ALD non-invasive differential diagnosis model based on plasma metabolite and construction method and application of ALD non-invasive differential diagnosis model

Abstract

The invention relates to the technical field of biological medicines, and provides a plasma metabolite-based ALD non-invasive differential diagnosis model, which is characterized in that five key metabolites in plasma, namely vecuronium bromide, N-(3-amino-3-oxopropyl)-L-valine, geranyl citronellol, phenylpropanolamine and S-allylcysteine, are used as core diagnosis indexes; or by taking alanine aminotransferase as an auxiliary index at the same time, calculating the ALD disease probability through a logistic regression equation. The invention further provides a construction method and application of the model. The method has the advantages that by accurately capturing plasma metabolite differences and combining clinical indexes, efficient identification of healthy people and ALD patients is achieved, and a reliable tool is provided for early screening and clinical diagnosis of ALD.

Description

Technical Field

[0001] This invention relates to the field of biomedical technology, and in particular to a non-invasive differential diagnostic model for alcoholic liver disease (ALD) based on plasma metabolites, its construction method, and its application. Background Technology

[0002] Alcoholic liver disease (ALD) is one of the most common types of chronic liver disease worldwide. Its course is clearly progressive, encompassing stages such as simple steatosis, alcoholic hepatitis (AH), alcoholic cirrhosis (ALC), and even hepatocellular carcinoma, posing a serious threat to human liver health and overall life safety. From a pathogenesis perspective, the metabolism of ethanol in the body is the core link in inducing ALD: acetaldehyde, a metabolite of ethanol, has strong cytotoxicity and can directly induce lipid peroxidation and DNA damage in hepatocytes; simultaneously, NAD3 caused by ethanol metabolism... + An imbalance in the NADH ratio significantly inhibits the mitochondrial β-oxidation pathway, leading to the accumulation of lipotoxic metabolites such as fatty acids and acylcarnitine in hepatocytes. This, in turn, induces endoplasmic reticulum stress and exacerbates oxidative stress, promoting hepatocyte apoptosis. Furthermore, long-term alcohol intake disrupts the regulatory balance between bile acid synthesis and enterohepatic circulation, reducing hydrophilic protective bile acids and enriching toxic bile acids such as deoxycholic acid (DCA) and lithocholic acid (LCA). This exacerbates hepatocyte membrane damage and cholestasis, creating a vicious cycle of "metabolic disorder-hepatocyte damage." Alcohol also disrupts the gut microbiota, causing abnormal release of gut microbiota metabolites and activating hepatic immune inflammatory pathways, further accelerating the progression of ALD and highlighting the dual role of metabolites as both "pathological markers" and "pathogenic drivers" in ALD development.

[0003] From an epidemiological perspective, the disease burden of ALD continues to increase with the shift in global alcohol consumption patterns. In my country, with the rise in social drinking rates and the evolution of drinking culture, the potential population for ALD is constantly expanding, and it is predicted that it may surpass hepatitis B in the future, becoming the core challenge for liver disease prevention and control in my country. Data from the Global Burden of Disease (GBD) study shows that alcohol abuse causes approximately 3 million deaths globally each year, nearly 50% of which are related to liver disease. The risk of developing ALD is significantly increased among long-term heavy drinkers. When men consume ≥40 grams of alcohol per day and women ≥20 grams per day for more than 10 years, the risk of developing cirrhosis is significantly increased compared to the general population, highlighting the urgency of ALD prevention and control.

[0004] However, the current clinical diagnostic system for ALD has significant limitations, making it difficult to meet the needs for early and accurate identification and risk stratification. Existing diagnoses primarily rely on comprehensive assessments, including a detailed history of alcohol consumption, changes in liver enzyme levels (e.g., AST / ALT ratio >1.5), other laboratory indicators (e.g., gamma-glutamyl transferase, total bilirubin, albumin, and coagulation function), imaging findings, and, when necessary, liver biopsy, while ruling out other possible causes of liver disease. However, in clinical practice, patients often underreport or conceal their alcohol consumption history, and traditional biochemical indicators have limited sensitivity and specificity in early lesions or acute exacerbations, easily leading to missed diagnoses or misdiagnoses. In the natural course of ALD, early identification and intervention are crucial for assessing disease severity, guiding individualized treatment plans, and developing long-term management strategies. While traditional diagnostic methods are fundamental, they still fall short in terms of precise subtyping.

[0005] In recent years, metabolomics technology, with its high sensitivity and specificity to dynamic changes in small molecule metabolites in vivo, has provided a new direction for the non-invasive diagnosis of ALD. Studies have confirmed that the plasma metabolite profile of ALD patients exhibits disease-specific changes: elevated fatty acylcarnitine reflects mitochondrial β-oxidation disorders, abnormal phospholipids indicate cell membrane damage, bile acid metabolism imbalance is associated with enterohepatic circulation disorders, and fluctuations in small molecules related to oxidative stress can quantify the degree of hepatocyte damage. These plasma metabolites not only show detectable changes in the early stages of ALD (such as in the stage of alcoholics alone), but can also distinguish different pathological stages (such as AH and ALC), providing a molecular basis for early screening and accurate subtyping, and can synergistically improve diagnostic efficacy with clinical indicators.

[0006] In summary, current ALD diagnosis faces challenges such as difficulty in early identification, low efficacy of traditional indicators, and poor accessibility of invasive diagnostic methods. There is an urgent need to develop a non-invasive differential diagnostic model for ALD based on plasma metabolites to overcome technical bottlenecks, achieve early and accurate identification and risk stratification, and lay the foundation for timely intervention, treatment optimization, and prognosis improvement. This has significant clinical and social implications. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide a non-invasive differential diagnostic model for alcoholic liver disease (ALD) based on plasma metabolites, and its construction method and application. By accurately capturing differences in plasma metabolites and combining them with clinical indicators, the model can achieve efficient differentiation between healthy individuals (HC) and patients with alcoholic liver disease (ALD).

[0008] The present invention solves the above-mentioned technical problems by adopting the following technical solutions:

[0009] A non-invasive differential diagnostic model for alcoholic liver disease (ALD) based on plasma metabolites, wherein the model uses five key metabolites in plasma as core diagnostic indicators; the five key metabolites are vecuronium bromide, N-(3-amino-3-oxopropyl)-L-valine, geranylide citronellol, phenylpropanolamine, and S-allyl cysteine.

[0010] The model calculates the probability of ALD using a logistic regression equation, which is: p(positive) = 1 / (1 + exp(-(-5.81898253 + 2.53176766 × relative abundance of vecuronium bromide + 13.84667083 × relative abundance of N-(3-amino-3-oxopropyl)-L-valine + (-0.21580122 × relative abundance of geranylidene citronellol) + 5.96063047 × relative abundance of phenylpropanolamine + (-17.46093785 × relative abundance of S-allyl cysteine)))); When the calculated p(positive) > 0.654, the patient is identified as having alcoholic liver disease (ALD).

[0011] As one of the preferred embodiments of the present invention, the relative abundance of the five key plasma metabolites is a relative value after internal standard correction and peak area normalization.

[0012] As one of the preferred embodiments of the present invention, the model also integrates alanine aminotransferase as an auxiliary diagnostic indicator to form a metabolite-alanine aminotransferase joint diagnostic model.

[0013] The logistic regression equation for the combined diagnostic model of metabolites and alanine aminotransferase is: p(positive) = 1 / (1 + exp(-(-0.94625651 + 0.96008273 × relative abundance of vecuronium bromide + 4.42839504 × relative abundance of N-(3-amino-3-oxopropyl)-L-valine + 0.93529431 × relative abundance of geraniol + 1.59791365 × relative abundance of phenylpropanolamine + (-6.56291868 × relative abundance of S-allyl cysteine) + 0.02671843 × alanine aminotransferase activity))); When the calculated p(positive) > 0.746, the patient is identified as having alcoholic liver disease (ALD).

[0014] A method for constructing a non-invasive differential diagnostic model for alcoholic liver disease (ALD) based on plasma metabolites includes the following steps:

[0015] (1) Sample collection and screening: Collect basic information and blood samples of the study subjects. The study subjects include healthy controls (HC), patients with alcoholic hepatitis (AH), and patients with alcoholic cirrhosis (ALC); clarify the inclusion and exclusion criteria, and screen samples that meet the requirements.

[0016] (2) Extraction and detection of plasma metabolites: The selected blood samples were preserved and metabolites were extracted. The extracted plasma metabolites were analyzed by LC-MS / MS using an ultra-high performance liquid chromatography-tandem Fourier transform mass spectrometry system. The corresponding chromatographic and mass spectrometric conditions were set.

[0017] (3) Data processing and sample division: Excluding research subjects that do not meet the standards, the train_test_split function of the sklearn package in Python is used to perform stratified random sampling in a ratio of 6:2:2 to divide the set into training set, validation set and test set. The K-fold crossover method is used for model prediction, K=5;

[0018] (4) Screening of key metabolites: In the training set and validation set, the random forest algorithm was used to screen out the key metabolites that have the most discriminative power for classifying alcohol-related liver disease in healthy control populations based on MeanDecrease Gini index analysis, n_estimators=100, max_depth=None.

[0019] (5) Model construction and validation: Based on the key metabolites screened, the diagnostic efficacy of a single metabolite is evaluated by receiver operating characteristic curve analysis and its optimal threshold is determined. A multi-metabolite joint diagnostic model is constructed by logistic regression algorithm. If a metabolite-alanine aminotransferase joint diagnostic model is constructed, the alanine aminotransferase index is further integrated and the corresponding model is established by logistic regression algorithm.

[0020] As one of the preferred embodiments of the present invention, in step (1):

[0021] Inclusion criteria were: age > 18 years; healthy controls with no history of alcohol consumption or serious illness; patients with alcoholic hepatitis or alcoholic cirrhosis with a history of alcohol consumption for more than 5 years, and men with a daily ethanol intake ≥ 40 grams and women with a daily ethanol intake ≥ 20 grams.

[0022] Exclusion criteria are: co-infection with hepatitis A, B, C, D, or E viruses or human immunodeficiency virus; co-infection with non-alcoholic fatty liver disease, drug-induced liver injury, autoimmune liver disease, or congenital liver disease; presence of primary liver cancer or liver metastases; presence of serious organic diseases affecting other organs; being pregnant or lactating; or a history of antibiotic or probiotic use within the past 3 months.

[0023] As one of the preferred embodiments of the present invention, in step (2):

[0024] The plasma metabolite extraction process is as follows: blood samples are placed in blood collection tubes, and the extractant is vortexed and mixed, extracted by low-temperature ultrasonication, allowed to stand at low temperature, centrifuged, dried by helium, reconstituted with the reconstituted solution, extracted by low-temperature ultrasonication again and centrifuged, and the supernatant is taken for analysis.

[0025] The chromatographic conditions were as follows: 3 μL of sample was separated using an HSST T3 column (100 mm × 2.1 mm, 1.8 μm); mobile phase A was a water / acetonitrile solution containing 0.1% formic acid (water / acetonitrile volume ratio 95 / 5); mobile phase B was an acetonitrile / isopropanol / water solution containing 0.1% formic acid (acetonitrile / isopropanol / water volume ratio 47.5 / 47.5 / 5); flow rate was 0.40 mL / min; column temperature was 40 °C.

[0026] The mass spectrometry conditions were as follows: positive and negative ion scanning mode, mass scan range m / z 10~1050, positive ion spray voltage 3500V, negative ion spray voltage -3000V, sheath gas 50arb, auxiliary heating gas 13arb, ion source heating temperature 450℃, and cyclic collision energy 20-40-60V.

[0027] As one of the preferred embodiments of the present invention, in step (3):

[0028] After excluding ineligible participants, a total of 123 participants were included. Stratified random sampling was performed in a 6:2:2 ratio, resulting in a training set of 73 participants, a validation set of 24 participants, and a test set of 24 participants. The training and validation sets together included 18 healthy controls, 43 patients with alcoholic hepatitis, and 36 patients with alcoholic cirrhosis. The test set included 5 healthy controls, 9 patients with alcoholic hepatitis, and 10 patients with alcoholic cirrhosis.

[0029] As one of the preferred embodiments of the present invention, in step (4), the key metabolites screened are vecuronium bromide, N-(3-amino-3-oxopropyl)-L-valine, geraniol, phenylpropanolamine, and S-allyl cysteine.

[0030] As one of the preferred embodiments of the present invention, in step (5):

[0031] The optimal thresholds for each key metabolite were: vecuronium bromide 5.866, N-(3-amino-3-oxopropyl)-L-valine 3.406, geraniol citronellol 3.451, phenylpropanolamine 4.315, and S-allyl cysteine ​​4.620.

[0032] The logistic regression equation for the multi-metabolite combined diagnostic model is: p(positive) = 1 / (1 + exp(-(-5.81898253 + 2.53176766 × relative abundance of vecuronium bromide + 13.84667083 × relative abundance of N-(3-amino-3-oxopropyl)-L-valine + (-0.21580122 × relative abundance of geranylideol) + 5.96063047 × relative abundance of phenylpropanolamine + (-17.46093785 × relative abundance of S-allylcysteine)))); When p(positive) > 0.480, the patient is diagnosed with ALD.

[0033] As one of the preferred embodiments of the present invention, in step (5):

[0034] The logistic regression equation for the combined diagnostic model of metabolites and alanine aminotransferase is: p(positive) = 1 / (1 + exp(-(-0.94625651 + 0.96008273 × relative abundance of vecuronium bromide + 4.42839504 × relative abundance of N-(3-amino-3-oxopropyl)-L-valine + 0.93529431 × relative abundance of geranylidene citronellol + 1.59791365 × relative abundance of phenylpropanolamine + (-6.56291868 × relative abundance of S-allyl cysteine) + 0.02671843 × alanine aminotransferase activity))); When the calculated p(positive) > 0.746, the patient is identified as an ALD patient.

[0035] Application of the above-mentioned non-invasive differential diagnostic model for ALD based on plasma metabolites in the preparation of diagnostic reagents for alcohol-related liver disease (ALD).

[0036] As one of the preferred embodiments of the present invention, the diagnostic reagent is used to detect the relative abundance of vecuronium bromide, N-(3-amino-3-oxopropyl)-L-valine, geranylide, phenylpropanolamine, and S-allyl cysteine ​​in the plasma of a subject, as well as the optional alanine aminotransferase activity value.

[0037] The advantages of this invention compared to the prior art are:

[0038] (1) The diagnostic efficacy is significantly higher, solving the problem of missed diagnosis and misjudgment by traditional indicators.

[0039] Existing technologies rely on traditional biochemical indicators, which have limited sensitivity and specificity in the early or acute exacerbation stages of ALD. In this invention, the combined model of five metabolites has an AUC of 1.000 (sensitivity 100.00%, specificity 100.00%) in the test set. After integrating ALT, the combined model still achieves a perfect AUC of 1.000, with positive / negative predictive values ​​of 100.00%. It can accurately distinguish between HC and ALD, and is especially suitable for early ALD screening.

[0040] (2) Completely non-invasive, avoiding the risks of invasive diagnosis.

[0041] Existing technologies for diagnosing ALD often require liver biopsy, which is an invasive procedure that is prone to complications such as bleeding and infection. Furthermore, patient acceptance is low and sampling errors exist. This invention only requires the collection of peripheral blood samples, and diagnosis can be completed by analyzing plasma metabolites using LC-MS / MS. The samples are easy to obtain and preserve, and there is no invasive damage. It is suitable for large-scale screening of high-risk populations and dynamic monitoring of the disease.

[0042] (3) The technical solution is highly operable and easy to promote in clinical practice.

[0043] Existing metabolomics diagnostic solutions are complex to operate and rely on special equipment; this invention uses a conventional ultra-high performance liquid chromatography-tandem Fourier transform mass spectrometry system for detection, and discloses in detail the sample pretreatment process, chromatographic conditions and mass spectrometry conditions, which can be directly reproduced by technicians; the model application only requires substituting into the logistic regression equation to calculate the probability, which has a low threshold and is suitable for use by medical institutions at all levels. Attached Figure Description

[0044] Figure 1 The results of plasma metabolite PLS-DA analysis in Example 1 are as follows: (Figure A shows the distribution of HC and AH metabolites, with significant differences between the two groups; Figure B shows the distribution of HC and ALC metabolites, with clear cluster boundaries; Figure C shows the distribution of HC and ALD metabolites, with high separation between the ALD and HC groups; Figure D shows the distribution of AH and ALC metabolites, with some differentiation but more overlap than the HC and ALD groups).

[0045] Figure 2 This is the feature importance map of key metabolites screened by random forest in Example 2 (using the Mean DecreaseGini index as an indicator, showing the top ten metabolites with the greatest influence on HC and ALD classification, with the top five being the core indicators of the model).

[0046] Figure 3The figures show the ROC curves of the HC and ALD differential diagnosis model constructed using the five key metabolites in Example 2 (Figure A shows the ROC curves of individual metabolites and the combined model in the training and validation sets; the combined model has an AUC of 0.998, sensitivity of 0.988, and specificity of 1.000; Figure B shows the ROC curves of the combined model in the validation and test sets; the validation set has an AUC of 0.998, sensitivity of 98.80%, and specificity of 100.00%; the test set has an AUC of 1.000, sensitivity of 100.00%, and specificity of 100.00%).

[0047] Figure 4 The ROC curves for the differential diagnosis model between HC and ALD constructed using "five metabolites + ALT" in Example 3 are shown in Figure A (ROC curves for the training set with a single metabolite, ALT alone, and the combined model; ALT alone, AUC = 0.706; combined model, AUC = 0.998, sensitivity 0.968, specificity 1.000; Figure B shows the ROC curves for the combined model of the validation and test sets; validation set, AUC = 0.998, sensitivity 100.00%, specificity 100.00%; test set, AUC = 1.000, sensitivity 100.00%, specificity 100.00%). Detailed Implementation

[0048] The embodiments of the present invention are described in detail below. These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments. Furthermore, unless otherwise specified, the reagents and instruments used in the following embodiments are all conventional reagents and instruments in the art; the experimental methods used, unless otherwise specified, are all conventional methods in the art and will not be described further.

[0049] Example 1: Detection and differential analysis of plasma metabolites in healthy controls and ALD patients:

[0050] 1. Determine the research subjects

[0051] We prospectively collected basic information and blood samples from hospitalized patients (including healthy controls (HC), patients with alcoholic hepatitis (AH), and patients with alcoholic cirrhosis (ALC)).

[0052] 2. Clarify the standards for inclusion and exclusion.

[0053] The inclusion criteria were: (1) age > 18 years; (2) healthy controls with no history of drinking or serious illness; (3) patients with alcoholic hepatitis or alcoholic cirrhosis with a history of drinking for more than 5 years, and men with a daily ethanol intake of ≥ 40 grams and women with a daily ethanol intake of ≥ 20 grams.

[0054] Exclusion criteria are: (1) co-infection with hepatitis A, B, C, D, or E virus or human immunodeficiency virus; (2) co-infection with non-alcoholic fatty liver disease, drug-induced liver injury, autoimmune liver disease, or congenital liver disease; (3) presence of primary liver cancer or liver metastases; (4) presence of serious organic diseases affecting other organs; (5) being pregnant or lactating; (6) history of antibiotic or probiotic use within the past 3 months.

[0055] 3. Collection and preservation of blood samples

[0056] Blood samples were collected from the subjects upon admission, placed in blood collection tubes, and stored at -80°C until they were sent to the sample center of Wuhan Union Hospital.

[0057] 4. Extraction of plasma metabolites

[0058] Pipette 100 μL of sample into a 1.5 mL centrifuge tube, add 400 μL of extraction buffer (acetonitrile:methanol = 1:1, containing four internal standards), vortex for 30 s, and then sonicate at low temperature for 30 min (5 °C, 40 kHz). Place the sample at -20 °C for 30 min, centrifuge at 13000 g for 15 min at 4 °C, transfer the supernatant, dry it with helium, reconstitute it with 100 μL of reconstitution solution (acetonitrile:water = 1:1), and sonicate at low temperature for 5 min (5 °C, 40 kHz). Centrifuge at 13000 g for 10 min at 4 °C, and transfer the supernatant to a vial with an inner tube for subsequent LC-MS / MS analysis.

[0059] The four internal standards in the extract include: L-2-chlorophenylalanine (CAS No.: 103616-89-3), L-phenyl-D5-alanine (CAS No.: 56253-90-8), cholic acid-[d4] (CAS No.: 116380-66-6), and acetyl-L-carnitine-d3 hydrochloride (N-methyl-d3) (CAS No.: 1334532-26-1). All four internal standards are commercially available, and their concentrations in the extract are as follows: L-2-chlorophenylalanine 0.01 mg / mL, L-phenyl-D5-alanine 0.003 mg / mL, cholic acid-[d4] 0.0005 mg / mL, and acetyl-L-carnitine-d3 hydrochloride (N-methyl-d3) 0.0001 mg / mL.

[0060] 5. LC-MS / MS analysis:

[0061] The instrument platform for LC-MS analysis was Thermo Fisher Scientific's UHPLC-Q Exactive HF-X ultra-high performance liquid chromatography-tandem Fourier transform mass spectrometry system.

[0062] Chromatographic conditions: 3 μL of supernatant was separated using an HSST T3 column (100 mm × 2.1 mm id, 1.8 μm) at a column temperature of 40 °C; mobile phase A was a water-acetonitrile (95:5, v / v) solution containing 0.1% formic acid, and mobile phase B was an acetonitrile-isopropanol-water (47.5:47.5:5, v / v) solution containing 0.1% formic acid; flow rate was 0.40 mL / min.

[0063] Mass spectrometry conditions: Sample mass spectrometry signal acquisition adopted positive and negative ion scanning mode, mass scanning range m / z: 10~1050, ion spray voltage: positive ion voltage 3500V, negative ion voltage -3000V, sheath gas 50arb, auxiliary heating gas 13arb, ion source heating temperature 450℃, cyclic collision energy 20-40-60V.

[0064] 6. Preprocessing data

[0065] Professional software was used to identify, align, baseline correct, and denoise the raw data peaks; the peak areas of metabolites were corrected using the four internal standards, including L-2-chlorophenylalanine, and then the relative abundance was obtained by normalizing the total peak area of ​​the sample.

[0066] 7. Analysis Results:

[0067] Figure 1 This presents the PLS-DA analysis results of plasma metabolites in healthy controls (HC) and patients with alcoholic liver disease (ALD) (including alcoholic hepatitis AH and alcoholic cirrhosis ALC). PLS-DA analysis revealed significant and stable categorical differences in plasma metabolites between ALD patients (including alcoholic hepatitis AH and alcoholic cirrhosis ALC) and healthy controls (HC), and these differences were clearly visible at different disease stages (AH and ALC). Figure 1 (A~C); especially, ALD patients have a very high degree of separation between their plasma metabolites and those of HC. Figure 1 D) This demonstrates that the occurrence and development of ALD are accompanied by characteristic abnormal distribution of plasma metabolites. This quantifiable difference in metabolites provides core data support and clear technical feasibility basis for subsequent screening of "key metabolites that can distinguish between HC and ALD" and construction of a non-invasive diagnostic model for ALD.

[0068] Example 2: Construction of a non-invasive differential diagnostic model for ALD based on plasma metabolites:

[0069] 1. Divide the sample

[0070] After excluding ineligible participants, a total of 123 participants were included. Stratified random sampling (6:2:2) using the `train_test_split` function from the Python sklearn package was employed to divide the participants into a training set (n=73), a validation set (n=24), and a test set (n=24), with values ​​rounded to the nearest whole number. The training and validation sets contained 18 cases of HC, 43 cases of AH, and 36 cases of ALC, while the test set contained 5 cases of HC, 9 cases of AH, and 10 cases of ALC. The proportions of each group in the training, validation, and test sets were maintained in a balanced manner. A K-fold crossover method (K=5) was used to predict the model.

[0071] 2. Screening key metabolites

[0072] In the training and validation sets, the Random Forest algorithm (n_estimators=100, max_depth=None) was used to evaluate the importance of metabolite classification based on the Mean Decrease Gini index, and the top ten key metabolites were selected as follows: Vecuronium, N-(3-Amino-3-oxopropyl)-L-valine, Geranylcitronellol, Phenylpropanolamine, S-Allylcysteine, N-cycloheptyl-6-thiophen-3-ylpyridine-3-carboxamide, N-Docosahexaenoyl Cysteine, and 2,6-diaminopimelic acid. acid (diaminopimelic acid), Butanal (n-butyraldehyde), Cyclohexanone (cyclohexanone) Figure 2 ).

[0073] The top 5 metabolites—Vecuronium, N-(3-Amino-3-oxopropyl)-L-valine, Geranylcitronellol, Phenylpropanolamine, and S-Allylcysteine—were selected as core indicators for subsequent model development.

[0074] 3. Construct and validate a multi-metabolite joint diagnostic model

[0075] The diagnostic efficacy of five metabolites—Vecuronium, N-(3-Amino-3-oxopropyl)-L-valine, Geranylcitronellol, Phenylpropanolamine, and S-Allylcysteine—was evaluated.

[0076] Receiver operating characteristic (ROC) curve analysis showed that these metabolites exhibited excellent differential diagnostic value for ALC: Vecuronium (AUC = 0.969, 95% CI: 0.932–0.995), N-(3-Amino-3-oxopropyl)-L-valine (AUC = 0.991, 95% CI: 0.976–1.000), Geranylcitronellol (AUC = 0.894, 95% CI: 0.796–0.969), Phenylpropanolamine (AUC = 0.929, 95% CI: 0.867–0.977), and S-Allylcysteine ​​(AUC = 0.887, 95% CI: 0.811–0.950). Notably, among the single metabolites, N-(3-Amino-3-oxopropyl)-L-valine exhibited the best diagnostic performance (AUC = 0.991), with a sensitivity of 0.949 and a specificity of 1.000. Figure 3 A, Table 1).

[0077] Table 1. Characteristics of the differential diagnostic model for HC and ALD plasma metabolites

[0078]

[0079]

[0080] Note: Combined Model is the model with the highest accuracy obtained from the training set through the validation set; Test is the result of the test set.

[0081] Meanwhile, based on the ROC curve analysis results, the optimal thresholds for each metabolite can be determined: Vecuronium 5.866, N-(3-Amino-3-oxopropyl)-L-valine 3.406, Geranylcitronellol 3.451, Phenylpropanolamine 4.315, and S-Allylcysteine ​​4.620.

[0082] A combined diagnostic model consisting of five metabolites—vecuronium (optimal threshold 5.866), N-(3-Amino-3-oxopropyl)-L-valine (optimal threshold 3.406), Geranylcitronellol (optimal threshold 3.451), Phenylpropanolamine (optimal threshold 4.315), and S-Allylcysteine ​​(optimal threshold -4.620)—demonstrated optimal disease predictive ability. Figure 3 A, B). Using the relative abundance of these five metabolites as independent variables and "whether or not the patient is an ALD patient" as the dependent variable, a logistic regression model was constructed, yielding the equation: p(positive)=1 / (1+exp(-(-5.81898253+2.53176766×Vecuronium relative abundance+13.84667083×N-(3-amino-3-oxopropyl)-L-valine relative abundance+(-0.21580122×Germainin) The model was calculated as follows: (relative abundance of phenylpropanol) + 5.96063047 × relative abundance of phenylpropanolamine + (-17.46093785 × relative abundance of S-allyl cysteine). This model demonstrated excellent diagnostic efficacy: validation set AUC = 0.998, test set AUC = 1.000, sensitivity, specificity, and other indicators (Table 1). When the calculated p(positive) > 0.654, the patient was identified as having alcoholic liver disease (ALD).

[0083] Example 3: Construction of a combined diagnostic model of metabolite-alanine aminotransferase (ALT):

[0084] Because the analysis of patients' basic clinical information revealed a significant difference in alanine aminotransferase (ALT) levels between the ALD group and the HC group (Table 2), a metabolite-alanine aminotransferase (ALT) combined diagnostic model was constructed based on the metabolite model established in Example 2, further integrating the important clinical variable ALT.

[0085] Table 2. Patient Basic Information and Clinical Data

[0086] Clinical indicators HC ALD P <![CDATA[Leukocyte,10 9 / L]]> 5.51±1.88 5.17±2.11 0.410 <![CDATA[Erythrocytes,10 12 / L]]> 4.33±0.48 4.01±0.96 0.020 <![CDATA[Platelet,10 9 / L]]> 210.55±58.28 151.46±69.36 0.000 Total bilirubin, μmol / L 10.82±4.0 30.79±43.72 0.000 ALT,U / L 22.04±12.26 62.7±90.12 0.000 AST, U / L 23.21±7.14 82.64±81.27 0.000 ALP,U / L 67.91±21.76 121.82±127.06 0.000 γ-GT,U / L 25.61±21.19 216.86±247.53 0.000 Albumin, g / l 43.71±4.03 41.46±9.51 0.070 Creatinine, μmol / L 67.23±19.55 73.88±23.1 0.150 Cholesterol, mmol / L 4.64±0.62 5.03±1.73 0.170 Triglycerides, mmol / L 1.38±0.72 2.06±1.66 0.020 HDL-C, mmol / L 1.22±0.4 1.26±0.76 0.870 LDL-C, mmol / L 2.68±0.86 2.64±1.28 0.910 D-dimer 0.25±0.06 1.47±2.53 0.000 PT,S 12.76±2.49 12.92±5.04 0.840 INR 1.42±2.27 2.66±4.15 0.070 APTT,S 39.26±12.07 46.48±21.18 0.040 FIB, g / L 4.41±6.9 7.13±11.34 0.160 TT,S 17.77±2.86 17.06±5.64 0.420

[0087] Note: HC (healthy control) refers to healthy controls; ALD (alcohol-associated liver disease) refers to alcohol-related liver disease; ALT (alanine aminotransferase) refers to alanine aminotransferase; AST (aspartate aminotransferase) refers to aspartate aminotransferase; ALP (alkaline phosphatase) refers to alkaline phosphatase; γ-GT (γ-glutamyl transpeptidase) refers to γ-glutamyl transpeptidase; HDL-C (High-Density Lipoprotein Cholesterol) refers to high-density lipoprotein cholesterol; LDL-C (Low-Density Lipoprotein Cholesterol) refers to low-density lipoprotein cholesterol; PT (prothrombin time) refers to prothrombin time; APTT (activated partial thromboplastin time) refers to activated partial thromboplastin time; FIB (fibrinogen) refers to fibrinogen; TT (thrombin time) refers to thrombin time.

[0088] 1. Obtain data

[0089] The relative abundance data of plasma metabolites were collected in the same manner as in Example 2. In addition, alanine aminotransferase (ALT) activity data (clinical routine enzymatic reaction method, unit U / L) were also collected.

[0090] 2. Construct a joint model

[0091] Using the relative abundance and ALT activity of five metabolites—vecuronium bromide, N-(3-amino-3-oxopropyl)-L-valine, geranylide, phenylpropanolamine, and S-allyl cysteine—confirmed in Example 2 as independent variables and "whether the patient is an ALD patient" as the dependent variable, a logistic regression model was constructed, yielding the equation p(positive) = 1 / (1+exp(-(-0.94625651+0.9600)). The relative abundance of vecuronium bromide was 8273 × 4.42839504 × N-(3-amino-3-oxopropyl)-L-valine + 0.93529431 × geraniol + 1.59791365 × phenylpropanolamine + (-6.56291868 × S-allylcysteine) + 0.02671843 × alanine aminotransferase activity. The model demonstrated near-perfect diagnostic performance in the training set (AUC = 0.998, 95% CI: 0.992–1.000), with a sensitivity of 96.80%, specificity of 100.00%, and both positive and negative predictive values ​​of 100.00%. When the calculated p(positive) > 0.746, the patient was diagnosed with ALD (Table 3). Figure 4 A, B).

[0092] Table 3. Characteristics of differential diagnostic models for HC and ALD plasma metabolites combined with ALT

[0093]

[0094] Note: Combined Model is the model with the highest accuracy obtained from the training set through the validation set; Test is the result of the test set.

[0095] Example 4: ALD diagnostic reagent based on a diagnostic model:

[0096] Reagent composition: Includes extraction components (extraction solution is acetonitrile-methanol = 1:1, containing four internal standards: L-2-chlorophenylalanine, L-phenyl-D5-alanine, cholic acid-[d4], and acetyl-L-carnitine-d3 hydrochloride (N-methyl-d3); reconstitution solution is acetonitrile-water = 1:1), detection components (mobile phase reagent containing 0.1% formic acid, adapted to UHPLC-Q Exactive HF-X system), and calibration components (five key metabolite standards).

[0097] Sample processing and detection: Select the sample to be validated (meeting the standards of Example 1), and use the extraction component to operate according to the plasma metabolite extraction steps of Example 1. Establish a standard curve through the calibration component to determine the relative abundance of metabolites, and at the same time detect the ALT activity value of the sample.

[0098] Result determination: Substitute the relative abundance of metabolites and the standardized ALT activity value into the metabolite-ALT combined diagnostic model equation of Example 3, calculate p(positive), and determine the patient as an ALD patient when p(positive) > 0.746.

[0099] Verification conclusion: Using comprehensive clinical diagnosis (combining alcohol consumption history, liver enzymes, imaging, etc.) as the gold standard, the diagnostic accuracy, sensitivity, and specificity of 40 samples were all 100.00%, indicating that the diagnostic reagent can accurately distinguish between HC and ALD and has the potential for clinical application.

[0100] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A plasma metabolite-based ALD non-invasive differential diagnosis model, characterized in that, The model takes five key metabolites in plasma as the core diagnostic indicators; the five key metabolites are vecuronium, N-(3-amino-3-oxopropyl)-L-valine, geranyl citronellol, phenyl ethanolamine, and S-allyl cysteine; The model calculates the ALD disease probability by a logistic regression equation: p(positive) = 1 / (1+exp(-(-5.81898253+2.53176766*vecuronium relative abundance+13.84667083*N-(3-amino-3-oxopropyl)-L-valine relative abundance+(-0.21580122*geranyl citronellol relative abundance)+5.96063047*phenyl ethanolamine relative abundance+(-17.46093785*S-allyl cysteine relative abundance)))); when the calculated p(positive) > 0.654, it is determined that the subject is an alcoholic liver disease ALD patient.

2. The model of claim 1, wherein, The relative abundance of the five key plasma metabolites is the relative value after internal standard correction and peak area normalization.

3. The model of claim 1, wherein, The model also integrates alanine aminotransferase as an auxiliary diagnostic indicator to form a metabolite-alanine aminotransferase combined diagnostic model; The logistic regression equation of the metabolite-alanine aminotransferase combined diagnostic model is: p(positive) = 1 / (1+exp(-(-0.94625651+0.96008273*vecuronium relative abundance+4.42839504*N-(3-amino-3-oxopropyl)-L-valine relative abundance+0.93529431*geranyl citronellol relative abundance+1.59791365*phenyl ethanolamine relative abundance+(-6.56291868*S-allyl cysteine relative abundance)+0.02671843*alanine aminotransferase activity value))); when the calculated p(positive) > 0.746, it is determined that the subject is an alcoholic liver disease ALD patient.

4. A method for constructing a plasma metabolite-based ALD non-invasive differential diagnosis model according to any one of claims 1 to 3, characterized in that, The method comprises the following steps: (1) Sample collection and screening: collect the basic information and blood samples of the research subjects, including healthy control population, alcoholic hepatitis patients, and alcoholic cirrhosis patients; determine the inclusion criteria and exclusion criteria, and screen the samples that meet the requirements; (2) Plasma metabolite extraction and detection: store and extract the metabolites of the screened blood samples, use an ultra-high performance liquid chromatography tandem Fourier transform mass spectrometry system to analyze the extracted plasma metabolites by LC-MS / MS, and set the corresponding chromatographic conditions and mass spectrometric conditions; (3) Data processing and sample division: exclude the research subjects who do not meet the criteria, use the train_test_split function of the sklearn package in python to perform stratified random sampling in a 6:2:2 ratio, divide into a training set, a validation set, and a test set, use K-fold cross-validation method for model prediction, and K = 5; (4) Key metabolite screening: In the training set and validation set, the key metabolites with the most discriminative power for the classification of healthy controls and alcohol-related liver disease were screened based on the Mean Decrease Gini index analysis using the random forest algorithm, n_estimators = 100, max_depth = None; (5) Model construction and validation: Based on the screened key metabolites, the diagnostic efficiency of individual metabolites was evaluated by receiver operating characteristic curve analysis to determine their optimal thresholds, and a multi-metabolite joint diagnostic model was constructed by the logistic regression algorithm. If a metabolite-alanine aminotransferase joint diagnostic model is constructed, the corresponding model is further established by the logistic regression algorithm by integrating the alanine aminotransferase index.

5. The construction method according to claim 4, characterized in that, In the step (1), the inclusion criteria are as follows: age > 18 years old; no drinking history and no history of serious illness in healthy control population; alcohol-related hepatitis patients and alcohol-related cirrhosis patients have a drinking history of more than 5 years, and the daily ethanol intake of males is ≥40 grams and the daily ethanol intake of females is ≥20 grams; The exclusion criteria are as follows: combined with hepatitis A, B, C, D, E virus or human immunodeficiency virus; combined with non-alcoholic fatty liver disease, drug-induced liver injury, autoimmune liver disease, congenital liver disease; presence of primary liver cancer or liver metastasis; presence of severe organic diseases affecting other organs; pregnancy or lactation; history of antibiotic or probiotic use within the past 3 months. In the step (2), the plasma metabolite extraction treatment is as follows: the blood sample is placed in a blood collection tube, mixed with an extraction solution by vortex, low-temperature ultrasonic extraction, low-temperature standing, centrifugation, helium blowing, re-dissolution, and re-extraction by low-temperature ultrasonic extraction and centrifugation, and the supernatant is taken for analysis; 6. The construction method of claim 4, wherein, The chromatographic conditions are as follows: 3 μL of sample is separated by HSST T3 column, the column specifications are 100 mm x 2.1 mm, 1.8 μm; the mobile phase A is 0.1% formic acid in water / acetonitrile solution, the volume ratio of water / acetonitrile is 95 / 5; the mobile phase B is 0.1% formic acid in acetonitrile / isopropanol / water solution, the volume ratio of acetonitrile / isopropanol / water is 47.5 / 47.5 / 5; the flow rate is 0.40 mL / min; the column temperature is 40℃; The mass spectrometry conditions are as follows: positive and negative ion scanning mode, mass scan range m / z 10-1050, positive ion spray voltage 3500V, negative ion spray voltage -3000V, sheath gas 50arb, auxiliary heating gas 13arb, ion source heating temperature 450℃, 20-40-60V cyclic collision energy. In the step (4), the screened key metabolites are verapamil, N-(3-amino-3-oxopropyl)-L-valine, geranyl citronellol, phenylpropanolamine, and S-allyl cysteine. In the step (5), the optimal thresholds of each key metabolite are as follows: verapamil 5.866, N-(3-amino-3-oxopropyl)-L-valine 3.406, geranyl citronellol 3.451, phenylpropanolamine 4.315, and S-allyl cysteine -4.

620.

7. The construction method of claim 4, wherein, ​ 8. The construction method according to claim 7, characterized in that, ​ ​ The multi-metabolite combined diagnostic model logistic regression equation is: p(positive) = 1 / (1+exp(-(-5.81898253+2.53176766 x relative abundance of vecuronium bromide+13.84667083 x relative abundance of N-(3-amino-3-oxopropyl)-L-valine+(-0.21580122 x relative abundance of geranyl citronellol)+5.96063047 x relative abundance of phenylserine+(-17.46093785 x relative abundance of S-allylcysteine)))); when p(positive) > 0.480, it is determined that the patient is an ALD patient. The metabolite-alanine aminotransferase combined diagnostic model logistic regression equation is: p(positive) = 1 / (1+exp(-(-0.94625651+0.96008273 x relative abundance of vecuronium bromide+4.42839504 x relative abundance of N-(3-amino-3-oxopropyl)-L-valine+0.93529431 x relative abundance of geranyl citronellol+1.59791365 x relative abundance of phenylserine+(-6.56291868 x relative abundance of S-allylcysteine)+0.02671843 x alanine aminotransferase activity value))); when the calculated p(positive) > 0.746, it is determined that the patient is an ALD patient.

9. Use of the plasma metabolite-based ALD non-invasive differential diagnosis model according to any one of claims 1 to 3 in the preparation of an ALD diagnostic reagent for alcohol-related liver disease.

10. Use according to claim 9, characterized in that, The diagnostic reagent is used to detect the relative abundance of vecuronium bromide, N-(3-amino-3-oxopropyl)-L-valine, geranyl citronellol, phenylserine, S-allylcysteine in the plasma of the subject, and optionally the alanine aminotransferase activity value. The diagnostic reagent is used to detect the relative abundance of vecuronium bromide, N-(3-amino-3-oxopropyl)-L-valine, geranyl citronellol, phenylserine, S-allylcysteine in the plasma of the subject, and optionally the alanine aminotransferase activity value.

Images