COPD-ph specific metabolite marker combinations based on metabolomics and machine learning screening

Abstract

The application discloses a chronic obstructive pulmonary disease related pulmonary hypertension (COPD-PH) specific metabolic marker combination based on metabolomics and machine learning screening, and relates to the technical field of biological medicine.The application first discovers specific metabolic characteristics of COPD-PH patients.Plasma samples of COPD-PH patients, COPD patients and healthy controls are subjected to non-targeted metabolomics analysis, and 163 differential metabolites are identified, then three machine learning algorithms, RF, SVM and LASSO, are comprehensively analyzed, and five key metabolites, ornithine, histidine, biotin, nicotinamide and homocysteine, are screened out.In a multi-center verification study, the five metabolite combination also exhibits excellent diagnostic efficiency [AUC = 0.814], which is significantly better than traditional diagnostic indexes.The application not only provides a new tool for non-invasive diagnosis of COPD-PH, but also further reveals the metabolic mechanism of disease occurrence and development.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of biological medicine, and particularly relates to a COPD-PH specific metabolic marker combination based on metabolomics and machine learning screening and an application thereof. BACKGROUND

[0002] Chronic obstructive pulmonary disease-related pulmonary arterial hypertension (COPD-PH) is a severe complication in the progression of COPD, but its diagnosis faces significant challenges. The current clinical practice has the following key problems:

[0003] Symptom overlap and high misdiagnosis rate: The clinical presentation of COPD and PH (e.g., dyspnea, decreased exercise tolerance) is highly similar, leading to a misdiagnosis rate of over 30% (Braganza M, Shaw J, Solverson K, Vis D, Janovcik J, Varughese RA, Thakrar MV, Hirani N, Helmersen D, Weatherald J: A Prospective Evaluation of the Diagnostic Accuracy of the Physical Examination for Pulmonary Hypertension. Chest 2019, 155(5):982-990.; Dauriat G, Reynaud-Gaubert M, Cottin V, Lamia B, Montani D, Canuet M, Boissin C, Tromeur C, Chaouat A, Degano B et al: Severe pulmonary hypertension associated with chronic obstructive pulmonary disease: A prospective French multicenter cohort. The Journal of heart and lung transplantation : the official publication of the International Society for Heart Transplantation 2021, 40(9): 1009-1018.). Existing diagnosis relies on the “Suspect-Support-Confirm-Stratify” strategy recommended by the 6th World Symposium on Pulmonary Hypertension (WSPH), but this process is complex and requires multiple modalities including chest CT, echocardiography, and pulmonary function (Galiè N, McLaughlin VV, Rubin LJ, Simonneau G: An overview of the 6th World Symposium on Pulmonary Hypertension. Eur Respir J 2019, 53(1).).

[0004] Limitations of the exclusive diagnostic approach: In the setting of chronic lung disease, the etiology of pulmonary hypertension is often complex, with coexisting risk factors such as HIV infection, connective tissue disease, portal hypertension, or chronic thromboembolic pulmonary hypertension, which makes the exclusive diagnostic approach currently used for COPD-PH difficult to perform in primary care settings. This leads to frequent referral of patients to specialized centers (Kerr KM, Elliott CG, Chin K, Benza RL, Channick RN, Davis RD, He F, LaCroix A, Madani MM, McLaughlin VV et al: Results From the United States Chronic Thromboembolic Pulmonary Hypertension Registry: Enrollment Characteristics and 1-Year Follow-up. Chest 2021, 160(5): 1822-1831.; Humbert M, Kovacs G, Hoeper MM, Badagliacca R, Berger RMF, Brida M, Carlsen J, Coats AJS, Escribano-Subias P, Ferrari P et al: 2022 ESC / ERS Guidelines for the diagnosis and treatment of pulmonary hypertension. Eur Respir J 2023, 61(1).).

[0005] The gold standard invasive defect: right heart catheterization (RHC) is the gold standard for diagnosis, but it is not suitable for routine screening of COPD patients due to high operation risk and strict technical threshold. In addition, although existing imaging markers (such as right ventricular diameter / left ventricular diameter ratio, pulmonary artery stiffness parameters) have potential, they rely on expensive equipment (such as cardiovascular magnetic resonance imaging) and lack standardization (Charters PFP, Rossdale J, Brown W, Burnett TA, Komber H, Thompson C, Robinson G, MacKenzie Ross R, Suntharalingam J, Rodrigues JCL: Diagnostic accuracy of an automated artificial intelligence derived right ventricular to left ventricular diameter ratio tool on CT pulmonary angiography to predict pulmonary hypertension at right heart catheterisation. Clinical radiology 2022, 77(7):e500-e508.; Agoston-Coldea L, Lupu S, Mocan T: Pulmonary Artery Stiffness by Cardiac Magnetic Resonance Imaging Predicts Major Adverse Cardiovascular Events in patients with Chronic Obstructive Pulmonary Disease. Sci Rep 2018, 8(1): 14447.). The existing diagnostic process is complex and time-consuming, and the current COPD-PH diagnosis requires a combination of multiple clinical symptoms and examination results. This complex evaluation process not only increases medical costs, but also may lead to delayed diagnosis and missed best treatment opportunity.

[0006] Limitations of existing technology: single marker and low specificity: previous studies have focused on a single metabolite (such as B-type natriuretic peptide), but the heterogeneity of COPD-PH requires multiple indicators for joint evaluation. Lack of non-invasive dynamic monitoring tools: imaging or hemodynamic examination cannot meet the needs of early screening and long-term follow-up of the disease.

[0007] Research gap of metabolic disorder and pulmonary vascular remodeling: Although the role of metabolic abnormalities in pulmonary arterial hypertension (PAH) has been widely studied, there is almost no research on the metabolic characteristics of COPD-PH, and there are few reports on specific metabolic markers of COPD-PH. SUMMARY

[0008] In view of the defects in the prior art, the present application first discovers the specific metabolic characteristics of COPD-PH patients through systematic metabolomics research. Non-targeted metabolomics analysis was performed on 85 plasma samples of COPD-PH patients, 99 plasma samples of COPD patients and 189 plasma samples of healthy controls, and 163 differential metabolites were identified. Then, through the comprehensive analysis of RF, SVM and LASSO three machine learning algorithms, five key metabolites were finally screened out: ornithine, histidine, biotin, nicotinamide and homocysteine. At the same time, in the multicenter verification research (newly added 116 samples), the combination of the five metabolites also showed excellent diagnostic efficiency (AUC=0.814), which was significantly better than the traditional diagnostic indicators. This discovery not only provides a new tool for the non-invasive diagnosis of COPD-PH, but also further reveals the metabolic mechanism of disease occurrence and development.

[0009] In order to achieve the above purpose, the present application adopts the following technical solutions:

[0010] One of the purposes of the present application is to provide a metabolic marker combination, which comprises ornithine, histidine, biotin, nicotinamide and homocysteine.

[0011] The second purpose of the present application is to provide the application of the metabolic marker combination in the preparation of COPD-PH disease diagnosis products, early screening products, disease monitoring products and / or drug efficacy monitoring products.

[0012] The third purpose of the present application is to provide a screening method of the metabolic marker combination, which comprises two stages: a retrospective modeling stage and a prospective verification stage; the retrospective modeling stage is based on the peak intensity characteristics of metabolites to construct a prediction model; the prospective verification stage reconstructs a clinical practical model through targeted metabolomics quantitative data.

[0013] Further, the research objects of the retrospective modeling stage and the prospective verification stage both include the COPD-PH group, the COPD group and the normal group.

[0014] Further, the acquisition of the peak intensity characteristics of metabolites in the retrospective modeling stage includes the use of non-targeted and targeted quantitative metabolomics analysis methods.

[0015] Further, the retrospective modeling stage first constructs a multi-dimensional feature screening system through three machine learning algorithms, random forest (RF), support vector machine (SVM) and least absolute shrinkage and selection operator (LASSO), and then obtains five candidate metabolites, ornithine, histidine, biotin, nicotinamide and homocysteine, through cross-validation.

[0016] The fourth object of the present application is to provide a product for the diagnosis, early screening, disease monitoring and / or drug efficacy monitoring of COPD-PH, which comprises reagents for detecting the combination of metabolic markers.

[0017] Further, the product form includes a kit.

[0018] The fifth object of the present application is to provide a model for the diagnosis, early screening, disease monitoring and / or drug efficacy monitoring of COPD-PH, wherein the construction method comprises using the combination of metabolic markers.

[0019] The sixth object of the present application is to provide a system for the diagnosis, early screening, disease monitoring and / or drug efficacy monitoring of COPD-PH, wherein the system comprises a module using the combination of metabolic markers.

[0020] Compared with the prior art, the present application has the following technical effects:

[0021] I. Beneficial effects

[0022] 1. Improved diagnostic performance

[0023] High precision: the five metabolic markers (ornithine, histidine, biotin, nicotinamide and homocysteine) screened based on multi-model machine learning algorithms have a combined diagnostic AUC of 0.879 (sensitivity 72.9%, specificity 91.9%) in a retrospective cohort, and the prospective verification AUC remains 0.814 (sensitivity 60.0%, specificity 83.9%), which is significantly better than traditional echocardiography or clinical scoring (such as B-type natriuretic peptide detection recommended by ESC / ERS guidelines).

[0024] Stability verification: the model cross-center applicability is verified by a double cohort design (retrospective n=85 COPD-PH / 99 COPD / 189 healthy; prospective n=50 COPD-PH / 35 COPD / 31 healthy), reducing single-center bias.

[0025] 2. Methodological optimization

[0026] Multi-algorithm joint screening: integrating RF, SVM and LASSO three machine learning algorithms, avoiding single model overfitting.

[0027] Metabolomics technology innovation: combination of non-targeted metabolomics [high-resolution liquid chromatography-mass spectrometry (LC-MS / MS)] global screening and targeted metabolomics (MRM mode) accurate quantification, detection sensitivity up to 0.1 ng / mL (R 2 >0.99), intra- / inter-batch CV <10%, meeting the clinical detection standard.

[0028] Cross-queue two-stage verification system: retrospective cohort (n=373) and prospective multicenter cohort (n=116) collaborative design, combined with the complementary nature of non-targeted metabolomics (global screening) and targeted metabolomics (accurate quantification), to ensure the universality and clinical transformation potential of the research conclusion. The prospective verification stage adopts double screening criteria (trend consistency test + Kruskal-Wallis multiple comparison) to lock 5 core markers (ornithine, histidine, biotin, nicotinamide, homocysteine), and the biological stability is fully verified by cross-platform (LC-MS vs. MRM) and cross-queue verification.

[0029] 3. Clinical practicability

[0030] The present application aims at the clinical pain point of the lack of non-invasive diagnosis method for COPD-PH, and for the first time proposes a metabolomics non-invasive screening system, breaking through the limitations of existing technology relying on invasive examination, and its core value and innovative application scenarios are as follows:

[0031] Simplify the detection process: the risk score formula only needs 5 kinds of metabolite concentrations (ornithine, histidine, etc.), and the calculation is simple, and the detection items are compatible with the conventional clinical laboratory platform. The current clinical diagnosis of COPD-PH relies on right heart catheterization (invasive, high cost, difficult to popularize in primary hospitals), while the present application only needs blood detection combined with conventional ultrasound to complete the high-risk population screening, especially suitable for:

[0032] Community early screening: in medical institutions lacking right heart catheterization equipment, the combination of "ultrasound + metabolomics Panel" realizes the initial screening of high-risk population, and reduces the missed diagnosis rate.

[0033] Efficacy monitoring: histidine (reflecting the inflammatory state) and homocysteine (oxidative stress marker) in the model can be used as biomarkers for drug (such as phosphodiesterase-5 inhibitor) responsiveness.

[0034] Fill in the blank of the diagnosis method: the current WHO guidelines lack specific non-invasive means for the diagnosis of COPD-PH, and the error rate of echocardiography in assessing pulmonary artery systolic pressure (PASP) in patients with emphysema is as high as 30%. The present application fills in the gap of the limitations of imaging through metabolomics characteristics.

[0035] II. Innovation

[0036] 1. Scientific discovery innovation

[0037] Metabolic pathway mechanism: First time to reveal the core role of urea cycle-nicotinamide metabolic axis in COPD-PH:

[0038] Ornithine (urea cycle) abnormalities lead to blood ammonia accumulation, driving pulmonary vascular endothelial injury;

[0039] Nicotinamide (NAD+ metabolism) depletion exacerbates mitochondrial dysfunction, promoting pulmonary artery smooth muscle proliferation.

[0040] Marker synergistic effect: Found the synergistic diagnostic value of homocysteine (oxidative stress) and histidine (inflammation regulation).

[0041] 2. Technical method innovation

[0042] Two-stage modeling strategy:

[0043] Retrospective stage: Construct high-sensitivity model based on peak intensity (research use);

[0044] Prospective stage: Convert categorical variables to establish clinically applicable models, addressing the clinical translation bottleneck of metabolomics data.

[0045] Dynamic threshold design: Set personalized cutoff values for different metabolites (e.g., ornithine 34,000, homocysteine 350) to optimize diagnostic specificity.

[0046] 3. Research design innovation

[0047] Multi-center cohort verification: Prospective cohort covers Guangzhou Medical University and Shanghai Pulmonary Hospital, covering Chinese populations in north and south, enhancing the universality of results.

[0048] Three, clinical application value

[0049] 1. Early diagnosis

[0050] Non-invasive alternative: Compared with right heart catheterization (gold standard), plasma metabolic marker detection can achieve non-invasive, low-cost (cost reduction of about 60%) screening, suitable for promotion in primary hospitals.

[0051] High-risk population stratification: Metabolic screening of COPD patients to early warn PH risk.

[0052] 2. Precision treatment guidance

[0053] Targeted intervention targets: Model markers point to intervenable pathways (such as supplementing nicotinamide to improve NAD+ levels and regulating urea cycle to reduce ammonia toxicity), providing the basis for individualized treatment.

[0054] Efficacy monitoring: Dynamic detection of ornithine / homocysteine ratio changes to evaluate the response rate of vasodilator drugs (such as phosphodiesterase 5 inhibitors) (preliminary data show AUC 0.72).

[0055] 3. Health economic benefits

[0056] Reducing unnecessary right heart catheterization: Model negative predictive value up to 89%, avoiding about 40% of invasive examinations.

[0057] Optimizing health insurance expenditure: The cost per test is significantly lower than right heart catheterization.

[0058] IV. Conclusion

[0059] The present application establishes the first COPD-PH non-invasive diagnostic system based on metabolomics through multi-model machine learning algorithm screening, double cohort clinical verification and deep analysis of metabolic mechanisms. Its innovation lies in:

[0060] Methodology: Multi-algorithm combined to improve marker screening efficiency;

[0061] Scientific discovery: Revealing the core role of the urea cycle-energy metabolism axis;

[0062] Clinical translation: Providing a simple, low-cost, and generalizable detection scheme.

[0063] Future multi-center randomized controlled trials can further verify its value in improving clinical decision-making.

[0064] In summary, the present application first systematically describes the metabolic characteristics of COPD-PH; establishes an efficient non-invasive diagnostic model; discovers new potential therapeutic targets; and solves the problems of existing technology such as invasiveness, high cost, and complexity. The developed five-marker combination has high sample stability and low detection cost, providing a generalizable screening tool for primary medical institutions. Not only does it fill the gap in the field of metabolic diagnosis of COPD-PH, but it also opens up new avenues for early diagnosis and precision treatment of COPD-PH, with important clinical application value. BRIEF DESCRIPTION OF DRAWINGS

[0065] Figure 1 The process and results of metabolic difference analysis of COPD-PH, COPD and normal groups in Example 1 of the present application;

[0066] Figure 2 The process of constructing a multi-dimensional feature screening system using three machine learning algorithms with complementary characteristics in Example 1 of the present application;

[0067] Figure 3 The peak expression characteristics of 5 metabolites in the retrospective cohort and the quantitative detection results in the prospective cohort in Example 1 of the present application;

[0068] Figure 4 This refers to the diagnostic performance of the predictive model in the retrospective modeling stage and the predictive efficacy of the clinically applicable model in the prospective validation stage of Embodiment 1 of the present invention.

[0069] Figure 5 The ROC analysis results for predicting COPD-PH individually for each metabolite in the retrospective and prospective cohorts of Example 1 of this invention are shown. Detailed Implementation

[0070] The following examples are used to illustrate the present invention, but are not intended to limit the scope of the invention. Any modifications or substitutions made to the methods, steps, or conditions of the present invention without departing from the spirit and essence of the invention are within the scope of the invention. The reagents, products, and instruments used in the following examples are all commercially available, and the methods used in the examples, unless otherwise specified, are consistent with conventional methods.

[0071] The technical solution of the present invention will be further described in detail below with reference to the embodiments.

[0072] Example 1

[0073] 1. Retrospective cohort study subjects

[0074] 1.1 COPD-PH group:

[0075] From December 2008 to December 2024, the First Affiliated Hospital of Guangzhou Medical University enrolled 250 patients diagnosed with COPD and pulmonary hypertension (PH) by right heart catheterization. The inclusion criteria for the COPD-PH group included: (I) age ≥ 18 years; (II) meeting the Global Initiative for Chronic Obstructive Lung Disease (GOLD) pulmonary function diagnostic criteria, namely, forced expiratory volume in one second (FEV1) to forced vital capacity (FVC) ratio < 70% after bronchodilator test, and having corresponding clinical manifestations and medical history; (III) chest X-ray or chest CT showing signs of lung parenchymal lesions; (IV) the hemodynamic diagnostic criteria for pulmonary hypertension were mean pulmonary artery pressure (mPAP) > 20 mmHg, pulmonary artery wedge pressure (PAWP) ≤ 15 mmHg, and pulmonary vascular resistance (PVR) ≥ 2 Wood units; (V) patients met the WHO category 3 pulmonary hypertension COPD-related subtype (group 3.1).

[0076] Exclusion criteria were: (I) lack of right heart catheterization data; (II) combined congenital heart disease or left heart system disease; (III) history of pulmonary diseases such as pulmonary embolism, interstitial lung disease, pulmonary fibrosis combined with emphysema, active tuberculosis, or severe bronchiectasis; (IV) combined renal insufficiency, hematological disease, rheumatic immune disease, immune deficiency, history of hyperthyroidism, connective tissue disease (such as systemic sclerosis / scleroderma, systemic lupus erythematosus, etc.), sarcoidosis, human immunodeficiency virus infection, mediastinal fibrosis, and other WHO Class 1, 2, 4, 5 pulmonary hypertension related diseases; (IV) malignancy.

[0077] 1.2 COPD group:

[0078] A total of 357 COPD patients with pulmonary arterial systolic pressure (PASP) < 35 mmHg by echocardiography or mPAP < 20 mmHg by right heart catheterization were included in the First Affiliated Hospital of Guangzhou Medical University from October 2009 to September 2023. The inclusion criteria for the COPD group included: (I) age ≥ 18 years; (II) meeting the GOLD pulmonary function diagnostic criteria, i.e. FEV1 / FVC < 70%, with corresponding clinical manifestations and medical history; (III) chest X-ray or chest CT showing lung parenchymal lesion signs; (IV) hemodynamic diagnostic criteria for pulmonary hypertension as mPAP < 20 mmHg; or echocardiography showing pulmonary arterial systolic pressure (PASP) < 35 mmHg or no tricuspid regurgitation.

[0079] Exclusion criteria were: (I) severe cardiovascular disease; (II) history of pulmonary diseases such as pulmonary embolism, interstitial lung disease, pulmonary fibrosis combined with emphysema, active tuberculosis, or severe bronchiectasis; (III) presence of acute pulmonary infection within 4 weeks; (IV) combined renal insufficiency, hematological disease, rheumatic immune disease, immune deficiency, history of hyperthyroidism, connective tissue disease (such as systemic sclerosis / scleroderma, systemic lupus erythematosus, etc.), sarcoidosis, human immunodeficiency virus infection, mediastinal fibrosis, and other diseases; (IV) malignancy.

[0080] 1.3 Normal group: 189 gender and age matched healthy subjects were recruited.

[0081] The clinical research protocol has been approved by the Medical Ethics Committee of the First Affiliated Hospital of Guangzhou Medical University (Medical Research Ethics Review 2021 No. 09).

[0082] 2. Prospective cohort study subjects

[0083] From October 2023 to November 2024, 50 COPD patients diagnosed as PH by right heart catheterization were included in the First Affiliated Hospital of Guangzhou Medical University and Shanghai Pulmonary Hospital. From July 2024 to November 2024, 35 COPD patients with PASP < 35 mmHg or mPAP < 20 mmHg by echocardiography or right heart catheterization were included. From September 2024 to November 2024, 31 gender and age matched healthy subjects were recruited. The inclusion and exclusion criteria of each group were consistent with the previous retrospective cohort study.

[0084] The clinical research protocol has been approved by the Medical Ethics Committee of the First Affiliated Hospital of Guangzhou Medical University (Ethical Number: ES-2024-043-01).

[0085] 3. Untargeted and targeted metabolomics

[0086] Untargeted metabolomics uses LC-MS / MS to comprehensively and unbiasedly screen small molecule metabolites in biological samples of plasma. First, proteins are removed by methanol or acetonitrile precipitation method and metabolites are extracted. Reverse phase chromatographic column (such as Waters ACQUITY UPLC HSS T3, 1.8 μm, 2.1 x 100 mm) is used for separation with mobile phase of water containing 0.1% formic acid (A phase) and acetonitrile (B phase), gradient elution program (such as 2-95% B phase, 20 minutes) and combined with electrospray ionization (ESI) positive and negative ion mode full scan (m / z 50-1500 Da) and data dependent acquisition (DDA) to obtain metabolite mass spectrum information. After peak extraction, alignment and normalization processing by Progenesis QI software, quality control (QC) samples are prepared by mixing the same amount of plasma in each sample and using the same program as the test sample to extract metabolites. Differential metabolites are screened by principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) (VIP≥1 and P<0.05), and metabolite annotation is based on databases such as HMDB. Metabolic pathway enrichment analysis is performed using MetaboAnalyst (http: / / www.metaboanalyst.ca / ) to elucidate the potential pathological mechanism. Targeted metabolomics targets key metabolites screened by untargeted screening and uses multiple reaction monitoring (MRM) mode for accurate quantification. By optimizing chromatographic conditions (such as the retention time of specific metabolites) and mass spectrometry parameters (such as collision energy), isotopic internal standards are used for correction and standard curves (0.1-1000 ng / mL, R 2>0.99), validate methodological parameters (recovery rate 85-115%, intra-batch / inter-batch CV <10%). The two methods complement each other; non-targeted metabolomics provides a global metabolic profile, while targeted metabolomics ensures quantitative accuracy and clinical translational value.

[0087] 4. Statistical Analysis

[0088] Continuous variables: Normally distributed data are expressed as mean ± standard deviation (Mean ± SD), and comparisons between groups are performed using one-way ANOVA or independent samples t-test; non-normally distributed data are expressed as median (interquartile range) [M(IQR)], and comparisons between groups are performed using Mann-Whitney U test (two groups) or Kruskal-Wallis test (multiple groups). Categorical variables: Expressed as frequency (percentage) [n(%)], and comparisons between groups are performed using chi-square test (χ²). 2 The expected frequency was <5 or Fisher's exact test. Metabolite expression levels among the three groups were assessed using the Kruskal-Wallis test. AUC analysis was performed using the pROC software package (version 1.16.2). Spearman correlation analysis was used to determine the correlation coefficients between relevant clinical indicators and major differentially expressed metabolites. All statistical analyses were performed using SPSS statistical software (version 26.0; IBM, Armonk, NY, USA) and R software (version 4.2.2; R Foundation for Statistical Computation, Vienna, Austria). P < 0.05 was considered statistically significant.

[0089] 5. Results

[0090] 5.1 Plasma metabolic landscape of COPD-PH, COPD and normal groups

[0091] To obtain a comprehensive metabolomics landscape from the normal group to COPD and COPD-PH, we retrospectively collected plasma samples from 85 COPD-PH patients, 99 COPD patients, and 189 sex- and age-matched healthy controls for large-scale LC-MS-based metabolomics analysis. In addition, we prospectively collected another independent cohort (COPD-PH n=50, COPD n=35, and normal group n=31) for targeted metabolomics analysis.

[0092] We detected 1619 metabolites in plasma samples, covering a wide range of biochemical components, including amino acids, organic acids, bile acids, nucleotides, acylcarnitines, coenzymes and vitamins, carbohydrates, purines and pyrimidines, and other classes of metabolites. PCA showed differences in metabolic characteristics between COPD-PH, COPD, and normal control samples. Figure 1 a). Partial PLS-DA clearly distinguishes the metabolic characteristics of COPD-PH, COPD, and normal groups (Figure 1 b) The partial overlap of metabolic signatures observed in COPD-PH and COPD groups by PCA and PLS-DA analysis strongly suggests a progressive metabolic evolution between these two disease states. This continuous but incomplete separation pattern reflects that the progression from simple COPD to COPD-PH is not a sudden pathological transition, but a dynamic process of gradual metabolic network remodeling. The overlapping region might represent: 1) a high-risk subgroup of COPD patients who already have pre-PH metabolic signatures; 2) the heterogeneity of disease progression speed in different individuals; 3) a pathophysiological turning point triggered after certain metabolite concentrations reach a threshold.

[0093] 5.2 Dynamic evolution of systemic metabolic reprogramming in COPD progression to COPD-PH

[0094] 5.2.1 Comparison of metabolic signatures between healthy controls and COPD-PH groups

[0095] Metabolic reprogramming is a core mechanism of COPD-PH development. PCA showed that the plasma metabolomic signature of COPD-PH patients was significantly separated from that of healthy controls, and the two groups formed independent cluster distributions in the reduced dimension space, while PLS-DA further identified the core metabolic features driving this differentiation. Differential metabolite analysis (Wilcoxon signed rank test, P < 0.05, FC > 1.2 or < 0.83) identified a total of 646 significantly changed metabolites, among which characteristic metabolites such as ethylenediamine-N,N'-diacetic acid, aspartate-serine complex, and ADP-glucose showed the most significant changes in abundance. Functional enrichment analysis showed that these differential metabolites were mainly involved in purine metabolism, Warburg effect, and various metabolic pathways of arginine and proline, urea cycle, etc. Pathway topology analysis further confirmed the significant activation of the alanine-aspartate-glutamate metabolic axis and the phosphotriose pathway, suggesting that the synergistic effect of energy metabolism abnormalities (such as enhanced glycolysis and inhibited oxidative phosphorylation) and oxidative stress (urea cycle disorder) might drive the progression of pulmonary vascular remodeling and pulmonary arterial hypertension.

[0096] 5.2.2 Metabolic difference characteristics between healthy controls and COPD groups

[0097] Metabolic perturbation was significantly featured in the development of COPD, and COPD patients also showed metabolic profile features that were significantly distinguished from healthy controls. PCA model showed that the distribution of COPD group samples in the metabolic space was significantly deviated from healthy controls, and PLS-DA model accurately identified the key metabolic markers that distinguished the two groups. Differential analysis identified 634 significantly changed metabolites, including characteristic molecules such as N5-modified ornithine (possibly related to nitric oxide metabolism) and acetylcarnitine (a marker of mitochondrial fatty acid beta oxidation). Notably, these metabolic changes were mainly enriched in the arginine-proline metabolic network and the pentose phosphate pathway, showing a metabolic disorder pattern that partially overlapped with the COPD-PH group but was less severe. The disorder of the arginine / proline metabolic pathway and the abnormal urea cycle suggested that nitric oxide synthesis disorders may exacerbate airway remodeling. In addition, abnormalities in nicotinamide metabolism and phosphatidylcholine biosynthesis may reflect compensatory regulation of mitochondrial function and impaired cell membrane stability, providing potential targets for early intervention.

[0098] 5.2.3 Metabolic progression features of COPD and COPD-PH groups

[0099] The progression from COPD to COPD-PH was accompanied by a significant expansion of metabolic disorders. PCA and PLS-DA showed that the metabolic profiles of COPD-PH and COPD groups partially overlapped, but there was a clear boundary, indicating that disease progression was accompanied by continuous changes in metabolic status and specific deterioration. A total of 499 differential metabolites were statistically different in COPD-PH compared to COPD, with abnormalities in inosine-5'-tetraphosphate and cysteine-glycine complex suggesting purine metabolism imbalance and redox homeostasis collapse. Pathway analysis showed that these metabolic changes were mainly related to the urea cycle and nicotinamide metabolism, reflecting the pathological features of ammonia metabolism disorder and energy crisis exacerbation during disease progression.

[0100] 5.2.4 Intersection of common differential metabolites reveals core metabolic pathway disorders

[0101] We took the intersection of differential metabolites between healthy controls and COPD-PH groups, healthy controls and COPD groups, and COPD and COPD-PH groups to find common differential metabolites among the three groups, and there were 163 common differential metabolites ( Figure 1c) This includes 19 organic acids and their derivatives, 11 amino acids, 11 amino acid derivatives, 6 nucleotides and their metabolites, 4 sugars, 4 sugar phosphates, 3 benzene and its derivatives, 2 phenols, 2 coenzymes and vitamins, 2 alcohols and polyols, 2 purines and pyrimidines, 2 fatty acids, 2 sugar alcohols, 3 sugar acids, 2 indoles and their derivatives, 4 phosphoric acid compounds, 1 pyridine and its derivatives, 1 imidazole, 9 other categories. Enrichment analysis shows that the differential metabolites are significantly enriched in the pentose phosphate pathway, histidine metabolism, methyl group histidine, glycine and serine metabolism, arginine and proline metabolism, urea cycle, methionine metabolism and alanine metabolism, etc. Figure 1 d) Pathway enrichment analysis shows that the metabolic pathways of pentose phosphate pathway, histidine metabolism, interconversion of pentose and glucuronate, arginine and proline metabolism, one folate carbon pool, glycine, serine and threonine metabolism, purine metabolism, B-alanine metabolism, amino sugar and nucleotide sugar metabolism, etc. are significantly enriched Figure 1 e).

[0102] In the progression of COPD to COPD-PH, metabolic disorders show the dual characteristics of stepwise accumulation and qualitative deterioration. The number of differential metabolites increases with disease staging (634 in health→COPD, 646 in health→COPD-PH, and 499 in COPD→COPD-PH), suggesting irreversible accumulation of metabolic disorders. From health to COPD, and then to COPD-PH, the pentose phosphate pathway, arginine / proline metabolism, and urea cycle are continuously disturbed, showing enhanced oxidative stress (increased NADPH demand), impaired nitric oxide synthesis, and gradually accumulated ammonia toxicity. In the COPD stage, these pathways have shown significant disturbance, but after entering COPD-PH, ammonia metabolism collapses and energy crisis worsens exponentially: complete disorder of the urea cycle leads to a surge in blood ammonia, abnormal nicotinamide metabolism (NAD+ depletion) aggravates mitochondrial dysfunction, and the Warburg effect (aerobic glycolysis dominated) and imbalance of purine metabolism (such as inosine-5'-tetraphosphate accumulation) together exacerbate ATP depletion. This vicious cycle drives pulmonary vascular oxidative damage, smooth muscle proliferation, and irreversible remodeling, ultimately establishing the pathological process of COPD-PH. In addition, the 163 differential metabolites common to the three groups (such as aspartate-serine complex, N5-ornithine) and their enriched pathways (such as histidine metabolism, folate cycle) further confirm that the ammonia toxicity-energy metabolism axis is the core driving force of disease evolution.

[0103] 5.3 COPD-PH metabolic marker discovery and double cohort validation based on multi-model machine learning

[0104] The present application adopts a multi-stage machine learning combined with metabolomics strategy to systematically screen and verify the key metabolic markers of COPD-PH. First, three machine learning algorithms with complementary characteristics (RF, SVM and LASSO) are used to construct a multi-dimensional feature screening system. Figure 2 After cross-validation, five candidate metabolites are obtained: ornithine, histidine, biotin, nicotinamide and homocysteine. This multi-model joint screening method effectively reduces the selection bias of a single algorithm.

[0105] To enhance the clinical translation value of the research conclusion, a multi-center prospective verification cohort (COPD-PH group n=50, COPD group n=35, healthy control group n=31) is specially designed, and target metabolomics technology is used to accurately quantify the five metabolites. These metabolites have important regulatory roles in pathological and physiological pathways such as purine metabolism and one-carbon unit metabolism, and may have potential correlation with the oxidative stress mechanism of COPD-PH.

[0106] Notably, the peak expression characteristics of the five metabolites in the retrospective cohort ( Figure 3 a-e) and the quantitative detection results in the prospective cohort ( Figure 3 f-j) exhibit significant biological consistency. In the retrospective cohort, the AUC of each metabolite individually predicted ranges from 0.709 (ornithine) to 0.793 (homocysteine) ( Figure 5 a-e), among which homocysteine shows the best discrimination performance. The target quantitative analysis of the prospective cohort further determines the optimal diagnostic cutoffs for each metabolite (ornithine: 34,000; histidine: 2,600; biotin: 430; nicotinamide: 92; homocysteine: 350; Figure S5). Notably, when these continuous variables are converted into categorical variables, their diagnostic performance (AUC 0.562-0.686, Figure 5 f-j) decreases, but remains statistically significant, indicating that the absolute concentration changes of metabolites may have more direct clinical significance in disease diagnosis.

[0107] The present application adopts a two-stage research design: in the retrospective modeling phase, the prediction model constructed based on the peak intensity characteristics of metabolites exhibits excellent diagnostic performance (AUC=0.879, sensitivity 72.9%, specificity 91.9%) ( Figure 4 a); in the prospective verification phase, the clinically practical model (risk score = 0.581 x ornithine + 1.252 x histidine + 1.299 x homocysteine - 0.651 x nicotinamide + 1.414 x biotin - 1.176) reconstructed by target metabolomics quantitative data maintains robust prediction performance (AUC=0.814, sensitivity 60.0%, specificity 83.9%) in the prospective cohort.Figure 4 b) In particular, it is worth emphasizing that the metabolites involved in the model are related to several key pathophysiological processes such as urea cycle (ornithine), inflammation regulation (histidine), energy metabolism (nicotinamide) and oxidative stress (homocysteine), which provides new research directions for in-depth understanding of the molecular mechanisms of COPD-PH.

[0108] The above-described embodiments are only to describe the preferred modes of the present application, and are not intended to limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements to the technical solutions of the present application made by those skilled in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. The use of a detection reagent of a metabolic marker combination in the preparation of a COPD-PH disease diagnosis product and / or an early screening product, characterized in that, The metabolic marker combination consists of ornithine, histidine, biotin, nicotinamide, and homocysteine.

2. Use according to claim 1, characterized in that, In the application, the screening method for the combination of metabolic biomarkers includes two stages: a retrospective modeling stage and a prospective validation stage; the retrospective modeling stage constructs a predictive model based on the peak intensity characteristics of metabolites. The prospective validation phase reconstructs clinically applicable models using targeted metabolomics quantitative data.

3. Use according to claim 2, characterized in that, The subjects in both the retrospective modeling phase and the prospective validation phase included the COPD-PH group, the COPD group, and the normal group.

4. The application according to claim 3, characterized in that, The acquisition of peak metabolite intensity characteristics during the retrospective modeling phase includes the use of non-targeted and targeted quantitative metabolomics analysis methods.

5. The application according to claim 4, characterized in that, The retrospective modeling phase first constructs a multidimensional feature screening system using three machine learning algorithms: Random Forest (RF), Support Vector Machine (SVM), and Least Absolute Shrinkage and Selection Operator (LASSO). Then, after cross-validation, five candidate metabolites are obtained: ornithine, histidine, biotin, nicotinamide, and homocysteine.

6. The application according to claim 5, characterized in that, The COPD-PH disease diagnosis and / or early screening products include COPD-PH disease diagnosis and / or early screening kits, which contain reagents for detecting the combination of said metabolic markers.

7. The application according to claim 5, characterized in that, The COPD-PH disease diagnosis and / or early screening product includes a COPD-PH disease diagnosis and / or early screening model, the construction method of which includes the step of using the combination of metabolic biomarkers.

8. The application according to claim 5, characterized in that, The COPD-PH disease diagnosis and / or early screening product includes a COPD-PH disease diagnosis and / or early screening system, which includes a module containing the combination of the metabolic markers.

Images