present application also provides an application scenario, which applies the above-mentioned plasma metabolic marker screening method. Specifically: The plasma metabolic marker screening method provided in this embodiment can be applied in the early screening scenario of renal cancer. The early screening scenario of renal cancer includes: the renal cancer screening model construction link, the design and production of early renal cancer detection equipment, the collection of plasma samples from the subjects to be tested, and the use of early renal cancer detection equipment. The plasma metabolic marker screening method provided in this embodiment belongs to the plasma marker screening and selection link in the renal cancer screening model construction link.

[0201] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig.31 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store video tag processing data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for screening plasma metabolic markers is implemented.

[0202] Those skilled in the art will understand that Fig.31 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0203] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0204] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0205] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0206] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0207] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0208] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0209] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for screening plasma metabolic markers, characterized in that: include: Acquire a plasma sample set; any plasma sample in the plasma sample set is labeled as a renal cancer patient or a healthy control; Each plasma sample in the plasma sample set is processed separately by using liquid chromatography-mass spectrometry technology to obtain the original metabolic fingerprint corresponding to each plasma sample; Performing first feature extraction processing on the plurality of original metabolic fingerprints respectively to obtain a first feature matrix corresponding to each plasma sample; the first feature matrix is ​​used to describe the metabolic features of the corresponding plasma sample; The support vector machine model is trained with the first feature matrix as input and the label as output to obtain a metabolic feature sequence; the metabolic feature sequence is obtained by arranging all metabolic features in the feature matrix in descending order according to the weights output by the support vector machine model, and intercepting the first preset number of metabolic features; According to the metabolic feature sequence, a plasma metabolic marker set is determined using a greedy algorithm and a support vector machine model.

2. The method for screening plasma metabolic markers according to claim 1, characterized in that: After determining the plasma metabolic marker set according to the metabolic feature sequence using a greedy algorithm and a support vector machine model, the method further includes: Obtaining a second preset quantity; Selecting a second preset number of plasma metabolite markers from the plasma metabolite marker set as target metabolic features; preparing an extract; the extract contains a plurality of isotope internal standards; the isotope internal standards correspond one to one with the target metabolic characteristics; the isotope internal standards are used to absolutely quantify the target metabolic characteristics; Using the extraction liquid to process each plasma sample in the plasma sample set respectively to obtain multiple extraction samples; The parallel reaction monitoring targeted metabolomics quantitative method was used to process multiple extraction samples separately to obtain multiple targeted metabolic fingerprints. Performing second feature extraction processing on multiple targeted original metabolic fingerprints respectively to obtain a second feature matrix corresponding to each plasma sample; the second feature matrix is ​​used to describe the target metabolic features of the corresponding plasma sample; Using the second feature matrix as input and the label as output, training a support vector machine model to obtain a kidney cancer screening model; A computer program product that implements the kidney cancer screening model is embedded in a processor to construct an early detection device for kidney cancer; the input of the early detection device for kidney cancer is a second characteristic matrix to be tested; the second characteristic matrix to be tested is obtained by processing the plasma sample of the subject to be tested using liquid chromatography-mass spectrometry technology to obtain an original metabolic fingerprint map to be tested, and performing a second feature extraction process on the original metabolic fingerprint map to be tested.

3. The method for screening plasma metabolic markers according to claim 2, characterized in that: After the computer program product realizing the renal cancer screening model is embedded in the processor to construct the early detection device for renal cancer, the method further includes: The sample to be tested is processed by using a parallel reaction monitoring targeted metabolomics quantitative method to obtain a targeted metabolic fingerprint to be tested; the sample to be tested is obtained by extracting a plasma sample of a subject to be tested by using an extraction solution; Performing feature extraction processing on the targeted metabolic fingerprint to be tested to construct a second feature matrix to be tested; the second feature matrix to be tested is used to describe the target metabolic features of the corresponding plasma sample; The second characteristic matrix to be tested is input into the early detection device for renal cancer to obtain the early detection result of renal cancer of the subject to be tested.

4. The method for screening plasma metabolic markers according to claim 1, characterized in that: According to the metabolic feature sequence, a set of plasma metabolic markers is determined using a greedy algorithm and a support vector machine model, including: The empty set is constructed as the initial plasma metabolic marker set; Let the number of iterations i = 1; Let the number of target undetermined metabolic features n=1, and determine the first n metabolic features in the metabolic feature sequence as target undetermined metabolic features; Performing feature extraction processing on the plurality of original metabolic fingerprints respectively, selecting parameters corresponding to the target metabolic features to be determined to construct a plurality of feature matrices at the i-th iteration; Constructing a training set at the i-th iteration and a validation set at the i-th iteration according to multiple feature matrices at the i-th iteration; Taking multiple feature matrices in the training set at the i-th iteration as input and the labels as output, training the support vector machine model to obtain the training model at the i-th iteration; The training model at the i-th iteration is verified using the validation set at the i-th iteration to obtain the accuracy of the training model at the i-th iteration; Determine whether the number of iterations i is less than 2, and obtain a first determination result; If the first judgment result is yes, then the i-th metabolic feature in the metabolic feature sequence is added to the initial plasma metabolic marker set, the value of i is increased by 1, and the process returns to the step of "determining the first i metabolic features in the metabolic feature sequence as target metabolic features to be determined"; If the first judgment result is no, then judging whether the accuracy of the training model at the i-th iteration is greater than the accuracy of the training model at the i-1-th iteration, and obtaining a second judgment result; If the second judgment result is yes, then the i-th metabolic feature in the metabolic feature sequence is added to the initial plasma metabolic marker set, the value of i is increased by 1, the value of n is increased by 1, and the process returns to the step of "determining the first n metabolic features in the metabolic feature sequence as target metabolic features to be determined"; If the second judgment result is no, the initial plasma metabolite marker set at the i-1th iteration is determined as the plasma metabolite marker set, the value of i is increased by 1, and the step of "determining the first n metabolic features in the metabolic feature sequence as the target metabolic features to be determined" is returned.

5. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the plasma metabolic marker screening method according to any one of claims 1 to 4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the plasma metabolic marker screening method according to any one of claims 1 to 4 is implemented.

7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the plasma metabolic marker screening method according to any one of claims 1 to 4 is implemented.

Citation Information

Patent Citations

  • Ensemble classification method based on randomized greedy feature selection

    CN106991296A

  • Lung cancer early diagnostic marker based on metabonomics and artificial intelligence technology and application thereof

    CN109884302A