CN109060977B

CN109060977B - Biomarkers and kits for diagnosis of liver fibrosis and cirrhosis and methods of use

Info

Publication number: CN109060977B
Application number: CN201810769995.6A
Authority: CN
Inventors: 贾伟; 谢国祥; 卫润民; 王京晔
Original assignee: Shenzhen Huiyun Biological Technology Co ltd
Current assignee: Shenzhen Huiyun Biological Technology Co ltd
Priority date: 2018-07-13
Filing date: 2018-07-13
Publication date: 2024-07-02
Anticipated expiration: 2038-07-13
Also published as: CN109060977A

Abstract

The present invention provides a set of biomarkers that can be used to detect liver fibrosis and cirrhosis. The biomarker of the present invention refers to a metabolite component present in a biological sample of a subject, comprising a plurality of metabolites selected from the group consisting of cholic acid, amino acid and fatty acid, said metabolite combination comprising at least one amino acid, fatty acid and cholic acid, being differentially expressed in at least one target plasma or serum compared to one control plasma or serum. The biomarker may optionally be used in combination with a clinical indicator for diagnosis of liver fibrosis in a subject. The biomarker combination has the characteristics of high sensitivity and specificity for pathological stage diagnosis of patients with liver fibrosis, and can be used as a non-invasive means in clinic to improve clinical diagnosis and reduce puncture pain of patients. The invention also provides a using method of the biomarker and a kit containing the biomarker.

Description

Biomarkers and kits for diagnosis of liver fibrosis and cirrhosis and methods of use

Technical Field

The invention belongs to the technical field of biological detection, and particularly relates to the field of biomarkers and kits for diagnosing and detecting liver fibrosis and liver cirrhosis.

Background

Cirrhosis (Cirrhosis) and liver cancer (Hepatocellular Carcinoma) are very serious liver diseases, while liver fibrosis (Hepatic Fibrosis) is an early stage of their development. Liver fibrosis is a diffuse hyper-sedimentary disease of the extracellular matrix (Extracellular Matrix, ECM) within the liver, which is manifested by elevated levels of intrahepatic fibrosis (Fibrogenesis, i.e. ECM mass deposition) and insufficient fibrosis (Fibrolysis, i.e. ECM degradation) [Tacke F,Weiskirchen R.Update on hepatic stellate cells:pathogenic role in liver fibrosis and novel isolation techniques.Expert Review of Gastroenterology&Hepatology.2012,6:67–80.]. liver fibrosis is a common and common pathological outcome of many liver diseases, especially various chronic liver diseases, the formation and development of which is a dynamic pathological process. Liver fibrosis, i.e. fibrous hyperplasia, is a repair response [Albanis E,Friedman SL.Hepatic fibrosis.Pathogenesis and principles of therapy.Clinical Liver Disease.2001,5:315–334,v–vi.], of the body to injury and a contributing factor to liver injury. Repeated or sustained chronic liver parenchymal inflammation caused by various causes such as viruses, alcohol, parasites and the like can lead to continuous fibrous hyperplasia of the liver to form liver fibrosis [Hernandez-Gea V,Friedman SL.Pathogenesis of liver fibrosis.Annual Review of Pathology.2011,6:425–456.],, wherein 25% -40% of the liver fibrosis finally develops into liver cirrhosis and causes fatal complications of liver cancer. In the process, ECM including collagen, non-collagen glycoprotein, proteoglycan and the like is formed in a large amount and deposited in liver, early deposition only occurs in liver cell interstitium, later fibrous hyperplasia enters liver parenchymal cell gaps, fibrous ropes and fibrous intervals are gradually formed, the fibrous intervals are connected with each other again to form fibrous packages, and the liver gradually enters a pathological stage of liver cirrhosis.

The morphological characteristics of liver fibrosis and its important role in chronic liver disease were explained more clearly in the twentieth century 60-80, but are mostly considered as a passive irreversible process [Popper H,Uenfriend S.Hepatic fibrosis.Correlation of biochemical and morphologic investigations.American Journal of Medicine.1970,49:707–721.], and even form the false concept of three parts of "hepatitis- & gtliver cirrhosis- & gtliver cancer". In recent 20 years, studies of liver fibrosis have progressed, and the following aspects have been mainly clarified: ① Liver fibrosis is determined as a repair reaction of an organism to chronic injury, and is an active matrix proliferation pathological process; ② The cytological basis of hepatic fibrosis formation is hepatic stellate cell activation, and the molecular mechanism for regulating hepatic stellate cell activation, such as free radical, ECM environment and cytokine, especially transforming growth factor-beta 1 and other signal transduction mechanisms for stimulating cell activation, is basically defined; ③ The natural progress of liver fibrosis, especially hepatitis C liver fibrosis, the occurring risk factors, genetic and environmental influencing factors are basically known, the histopathological diagnosis standard of the liver fibrosis is basically established, and the serological comprehensive diagnosis is remarkably progressed; ④ In terms of treatment, liver fibrosis and a degree of cirrhosis have been shown to be reversible, and recent studies of the promotion of reversal [Hammel P,et al.Regression of liver fibrosis after biliary drainage in patients with chronic pancreatitis and stenosis of the common bile duct.The New England Journal of Medicine.2001,344:418–423.]. of liver fibrosis by some drugs have shown that liver fibrosis is reversible in certain circumstances, but if the etiology persists, liver fibrosis will eventually progress to irreversible cirrhosis. Therefore, early diagnosis of liver fibrosis and quantification of the degree of liver fibrosis are of great clinical value for timely intervention treatment and reversal of liver fibrosis development and prevention of cirrhosis and liver cancer.

At present, the gold standard for clinically performing liver fibrosis detection and staging in China is liver puncture biopsy (Liver Biopsy), and after anesthesia is performed on a patient, a biopsy needle with the length of about 70 mm is used for performing living liver puncture detection, so that the method is not only invasive and painful, but also can cause complications in terms of patient experience, and has a high false negative rate. The whole puncture process causes great physical and psychological pain and inconvenience to patients suffering from liver diseases, and simultaneously, the single detection cost reaches up to 1500 yuan to 2000 yuan for the RMB, so that the diagnosis is abandoned due to the fact that the economic aspect is difficult to bear.

Imaging evaluation methods such as ultrasonic imaging, computed tomography, magnetic resonance imaging and the like are complementary to each other, so that the diagnosis level of liver diseases is greatly improved. At present, the imaging diagnosis of liver fibrosis is still most common by B ultrasonic, but no index for sensitivity and specificity exists. The main reason is that when the liver tissue is subjected to fibrosis pathological change, the difference of the acoustic interface impedance or the change of the acoustic scattering coefficient of the tissue is not obvious, so that the difference of the response to the B-ultrasonic image and the normal liver tissue is not great. Recently, two ultrasound-based systems have been approved for use in the clinic, shear wave elastography and transient elastography (FibroScan), but these imaging modalities have limited accuracy [Morikawa H.Real-Time Tissue Elastography and Transient Elastography for Evaluation of Hepatic Fibrosis,2012.]. in ascites, elevated central venous pressure and obese patients-the cost and limited availability of these diagnostic interventions are also major obstacles in early accurate diagnosis and advanced treatment of chronic liver disease patients in the subset affected by the differences.

There is a critical clinical need for blood-borne biomarkers for diagnosing patients with chronic liver disease, as well as for assessing disease risk and monitoring disease progression and treatment response, which are generally non-invasive, widely deployable and of low cost. In recent years, scholars at home and abroad aim at researching a noninvasive liver fibrosis detection method, and find that means such as clinic, biochemistry, image and the like have important reference values in the aspect of liver fibrosis evaluation. Clinical evaluation includes relevant observation parameters of etiology, age, sex, disease course, treatment condition, clinical manifestation and the like, and the degree of liver fibrosis is judged by integrating and comprehensively evaluating the parameters. This method involves considerable experience and subjectivity, lacks sufficient accuracy, and can only be used as an aid. Biochemical indicators include serum etiology indicators (e.g., HBeAg, HBV DNA, etc.), serum liver fibrosis markers (e.g., hyaluronic acid HA, laminin LN, procollagen type iii, etc.), related biochemical indicators (e.g., aspartate aminotransferase, y glutamyl transpeptidase, platelets, etc.), and urine biochemical indicators (e.g., elastin and collagen degradation products Lai Ansu and hydroxylysylpyridinium). Other serum markers, such as keratin 18, while more sensitive than traditional liver injury markers, are not specific [Feldstein AE,Wieckowska A,Lopez AR,et al.Cytokeratin-18fragment levels as noninvasive biomarkers for nonalcoholic steatohepatitis:a multicenter validation study.Hepatology 2009;50:1072-8.]. some scoring systems such as FibroTest[Imbert-Bismut F,Ratziu V,Pieroni L,et al.Biochemical markers of liver fibrosis in patients with hepatitis C virus infection:a prospective study.Lancet 2001;357:1069-75.], aspartate aminotransferase/alanine aminotransferase (AST to ALT) ratio [Park SY,Kang KH,Park JH,et al.[Clinical efficacy of AST/ALT ratio and platelet counts as predictors of degree of fibrosis in HBV infected patients without clinically evident liver cirrhosis].Korean J Gastroenterol 2004;43:246-51.], have recently developed AST to platelet ratio index (APRI)[Wai CT,Greenson JK,Fontana RJ,et al.A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C.Hepatology 2003;38:518-526.], and FIB-4 (patient age, AST, ALT and platelet )[Sterling RK,Lissen E,Clumeck N,et al.Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection.Hepatology 2006;43:1317-25.], may be used to predict fibrosis and cirrhosis in hepatitis c control guidelines first introduced by the 2014 world health organization, two simple and readily available noninvasive serum index scores of APRI and FIB4 are recommended, however these two scores still remain to be further improved for diagnostic accuracy of hepatitis c and other chronic hepatitis hepatic fibrosis.

The liver plays a very important role in human life activities. Play an important role in digestion, absorption, excretion, bioconversion and metabolism of various substances, which are known as "substance metabolism centers", including cholesterol, lipids, fatty acids, bile Acids (BAs), and carbohydrate metabolism, liver plays an important role in lipid metabolism by absorption of Free Fatty Acids (FFAs), production, storage and transport of lipid and phospholipid metabolites, insulin reaction, alpha-oxidation, alterations in lipid storage and transport and pathophysiological changes associated with the development of liver cancer by BA nuclear receptor signaling and other pathways, such as Short Chain Fatty Acids (SCFAs) and choline, we have recently elaborated (by natural review of gastroenterology and hepatology) the importance of the cholic acid-intestinal microbiome metabolic axis, which is the result of intestinal microbiologic imbalance, in addition, advanced liver fibrosis, particularly liver cirrhosis, is closely associated with abnormal metabolism of plasma Amino Acids (AA) with reduced levels of branched-chain amino acids and increased concentrations of the aromatic amino acids phenylalanine and tyrosine, in view of the range of diseases associated with exposure to these toxins [Campollo O,Sprengers D,McIntyre N.The BCAA/AAA ratio of plasma amino acids in three different groups of cirrhotics.Rev Invest Clin 1992;44:513-8.].

Therefore, a noninvasive biomarker diagnosis method easy to detect is found, and the liver fibrosis stage is predicted and diagnosed by combining the existing clinical indexes, so that the method has great significance for disease risk assessment and prognosis of liver disease high-risk groups. At present, no rapid and efficient diagnostic kit for liver fibrosis is available on the market, and early detection and treatment of liver fibrosis are seriously affected.

Disclosure of Invention

The present invention provides a set of biomarkers useful in the detection of liver fibrosis, further provides kits for detecting liver fibrosis comprising the biomarkers and methods of use thereof.

The biomarker of the present invention refers to a metabolite component present in a biological sample of a subject, which may be a human or a mammal. The biological sample may be selected from urine, plasma or serum derived from a subject; serum is preferred in the present invention.

In a first aspect, the invention discloses biomarkers useful in the diagnosis of liver fibrosis, comprising a plurality of metabolites selected from the group consisting of cholic acid, amino acid and fatty acid, said combination of metabolites comprising at least one amino acid, fatty acid and cholic acid, differentially expressed in at least one target plasma or serum compared to one control plasma or serum. The biomarker may optionally be used in combination with a clinical indicator for diagnosis of liver fibrosis in a subject.

The metabolites are selected from one or more of the following metabolites: glycohyodeoxycholic acid, glycocholic acid, glycoursodeoxycholic acid, taurochenodeoxycholic acid, taurocholic acid, 7-ketocholic acid, taurochenodeoxycholic acid, myristic acid, palmitoleic acid, erucic acid, cis-13, 16-docosadienoic acid, nervonic acid, arachidonic acid, linoleic acid, elaidic acid, cis-7,10,13,16-tetracosatetraenoic acid, eicosadienoic acid, tyrosine, aspartic acid, beta-alanine, and valine.

The combination of clinical indicators comprises at least one clinical indicator that is differentially expressed in at least one liver disease patient versus a healthy subject. Clinical indicators may be obtained according to conventional clinical testing methods, such as aspartic acid aminotransferase, alanine aminotransferase, total bilirubin, direct bilirubin, indirect bilirubin, alkaline phosphatase, glutamyl transferase, albumin, total bile acid, cholinesterase, creatinine, blood urea nitrogen, cholesterol, triglycerides, high density lipoproteins, low density lipoproteins, and platelet count, and the like.

Use of said biomarker for the preparation of a kit for testing the degree of liver fibrosis in a subject, said kit comprising, in a method of use, the degree of liver fibrosis and stage of liver fibrosis in a subject, by testing the level of a metabolite in the plasma or serum of the subject, optionally in combination with a clinical indicator, preferably the age of the subject, the platelet count in the plasma of the subject, the level of aspartate aminotransferase and alanine aminotransferase.

As one of the preferred embodiments, the metabolites include elaidic acid, taurocholate, tyrosine and valine. Further provided are diagnostic kits for liver fibrosis comprising the four metabolites, standard solutions comprising elaidic acid, taurocholate, tyrosine and valine and an internal standard solution, wherein the internal standard solution refers to the isotopically labeled diagnostic markers elaidic acid, taurocholate, tyrosine and valine, and the isotopically labeled mode can be 2H or 13C. The kit also comprises an extract liquid, wherein the extract liquid consists of methanol and acetonitrile, and the volume ratio of the methanol to the acetonitrile is as follows: 1:1-5:1, and further may comprise 700 μl of 96-well plates, 350 μl of V-shaped 96-well plates, shrouded silica gel, 96-well sealed aluminum membranes.

The use method of the kit for diagnosing liver fibrosis is characterized in that the levels of metabolites of elapsic acid, taurocholate, tyrosine and valine in serum or plasma of a test subject are measured, and the measured values are input into a random forest model or a gradient lifting tree model to obtain score cut-off values for judgment.

The use method of the kit for diagnosing liver fibrosis comprises the following steps:

a) Preparing metabolite standard substance solutions with different concentrations: preparing solutions with different concentrations by using standard solutions of elaidic acid, taurocholate, tyrosine and valine respectively, placing the solutions and a blank control into a centrifuge tube, and centrifuging for 10-30 minutes under 4000-10000 revolutions of a table centrifuge; adding 200 microliters of freshly prepared deionized water into each centrifuge tube, shaking vigorously, shaking and dissolving for 10-15 minutes at 800-1200 rpm, and standing for later use;

b) Preparing an internal standard substance solution: 3 ml of methanol is taken as an internal standard diluent, added into an internal standard solution, covered with a cover, vigorously shaken, and left to stand for about 15 minutes for dissolution. Diluting an internal standard solution, and adding the internal standard solution into a 96-well micro-pore plate;

c) Preparation of serum or plasma samples: taking out 700 microliter of the microplate provided by the kit, sequentially adding 5 microliter of standard 1 to standard 7 and blank control into the A1 to A8 wells, and adding 5 microliter of serum (or plasma) sample or 5 microliter of low, medium and high concentration quality control into other wells. To each well was added 25. Mu.l of the internal standard solution, covered with a silica gel cap and shaken at 1000rpm for 10min. Centrifuge at 2000g for 2 min. The silica gel cover is taken down lightly, so that liquid in the micro-pore plate is prevented from splashing, and the silica gel pad is properly placed to prevent pollution for standby; reacting at 1450rpm at 30 ℃ for 60 minutes, carefully taking down the silica gel pad, and properly placing the silica gel pad to prevent pollution for later use; 350 microliters of loading buffer is added to each well, and the silica gel cover is covered for vigorous shaking; standing at-20 ℃ for 20 minutes, and centrifuging 2000g for 20 minutes; taking down the silica gel cover, carefully sucking 150 microliters of supernatant into a clean V-shaped bottom micro-pore plate, covering an aluminum foil envelope, and putting into an automatic sampler;

d) Measuring the sample prepared in c) by liquid chromatography and mass spectrometry, and calculating the concentration of the metabolite in the sample;

e) Inputting the concentration of the biomarker obtained in the step d) into a random forest model or a gradient lifting tree model for calculation, and judging the hepatic fibrosis degree of the test body according to the score.

As one of the preferred embodiments, the metabolite includes taurocholate and tyrosine.

As one of the preferred embodiments, the metabolite includes taurocholate and tyrosine. Further provided is a diagnostic kit for liver fibrosis comprising the two metabolites, a standard solution comprising taurocholate and tyrosine and an internal standard solution, wherein the internal standard solution refers to the diagnostic markers taurocholate and tyrosine marked by isotopes, and the isotope marking mode can be 2H or 13C. The kit also comprises an extract liquid, wherein the extract liquid consists of methanol and acetonitrile, and the volume ratio of the methanol to the acetonitrile is as follows: 1:1-5:1, and further may comprise 700 μl of 96-well plates, 350 μl of V-shaped 96-well plates, shrouded silica gel, 96-well sealed aluminum membranes.

The use method of the kit for diagnosing liver fibrosis is to determine the score cut-off value obtained by measuring the levels of the metabolites taurocholate and tyrosine in serum or plasma of a test subject, combining the age, platelet count, aspartate aminotransferase and alanine aminotransferase levels of the test subject, and inputting the values into a random forest model or a gradient lifting tree model.

a) Preparing metabolite standard substance solutions with different concentrations: preparing solutions with different concentrations by using standard solutions of taurocholate and tyrosine respectively, placing the solutions and a blank control into a centrifuge tube, and centrifuging for 10-30 minutes under 4000-10000 revolutions of a table centrifuge; adding 200 microliters of freshly prepared deionized water into each centrifuge tube, shaking vigorously, shaking and dissolving for 10-15 minutes at 800-1200 rpm, and standing for later use;

F) And d) inputting the concentration of the biomarker calculated in the step d) into a random forest model or a gradient lifting tree model for calculation in combination with the age, platelet count, aspartate Aminotransferase (AST) and alanine Aminotransferase (ALT) levels of the test body, and judging the hepatic fibrosis degree of the test body according to the score.

The invention also provides a kit for distinguishing between the serum of a patient suffering from liver fibrosis and the serum of a healthy individual, comprising said biomarker and at least one control serum from a healthy individual.

The invention also provides a kit for distinguishing between the serum of a patient with early stage liver fibrosis and the serum of a patient with late stage liver fibrosis, comprising the biomarker and at least one control serum from a patient with late stage liver fibrosis.

The invention also provides a kit for distinguishing between the serum of a patient suffering from liver fibrosis and the serum of a patient suffering from liver cirrhosis, comprising said biomarker and at least one control serum from a patient suffering from liver cirrhosis.

The invention also provides a liver fibrosis diagnosis kit, which comprises a random forest model, wherein the level of metabolites of elapsic acid, taurocholate, tyrosine and valine in a biological sample of a test subject is measured, and the measured values are input into the random forest model to obtain a score cut-off value, and if the score is smaller than 0, the test subject can be judged to be healthy; if the model risk score for the subject is greater than 0, the subject may be identified as liver fibrosis.

The inventors found that based on the method provided by the invention, a conventional clinically used liver function test kit can also be used for liver fibrosis diagnosis. The method comprises the steps of determining the levels of aspartic acid aminotransferase and alanine aminotransferase in a biological sample of a test body, combining the age and platelet count of the test body, inputting a gradient lifting tree model to obtain a score cut-off value, and judging the test body as an early-stage hepatic fibrosis patient if the model risk score of the test body is less than-0.93; if the model risk score for an individual is greater than-0.93, the individual may be identified as a patient with advanced liver fibrosis.

The random forest model or gradient-lifting tree model to which the present invention applies may be commercially available software, which may optionally also be part of the above-described kit. Preferably, the software package is part of a kit.

Drawings

FIG. 1 is a design of a biomarker screening experiment of example 1

Figures 2-7 show the statistics of examples 3-8 in sequence.

Figure 2 illustrates a random forest model comprising the preferred metabolite combinations of the invention for determining at least one chronic liver disease target serum. (A) ROC profile of diagnostic value of preferred metabolite combinations in the training set when compared to the control group. (B) ROC profile of diagnostic value of preferred metabolite combinations when compared to control group in the test set chronic liver disease group.

Figure 3 illustrates a random forest model comprising a preferred metabolite combination of the invention for further differentiating liver fibrosis target serum and serum of patients with cirrhosis. (A) ROC profiles of diagnostic value of preferred metabolite combinations when comparing liver fibrosis with cirrhosis in training sets. (B) ROC profiles of diagnostic value of preferred metabolite combinations when comparing concentrated liver fibrosis with cirrhosis were tested. (C) ROC graphs validating diagnostic value of preferred metabolite combinations when focused liver fibrosis is compared to cirrhosis.

FIG. 4 illustrates a random forest model comprising a preferred metabolite combination of the invention for further differentiating between early liver fibrosis target serum (S0-2) and serum of patients with advanced liver fibrosis (S3-4). (A) ROC plots of diagnostic value for the preferred metabolite combinations when compared to advanced liver fibrosis in the training set. (B) ROC plots of diagnostic value for preferred metabolite combinations when comparing early liver fibrosis with late liver fibrosis in the test set. (C) ROC plots focusing on diagnostic value of preferred metabolite combinations when comparing early liver fibrosis with late liver fibrosis were validated.

FIG. 5 illustrates a random forest model comprising the preferred metabolite combinations of the invention for determining at least one chronic liver disease target serum. (A) The preferred metabolite combination bin for training and test sets. (B) The preferred metabolite combinations in the training set and in the test set are used to diagnose the unsupervised hierarchical clustering of the four markers of the predictive model. (C) ROC profile of diagnostic value of preferred metabolite combinations in the training set when compared to the control group. (D) Accuracy-recall plot of diagnostic value of preferred metabolite combinations when compared to control groups in chronic liver disease groups in the test set. (E) ROC profile of diagnostic value of preferred metabolite combinations in the training set when compared to the control group. (F) Accuracy-recall plot of diagnostic value of preferred metabolite combinations when compared to control groups in chronic liver disease groups in the test set. (G) The preferred metabolite combinations in the training set and test set are used to diagnose risk factor scores for the predicted model.

FIG. 6 illustrates a random forest model comprising a preferred metabolite combination of the invention for further differentiating liver fibrosis target serum and serum of patients with cirrhosis. (A) The preferred metabolite combination bin for training and test sets. (B) The preferred metabolite combinations in the training set and in the test set are used to diagnose the unsupervised hierarchical clustering of the four markers of the predictive model. (C) ROC profiles of diagnostic value of preferred metabolite combinations when comparing liver fibrosis with cirrhosis in training sets. (D) Accuracy-recall plot of diagnostic value for a preferred metabolite combination when comparing liver fibrosis with cirrhosis in a test set. (E) ROC profiles of diagnostic value of preferred metabolite combinations when comparing liver fibrosis with cirrhosis in training sets. (F) Accuracy-recall plot of diagnostic value for a preferred metabolite combination when comparing liver fibrosis with cirrhosis in a test set. (G) The preferred metabolite combinations in the training set and test set are used to diagnose risk factor scores for the predicted model.

FIG. 7 illustrates a random forest model comprising a preferred metabolite combination of the invention for further differentiating between early liver fibrosis target serum (S0-2) and serum of patients with advanced liver fibrosis (S3-4). (A) The preferred metabolite combination bin for training and test sets. (B) The preferred metabolite combinations in the training set and in the test set are used to diagnose the unsupervised hierarchical clustering of the four markers of the predictive model. (C) ROC plots of diagnostic value for the preferred metabolite combinations when compared to advanced liver fibrosis in the training set. (D) Accuracy-recall plot of diagnostic value of preferred metabolite combinations when comparing early liver fibrosis with late liver fibrosis in the test set. (E) ROC plots of diagnostic value for the preferred metabolite combinations when compared to advanced liver fibrosis in the training set. (F) Accuracy-recall plot of diagnostic value of preferred metabolite combinations when comparing early liver fibrosis with late liver fibrosis in the test set. (G) The preferred metabolite combinations in the training set and test set are used to diagnose risk factor scores for the predicted model.

FIG. 8 selects a machine learning model that stably enhances FIB-4 classification.

FIG. 9 ROC curve of gradient-lifting tree model to distinguish early and late liver fibrosis, liver fibrosis and cirrhosis

FIG. 10 gradient-lifting tree model PR curve distinguishing early stage and late stage liver fibrosis, liver fibrosis and cirrhosis

Fig. 11.Fib-4 and GB gradient-lifting tree model distinguish the best score values for early and late liver fibrosis, liver fibrosis and cirrhosis in the discovery set as well as in the validation set.

Detailed Description

The following examples are provided to further illustrate the substance of the present invention, but are not intended to limit the scope of the present invention. Although the invention has been described in detail with reference to preferred embodiments, those skilled in the art will understand that modifications and equivalents may be made to the present invention without departing from the spirit and scope of the invention. The experimental methods for which specific conditions are not specified in the examples are generally conducted under conventional conditions, such as those described in textbooks and experimental guidelines, or under conditions recommended by the manufacturer.

Example 1: screening for differential biomarkers

The experimental design method is shown in figure 1. The test specimens in the present invention were approved by the local ethics committee and informed consent was obtained for all subjects. The present invention shares 1374 subjects into a training set, a testing set and an independent verification set. In the training set and the testing set, the content of metabolites such as cholic acid and amino acid in fasting (12 hours) serum specimens of 502 healthy people and 504 patients with chronic liver diseases, which are proved by liver puncture biopsy and have different degrees and are caused by HBV infection, and the detection of corresponding clinical indexes are respectively detected by using an ultra-high performance liquid chromatography tandem mass spectrometry technology. In independent verification and concentration, the ultra-high performance liquid chromatography tandem mass spectrometry technology is used for detecting the content of metabolites such as cholic acid, amino acid and the like in a fasting (12 hours) serum specimen of a chronic liver disease patient caused by HBV infection, which is verified by pathological puncture biopsy, and detecting corresponding clinical indexes. The detection method is as follows. The results of the measurements are shown in tables 1 and 2.

We split the dataset into a 70% training set and a 30% test set using a 100 random sampling method, and train a Decision Tree (DT), random Forest (RF), and gradient lifting tree (gradient boosting, GB) on the training set, predict the results on the test set and compare to FIB-4, apri, and AST/ALT.

FIB-4 refers to Fibrosis 4Score, another method proposed by Stirling (Sterling)[Sterling RK,Lissen E,Clumeck N,et al.Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection.Hepatology.2006.43(6):1317-25.] in 2006 to noninvasively evaluate liver fibrosis in chronic liver disease patients. APRI is the ratio index of aspartate Aminotransferase (AST) to Platelet (PLT) (Aspartate aminotransferase-to-Platelet Ratio Index)[Wai CT,Greenson JK,Fontana RJ,et al.A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C.Hepatology.2003.38(2):518-26.]

1. Liver biopsy

Except for patients who have been diagnosed with decompensation of cirrhosis (cirrhosis with complications of varicose bleeding, ascites, hepatic encephalopathy, and jaundice), ultrasound guided liver biopsy was performed. Biopsy specimens with a minimum length of greater than 1.5 cm were taken in at least 6 of the catchment areas, fixed with 10% formalin, paraffin embedded, and tissue sections stained with hematoxylin-eosin and Masson. Liver fibrosis is divided into S0 phase to S4 phase according to Scheuer pathological stage criteria. The above pathology evaluation was independently evaluated by three pathology specialists of Shanghai medical college at double denier university by a blind method, and the results were verified for consistency by Kappa test. When the evaluation result fails the Kappa test, the sample is re-analyzed to agree on the result.

2. Serum sample collection and preparation

5ML of fasting venous blood was collected and placed in a plastic centrifuge tube.

Serum preparation:

1) Slowly turning the serum preparation tube 5 times.

2) The test tubes were placed vertically on a tube rack for 1.5 hours at room temperature (about 25 degrees celsius).

3) The tube was centrifuged at 2500rpm for 10 minutes (4 degrees celsius).

4) The supernatant (about 2.5 ml) was dispensed into plastic centrifuge tubes (eppendorf, 1.5ml centrifuge tubes) using a pipette, with 0.5ml serum per cryopreservation tube.

5) Sample numbers were marked on the centrifuge tube.

6) Quickly placing in a refrigerator at-80 ℃.

3. Serum clinical marker detection

Hematology and biochemistry tests were performed using a LH750 hematology analyzer and a syncron DXC800 clinical system (Beckman Coulter, usa) according to the manufacturer's protocol; detection of blood hyaluronic acid and laminin was performed using a chemiluminescent immunoassay (LUMO, shinova Systems, shanghai, china); the detection of coagulation function was performed using a coagulation function meter (STAGO Compact, diagnostica Stago, france); blood HBV-DNA detection was performed using a real-time polymerase chain reaction system (LightCycler 480, roche, USA). Specific detection indexes are shown in Table 1.

3.1 Cholic acid detection of serum samples:

Sample preparation: mu.l of serum was taken in a 1.5mL centrifuge tube and 150. Mu.l of methanol (containing an internal standard, 50nM d4-CA (cholic acid), d4-UDCA (ursodeoxycholic acid), d4-LCA (lithocholic acid)) was added. Vortex shaking and mixing for 10 min, standing for 10 min, centrifuging at 4 degree 13500 r for 20 min, and collecting supernatant for analysis by UPLC-TQMS (ultra high performance liquid chromatography-triple quadrupole mass spectrometry).

Analytical instrument testing: UPLC-TQMS: a binary solvent controller and sample control chamber were equipped using a waters ultra-high performance liquid chromatography system (waters, usa). Dual electrospray ion sources were equipped using a waters XEVO triple quadrupole mass spectrometer (waters, usa).

Chromatographic conditions: UPLC BEH C18 column (100 mm. Times.2.1 mm,1.7 μm) was used; column temperature 45 ℃; mobile phase a: water (0.1% formic acid), B: acetonitrile (0.1% formic acid,); the flow rate is 0.4mL/min; the sample injection amount is 5uL; gradient elution conditions ：0-1min(5％B),1-5min(5-25％B),5-15.5min(25-40％B),15.5-17.5min(40-95％B),17.5-19min(95％B),19-19.5min(95-5％B),19.6-21min(5％B).

Mass spectrometry conditions: the electrospray ion source adopts an anion scanning mode (ESI-) and the specific conditions are as follows: capillary voltage 1.2kV, taper hole voltage 55V, extraction taper hole voltage 4V, ion source temperature 150 ℃, desolventizing gas temperature 550 ℃, reverse taper hole gas flow 50L/h, desolventizing gas 650L/h, low-quality area resolution 4.7, high-quality area resolution 15, and multi-reaction detection mode data acquisition.

3.2 Amino acid detection of serum samples:

Sample preparation: taking 40 mu L of serum, adding 500 mu L of methanol acetonitrile mixed solvent (1:9, v:v), and carrying out vortex oscillation for 2min; centrifuge tube was placed at-20deg.C for 10min to promote protein precipitation and centrifuged at 12000rpm for 15min at 4deg.C. The supernatant was taken at 20. Mu.L and dried under vacuum at room temperature. To the dried centrifuge tube, 100. Mu.L of a methanol-water mixed solvent (1:1, v: v, containing 1. Mu.g/mL of dichlorophenylalanine as an internal standard) was added for reconstitution and analysis.

Chromatographic conditions: UPLC BEH C18 column (100 mm. Times.2.1 mm,1.7 μm) was used; column temperature 40 ℃; mobile phase a: water (0.1% formic acid), B: acetonitrile (0.1% formic acid,); the flow rate is 0.4mL/min; the sample injection amount is 5uL; gradient elution conditions: 0-0.5min (1% B), 0.5-9min (1-20% B), 9-11min (20-75% B), 11-16min (75-99% B), 16-16.5min (99% B).

Mass spectrometry conditions: the electrospray ion source adopts an anion scanning mode (ESI-) and the specific conditions are as follows: the capillary voltage is 3.0, the taper hole voltage is 55V, the extraction taper hole voltage is 4V, the ion source temperature is 150 ℃, the desolvation gas temperature is 450 ℃, the reverse taper hole gas flow is 50L/h, the desolvation gas is 800L/h, the low-quality area resolution is 4.7, the high-quality area resolution is 15, and the data are collected in a multi-reaction detection mode.

Determination of the concentration of diagnostic markers to be tested: drawing a standard curve according to the concentration of the standard substance solution of the diagnostic marker to be detected, the corresponding diagnostic marker to be detected and the stable isotope internal standard area ratio same as that of the diagnostic marker to be detected, and measuring by adopting the isotope internal standard quantity. And simultaneously, the quality control of the sample detection process is carried out by adding an isotope internal standard into the sample.

The invention randomly divides 502 healthy patients and 504 patients with chronic liver disease into a training set and a testing set according to the proportion of 70% and 30%. In a training set, to distinguish healthy and chronic liver disease patients, liver fibrosis patients (stage S0-S3) and liver cirrhosis patients (stage S4), and liver fibrosis patients of different stages (stage S0-S2 and stage S3-S4), we used a one-way Wilcoxon rank sum test and LASSO to pick and identify candidate biomarkers, and a random forest model to evaluate candidate variables and model, and then separately validated the above models in a test set and independent validation set, and found that multiple biomarker combinations such as age (age), platelet count (PLT), aspartate Aminotransferase (AST), alanine Aminotransferase (ALT), elapsic acid (C18:2n6t), taurocholate (TCA), tyrosine (Tyrosine) and Valine (Valine) have predictive capacity for predicting chronic liver disease stage. Specifically, the results are shown in Table 3. The random forest model was performed using Shenzhen Town Biotechnology Inc LiveForest software, copyright accession number 2018SR227394, software name: a machine learning diagnosis system V1.0 for chronic liver disease based on metabonomics.

Table 3: diagnostic capabilities of biomarker combinations

Labeling: the cut-off value is the maximum value of the sum of sensitivity and specificity in the training set.

Example 2 predictive Capacity of biomarker combinations (elaidic acid (C18:2n6t), taurocholate (TCA), tyrosine (Tyrosine) and valine

The sample and data sources are the same as in example 1.

Candidate variables were evaluated and modeled using a random forest model, and the predictive power of a combination of elapsic acid (C18:2n6t), taurocholate (TCA), tyrosine (Tyrosine) and valine for predicting the extent of liver fibrosis was verified, and the results are shown in Table 4. From table 4, it can be seen that these four metabolite combinations have better predictive power by comparison with the existing test indicators.

The random forest model was performed using Shenzhen Town Biotechnology Inc LiveForest software, copyright accession number 2018SR227394, software name: a machine learning diagnosis system V1.0 for chronic liver disease based on metabonomics.

TABLE 4 Table 4

Example 3: biomarker combination random forest model to distinguish healthy and chronic liver disease groups

The biomarkers selected were: TCA (taurocholate), tyrosine (Tyrosine), and age, AST, ALT, and platelets as described in example 1, 363 patients with chronic liver disease and 371 healthy persons in the training set, 141 patients with chronic liver disease and 131 healthy persons in the test set, the likelihood of the above-described subjects suffering from chronic liver disease patients can be output using a biomarker combination random forest model trained in the training set, and the total ability of the model to distinguish patients from healthy persons can be evaluated by finding the best cut-off value from the about best point in ROC analysis. Results (fig. 2): the ROC curve lower area and 95% confidence interval in the training set are 0.986 (0.979-0.991), the PR curve lower area and 95% confidence interval are 0.988 (0.983-0.993), the optimal cut-off value is 0.551, and the sensitivity and specificity percentages at the optimal cut-off value are 92.4% and 95.3%, respectively; the ROC curve area under the test set and 95% confidence interval were 0.992 (0.982-0.998), the PR curve area under the 95% confidence interval was 0.994 (0.987-0.999), the optimal cut-off value was 0.551, and the sensitivity and specificity percentages at the optimal cut-off value were 96.2% and 94.3%, respectively. Measuring marker levels in said subject, and taking these measurements into a random forest model, if the individual score threshold is greater than 0.551, indicating that the individual is at a higher risk for chronic liver disease; if the individual score threshold is less than 0.551, this indicates that the individual has a lower risk for patients with chronic liver disease.

Example 4: biomarker combination random forest model for distinguishing liver fibrosis group and liver cirrhosis group

The biomarkers selected were: TCA (taurocholate), tyrosine (Tyrosine), and age, AST, ALT, and platelets as described in example 1, 299 cases of liver fibrosis patients and 64 cases of liver cirrhosis patients in the training set, 101 cases of liver fibrosis patients and 40 cases of liver cirrhosis patients in the test set, 193 cases of liver fibrosis patients and 175 cases of liver cirrhosis patients in the test set were independently verified, the probability of liver cirrhosis of the above subjects was output using a biomarker combination random forest model trained in the training set, and the best cut-off value was found by about the optimal point in ROC analysis, and the total ability of the model to distinguish liver cirrhosis patients from liver fibrosis patients was evaluated. Results (fig. 3): the ROC curve lower area and 95% confidence interval in the training set is 0.987 (0.978-0.994), the PR curve lower area and 95% confidence interval is 0.981 (0.965-0.992), the optimal cut-off value is 0.413, and the sensitivity and specificity percentages at the optimal cut-off value are 96.8% and 95%, respectively; the area under the ROC curve and 95% confidence interval in the test set are 0.916 (0.867-0.957), the area under the PR curve and 95% confidence interval are 0.908 (0.851-0.957), the optimal cut-off value is 0.413, and the sensitivity and specificity percentages at the optimal cut-off value are 73.3% and 89.0%, respectively; the ROC curve area under and 95% confidence interval in the independent validation set was 0.877 (0.831-0.918), the PR curve area under and 95% confidence interval was 0.843 (0.776-0.898), the optimal cut-off value was 0.413, and the sensitivity and specificity percentages at the optimal cut-off value were 87.9% and 72.6%, respectively. Measuring marker levels in said subject, and taking these measurements into a random forest model, if the individual score threshold is greater than 0.413, indicating that the individual has a higher risk of cirrhosis; if the individual score threshold is less than 0.413, this individual is indicated to have a lower risk of cirrhosis.

Example 5: biomarker combinations random forest model to distinguish between different stages of liver fibrosis group (early stage S0-S2 and late stage S3-S4)

The biomarkers selected were: TCA (taurocholate), tyrosine (Tyrosine), and age, AST, ALT, and platelets were as described in the specification, 260 early stage liver fibrosis patients and 103 late stage liver fibrosis patients in the training set, 89 early stage liver fibrosis patients and 52 late stage liver fibrosis patients in the test set, 155 early stage liver fibrosis patients and 213 late stage liver fibrosis patients in the test set were independently verified, the likelihood of developing late stage liver fibrosis in the above subjects was output using a biomarker combination random forest model trained in the training set, and the total ability of the model to distinguish late stage liver fibrosis patients from early stage liver fibrosis patients was evaluated by finding the best cut-off value from about the mount best point in ROC analysis. Results (fig. 4): the area under the ROC curve and the 95% confidence interval in the training set are 1.00 (1.00-1.00), the area under the PR curve and the 95% confidence interval are 1.00 (1.00-1.00), the optimal cut-off value is 0.574, and the sensitivity and the specificity percentage at the optimal cut-off value are 99.6% and 100% respectively; the area under the ROC curve and the 95% confidence interval in the test set are 0.925 (0.872-0.971), the area under the PR curve and the 95% confidence interval are 0.843 (0.741-0.934), the optimal cut-off value is 0.574, and the sensitivity and the specificity percentage at the optimal cut-off value are 72.7% and 93.1%, respectively; the area under the ROC curve and 95% confidence interval in the independent validation set are 0.878 (0.832-0.917), the area under the PR curve and 95% confidence interval are 0.878 (0.834-0.916), the optimal cut-off value is 0.574, and the sensitivity and specificity percentages at the optimal cut-off value are 86.3% and 72.3%, respectively. Determining marker levels in said subject, and taking these determinations into a random forest model, if the individual score threshold is greater than 0.574, indicating that the individual is at a higher risk of developing advanced liver fibrosis; if the individual score threshold is less than 0.574, this individual is indicated to have a lower risk of developing advanced liver fibrosis.

Example 6: biomarker combination random forest model to distinguish healthy and chronic liver disease groups

The biomarkers selected were: c18:2n6t (elaidic acid), TCA (taurocholate), tyrosine (Tyrosine) and valine.

As described in the specification, 363 chronic liver disease patients and 371 healthy people in the training set, 141 chronic liver disease patients and 131 healthy people in the test set, the possibility of the chronic liver disease patients of the subjects can be output by using the biomarker combination random forest model trained in the training set, and the total capacity of the model for distinguishing the patients and the healthy people can be estimated by finding the optimal cut-off value through about the optimal point in ROC analysis. Results (fig. 5): the ROC curve area under the training set and the 95% confidence interval are 0.998 (0.996-1), the PR curve area under the PR curve and the 95% confidence interval are 0.999 (0.996-1), the optimal cut-off value is-0.798, and the sensitivity and the specificity percentage at the optimal cut-off value are 98.3% and 99.7% respectively; the area under the ROC curve and 95% confidence interval in the test set was 0.999 (0.997-1), the area under the PR curve and 95% confidence interval was 0.999 (0.998-1), the optimal cut-off was-0.798, and the sensitivity and specificity percentages at the optimal cut-off were 98.4% and 96.3%, respectively. Determining marker levels in said subject, and taking these determinations into a random forest model, if the individual score threshold is greater than-0.798, indicating that the individual is at high risk for a patient suffering from chronic liver disease; if the individual score threshold is less than-0.798, this individual is indicated to have a lower risk for patients with chronic liver disease.

Example 7: biomarker combination random forest model for distinguishing liver fibrosis group and liver cirrhosis group

The biomarkers selected were: age, C18:2n6t (elaidic acid), TCA (taurocholate), tyrosine (Tyrosine), tyrosine and Valine ratio (Tyrosine to Valine ratio)

As described in the specification, 299 patients with liver fibrosis and 64 patients with liver cirrhosis in the training set, 101 patients with liver fibrosis and 40 patients with liver cirrhosis in the test set, 193 patients with liver fibrosis and 175 patients with liver cirrhosis in the test set were independently verified, the possibility of liver cirrhosis in the above subjects was output by using the biomarker combination random forest model trained in the training set, and the best cut-off value was found by about the mount optimum point in ROC analysis, and the total ability of the model to distinguish between patients with liver cirrhosis and patients with liver fibrosis was evaluated. Results (fig. 6): the ROC curve area under the training set and the 95% confidence interval are 0.991 (0.982-0.997), the PR curve area under the PR curve and the 95% confidence interval are 0.962 (0.93-0.985), the optimal cut-off value is 0.632, and the sensitivity and the specificity percentage at the optimal cut-off value are 91.4% and 100%, respectively; the ROC curve area under the test set and 95% confidence interval are 0.889 (0.829-0.939), the PR curve area under the PR curve and 95% confidence interval are 0.901 (0.851-0.945), the optimal cut-off value is 0.632, and the sensitivity and specificity percentages at the optimal cut-off value are 85.9% and 75.0%, respectively; the area under the ROC curve and 95% confidence interval in the independent validation set are 0.855 (0.8-0.899), the area under the PR curve and 95% confidence interval are 0.821 (0.751-0.889), the optimal cut-off value is 0.632, and the sensitivity and specificity percentages at the optimal cut-off value are 82.3% and 75.5%, respectively. Measuring marker levels in the subject, and taking these measurements into a random forest model, if the individual score threshold is greater than 0.632, indicating that the individual has a higher risk of cirrhosis; if the individual score threshold is less than 0.632, this individual is indicated to have a lower risk of cirrhosis.

Example 8: biomarker combinations random forest model to distinguish between different stages of liver fibrosis group (early stage S0-S2 and late stage S3-S4)

As described in the specification, 260 patients with early stage liver fibrosis and 103 patients with late stage liver fibrosis in the training set, 89 patients with early stage liver fibrosis and 52 patients with late stage liver fibrosis in the test set, 155 patients with early stage liver fibrosis and 213 patients with late stage liver fibrosis in the test set are independently verified, the possibility of suffering from late stage liver fibrosis of the subjects can be output by using a biomarker combination random forest model trained in the training set, and the optimal cut-off value can be found by about the optimal point in ROC analysis, and the total capacity of the model for distinguishing the patients with late stage liver fibrosis from the patients with early stage liver fibrosis can be evaluated. Results (fig. 7): the ROC curve area and 95% confidence interval in the training set is 0.993 (0.987-0.997), the PR curve area and 95% confidence interval is 0.981 (0.966-0.993), the optimal cut-off value is 0.588, and the sensitivity and specificity percentages at the optimal cut-off value are 96.1% and 100%, respectively; the ROC curve area under the test set and 95% confidence interval are 0.923 (0.86-0.974), the PR curve area under the PR curve and 95% confidence interval are 0.878 (0.786-0.95), the optimal cut-off value is 0.588, and the sensitivity and specificity percentages at the optimal cut-off value are 88.6% and 81.8%, respectively; the area under the ROC curve and 95% confidence interval in the independent validation set are 0.827 (0.773-0.874), the area under the PR curve and 95% confidence interval are 0.812 (0.751-0.871), the optimal cut-off value is 0.588, and the sensitivity and specificity percentages at the optimal cut-off value are 76.1% and 74.1%, respectively. Measuring marker levels in the subject, and taking these measurements into a random forest model, if the individual score threshold is greater than 0.588, indicating that the individual has a higher risk of developing advanced liver fibrosis; if the individual score threshold is less than 0.588, this individual is shown to have a lower risk of developing advanced liver fibrosis.

Example 9 selection of machine learning models that stabilize and improve Classification Effect

We split the dataset into 70% training set and 30% test set using 100 random samplings on the model build set, and train an example Decision Tree (DT), random Forest (RF), and gradient lifting tree (gradient boosting, GB) on the training set, predict and compare the results with FIB-4 on the test set, calculate AUPR and AUROC for each iteration to evaluate the performance of the different classifiers. In comparison to FIB-4 scoring, for prediction of advanced fibrosis we found: among the 100 iteration results 1) DT showed no significant difference from FIB-4 results at AUPR and even worse in AUROC, 2) RF AUPR was significantly improved, 3) GB showed the most significant improvement in AUROC and AUPR. For cirrhosis classification, we observed: 1) DT performed the worst on AUROC and AUPR, 2) RF significantly improved with respect to FIB-4 classification performance by 3) GB improvement was most pronounced. Furthermore, we consider GB as the optimal machine learning model that can stably enhance the FIB-4 classification effect (FIG. 8).

Example 10: gradient-lifting tree model distinguishes early liver fibrosis and late liver fibrosis, liver fibrosis and cirrhosis

We trained two GB models on a model set-up for distinguishing early and late liver fibrosis, liver fibrosis and cirrhosis, respectively, and validated the models on a separate validation set. Results (fig. 9, fig. 10): we find that: in the model establishment set, the GB model has the advantages that when the AUROC for distinguishing early liver fibrosis and late liver fibrosis is 0.904, AUPR is 0.836, the optimal cut-off value is-0.93, and compared with the FIB-4, the AUROC is remarkably improved (AUROC=0.817, AUPR=0.688); the GB model has the advantages that compared with the FIB-4, the AUROC for distinguishing hepatic fibrosis and liver cirrhosis is 0.961, AUPR is 0.891, the optimal cut-off value is-1.39, and the AUROC is remarkably improved (AUROC=0.864, AUPR=0.671); in independent validation sets, GB model significantly improved with respect to FIB-4 when AUROC was 0.918 and aupr was 0.925, which distinguishes between early and late liver fibrosis (auroc=0.841, aupr=0.844); the GB model showed a significant improvement over FIB-4 in terms of AUROC 0.871 and aupr 0.833 for liver fibrosis and cirrhosis (auroc=0.83 and aupr=0.738). Measuring the level of the hematological marker as described above in the subject, and taking these measurements into a gradient-lifting tree model, if the individual score threshold is greater than-0.93 in a model that distinguishes between early liver fibrosis and late liver fibrosis, indicating that the individual is at a higher risk of developing late liver fibrosis; if the individual score threshold is less than-0.93, indicating that the individual has a lower risk of developing advanced liver fibrosis; in a model that distinguishes between liver fibrosis and cirrhosis, if the individual score threshold is greater than-1.39, this individual is indicated to have a higher risk of developing cirrhosis; if the individual score threshold is less than-1.39, this individual is indicated to have a lower risk of cirrhosis (FIG. 11). Gradient-lifted trees were performed using Shenzhen Toku biotechnology Co., ltd LiveBoost software, copyright accession number XXXXX, software name: XXXX.

Claims

1. A biomarker combination useful for diagnosis of liver fibrosis, wherein the biomarker combination is a combination of taurocholate and tyrosine, and is useful for diagnosis of liver fibrosis in a subject in combination with a clinical indicator of age and platelet count, aspartate aminotransferase and alanine aminotransferase levels in plasma of the subject.

2. Use of a biomarker combination according to claim 1 for the preparation of a kit for the diagnosis of liver fibrosis, characterized in that the kit comprises a standard solution of the biomarker combination according to claim 1 and an internal standard solution, which refers to isotopically labelled diagnostic markers, taurocholate and tyrosine, in an isotopically labelled manner ² H or ¹³ C, and a calculation model selected from one of a gradient lifting tree model and a random forest model.

3. Use of a biomarker combination according to claim 2, for the preparation of a kit for the diagnosis of liver fibrosis, characterized in that it further comprises an extract consisting of methanol and acetonitrile in a volume ratio of 1: 1-5: 1.

4. Use of a biomarker combination according to claim 3 for the preparation of a kit for the diagnosis of liver fibrosis, characterised in that it further comprises 700 μl 96-well plates, 350 μl V-96-well plates, sealing plate silica gel and 96-well sealing aluminium membranes.

5. Use of the biomarker combination according to claim 1 for the preparation of a kit for distinguishing between serum of a liver fibrosis patient and serum of a healthy individual, characterized in that it comprises the biomarker combination according to claim 1 and at least one control serum from a healthy individual and a gradient-lifting tree model.

6. Use of the biomarker combination according to claim 1 for the preparation of a kit for distinguishing between the serum of a patient with early liver fibrosis and the serum of a patient with late liver fibrosis, characterized in that it comprises the biomarker combination according to claim 1 and at least one control serum from a patient with late liver fibrosis and a gradient-lifting tree model; the early stage liver fibrosis patient refers to a patient in the S0-S2 phase, and the late stage liver fibrosis patient refers to a patient in the S3-S4 phase.

7. Use of the biomarker combination according to claim 1 for the preparation of a kit for distinguishing between serum of a liver fibrosis patient and serum of a liver cirrhosis patient, characterized in that it comprises the biomarker combination according to claim 1 and at least one control serum from a liver cirrhosis patient and a gradient lift tree model; the liver fibrosis patients refer to S0-S3 patients, and the liver cirrhosis patients refer to S4-phase patients.