Biomarker and system for predicting death of ACLF patient and application of biomarker and system

Abstract

The invention provides a biomarker and a system for predicting whether an ACLF patient is dead or not and application of the biomarker and the system, ACLF related samples are analyzed through non-targeted metabonomics, a series of biomarkers capable of predicting the short-term death risk of the ACLF patient are found, a biomarker combination is further screened out from the biomarkers, a short-term post-death risk prediction model of the ACLF patient is constructed, and the prediction model is used for predicting the short-term post-death risk of the ACLF patient. The method achieves convenient and efficient prediction, masters the golden window period of intervention for high-risk people, reduces the economic burden for low-risk patients, and meets the clinical needs.

Description

[0001] This application is a divisional application of parent case 2023105195594.

[0002] This application claims priority to the prior Chinese application, application number: 2023103293484, filing date March 30, 2023; claims priority to the prior Chinese application, application number: 2023103293785, filing date March 30, 2023; all contents thereof are regarded as part of the present invention. Technical Field

[0003] The present invention relates to the field of biotechnology, and in particular to a biomarker, system and application thereof for predicting whether an ACLF patient will die. Background Art

[0004] Acute-on-chronic liver failure (ACLF) is a rapidly progressive disease with rapid clinical deterioration and high mortality. It manifests as acute liver dysfunction in patients with chronic liver disease or cirrhosis, characterized by organ failure and high short-term mortality. More than 25% of hospitalized patients with cirrhosis will experience this condition. ACLF is known to be associated with acute injuries such as alcohol, viral infection, extrahepatic organ failure and systemic inflammation. Its pathophysiology is related to persistent inflammation, immune dysregulation with initial widespread immune activation, a state of systemic inflammatory response syndrome and subsequent sepsis due to immunosuppression. The prevalence of ACLF is 5.7 cases per 1000 people per year according to Asia-Pacific standards and 20.1 cases per year according to European standards. There is currently no effective treatment for ACLF. The existing treatment options are mainly treatment of emergencies, organ support and liver transplantation. Despite intensive care, the prognosis of ACLF is poor, with a short-term mortality rate of about 50% within 90 days. At the same time, up to 20% of patients initially diagnosed with non-ACLF (pre-ACLF patients) may develop ACLF within 3 months. Early identification of pre-ACLF patients for active disease treatment and monitoring can significantly improve their clinical outcomes. It can be seen that the key to improving the survival rate of ACLF is to achieve accurate prognosis through early diagnosis and disease management.

[0005] Traditional scoring systems for cirrhosis, including the Model for End-Stage Liver Disease (MELD) score, MELD-sodium (MELD-Na) score, and Child-Turcotte-Pugh (CTP) score, have limited accuracy for ACLF prognosis. The development of more accurate ACLF prognostic tests has been a subject of intensive research over the past decade. Based on clinical and laboratory tests, the CANONIC study of the European Chronic Liver Failure Research Foundation (CLIF) proposed the CLIF-C ACLF score, which significantly improved the accuracy of short-term mortality prediction compared with traditional scoring systems. The latest biomarker discovery approaches include esoteric clinical tests such as serum ammonia, inflammatory markers, D-dimers, proteomics, and metabolomics. However, the discovery and validation process often relies on relatively small clinical cohorts and requires further validation. Therefore, more accurate and effective ACLF prognostic tests need to be found to meet clinical needs.

[0006] The China Acute-on-Chronic Liver Failure Alliance found that among patients hospitalized due to acute exacerbation of chronic liver disease, two groups of patients have a golden window intervention period. One group is patients who were non-ACLF patients on the day of admission but developed organ failure or even ACLF within 28 days of admission (i.e., patients with pre-ACLF on the day of admission). This stage before the patients develop into ACLF is a good time window for active intervention; the other group is patients who were ACLF patients on the day of admission but died within 29 to 90 days of admission. For this group of patients, the 4 weeks after admission is the golden window period for drug intervention.

[0007] Therefore, there is an urgent need to find a risk assessment method for the short-term mortality risk of ACLF patients and the short-term progression to ACLF in non-ACLF patients. When patients with short-term mortality risk or short-term progression to ACLF are found, intensive treatment or early liver transplantation should be carried out in a timely manner to achieve accurate screening and rapid and active clinical intervention in order to improve the survival rate of patients. At the same time, it can also save medical resources and reduce the economic burden for low-risk patients. Summary of the invention

[0008] In view of the problems existing in the prior art, the present invention provides a biomarker, system and application for predicting whether ACLF patients will die. Through non-targeted metabolomics analysis of ACLF-related samples, a series of biomarkers that can predict whether ACLF patients have a risk of death in the short term are found. These new markers are effectively combined, and biomarker combinations are further screened out, and a short-term mortality risk prediction model for ACLF patients is constructed to achieve convenient and efficient prediction, grasp the golden window period for intervention for high-risk groups, reduce the economic burden for low-risk patients, and meet clinical needs.

[0009] Acute-on-chronic liver failure (ACLF) is a rapidly progressive disease with rapid clinical deterioration and high mortality, manifested by acute liver dysfunction in patients with chronic liver disease or cirrhosis, characterized by organ failure and high short-term mortality, which occurs in more than 25% of hospitalized patients with cirrhosis.

[0010] The prediction of whether an ACLF patient will die in the present invention refers to predicting among ACLF patients, finding ACLF patients with a high mortality rate in the future, which is an early prediction of whether the disease will rapidly worsen and lead to death in the short term in the future, and the prediction is relatively difficult.

[0011] The prediction of whether an ACLF patient is an ACLF patient in existing reports and the prediction of whether an ACLF patient will die as described in the present invention are two completely different concepts. Among them, to predict whether an ACLF patient is an ACLF patient, it is only necessary to distinguish the non-ACLF control population (including healthy people or stable chronic liver disease people) from the ACLF patients. There will be more differential markers in the samples of ACLF patients and the control population, which are easier to distinguish; while to predict whether an ACLF patient will die, it is necessary to distinguish which patients have a high mortality rate and which patients have a very low mortality rate from the patients with the same ACLF. The difficulty of distinguishing is bound to be much higher than the difficulty of distinguishing healthy people from ACLF patients, because those involved in the prediction have already suffered from ACLF, and the blood or body fluid samples of this part of the population will be closer, with fewer differential markers, and the differences between markers will also be smaller.

[0012] The prediction of whether an ACLF patient will die described in the present invention is completely different from the prediction of whether the patient is an ACLF patient.

[0013] It is understandable that the short term in "ACLF patients' short-term mortality risk" is not an absolute and unchanging time node. Although the commonly used time point in clinical practice is "ACLF group's 90-day mortality rate", this is only the time node summarized based on clinical experience. With the passage of time or the improvement of other treatment methods, the time node or the length of time may change. For example, within 90 days or 100 days, or within one year, 360 days, or within half a year, 180 days, or within 80 days, or within 20 days, within 4 weeks, within 3 weeks, within 2 weeks, within 1 week, etc., it is possible to progress from non-ACLF to ALCF, or from ACLF to death.

[0014] Existing ACLF-related scoring systems often rely on relatively small clinical cohorts and are difficult to validate. A more accurate and effective ACLF prognosis test remains an unmet clinical need. The present invention performed metabolomics analysis on 1,300 serum samples of chronic liver disease patients (including 420 ACLF and 880 non-ACLF) with acute exacerbation upon admission. The samples were randomly divided into two equal halves as a discovery set and a validation set. Statistical analysis and random forest analysis were used in the discovery set to identify metabolites associated with mortality within 90 days in the ACLF group and progression to ACLF within 28 days in the non-ACLF group. A predictive model was developed and well validated in the validation set.

[0015] Studies have shown that the serum metabolome, especially membrane lipid metabolism, steroid hormone metabolism, oxidative stress pathways and energy metabolism, was significantly changed between patients who died within 90 days of the ACLF group and those who survived within 90 days of the ACLF group; and there was also a significant correlation between the series of differential metabolites between patients who died within 90 days of the ACLF group and those who survived within 90 days of the ACLF group and the series of differential metabolites of whether the non-ACLF group would progress to ACLF patients within 28 days. The present invention trains and verifies a machine learning algorithm for predicting the mortality rate of ACLF within 90 days (area under the curve, AUC: 0.80) and the progression of non-ACLF to ACLF within 28 days (AUC: 0.85), uses liquid chromatography-mass spectrometry (LCMS) to conduct targeted analysis of selected metabolites, and the constructed model can be used clinically.

[0016] In one aspect, the present invention provides a use of a biomarker for preparing a reagent for predicting whether an ACLF patient has a risk of death or a probability of survival, wherein the biomarker comprises one or more of the following: 2-hydroxydadipic acid, hexadecanedioic acid, hydroxyproline, methionine sulfoxide, 3-hydroxybutylcarnitine, pipecolic acid, d), estrone sulfate, N-acetylamino-glutamic acid (NAAG), quinolinic acid, ureidopropionic acid, γ-CEHC (3-(6-hydroxy-2,7,8-trimethyl-3,4-dihydro-2H-benzopyran-2-yl) propionic acid), thyroxine, formiminoglutamic acid, γ-3 glutamyl dipeptide (gamma-glu-phe), and bilirubin.

[0017] Furthermore, predicting whether an ACLF patient has a risk of death refers to predicting whether an ACLF patient has a risk of death in the short term.

[0018] The "short term" here can refer to whether the ACLF patient has a risk of death within 90 days, 180 days, 360 days, etc., and of course it can also be less than 90 days, such as 80 days, 70 days, 60 days or shorter, or it can be within 1 year, 11 months, 10 months, 9 months, 8 months, 7 months, 6 months, 5 months, 4 months, 3 months, 2 months, 1 month, etc.

[0019] Because the short-term here is difficult to predict in advance, it is a time node summarized based on current clinical experience. With the passage of time or the improvement of other treatment methods, this time node or the length of time may change. It may also change according to the length of the patient's hospitalization.

[0020] The present invention collects serum samples from ACLF patients at the time of admission through non-targeted metabolomics research, and divides the patients into a survivor group and a non-survivor group (death) according to whether they survive in the short term. The serum samples of the survivor group and the non-survivor group at the time of admission are analyzed by UPLC-MS / MS ultra-high performance liquid chromatography-tandem mass spectrometry, and a series of biomarkers that can predict the short-term mortality risk of ACLF patients are found. Machine learning, statistical analysis and pathway enrichment analysis (including random forest, FDR test, t-test) are used to determine the candidate biomarkers in the discovery set and determine their priorities, and significantly different metabolites are screened, and finally 15 biomarkers are obtained, which can be used to efficiently predict the short-term mortality risk of ACLF patients.

[0021] In some embodiments, the biomarker reagent for predicting whether an ACLF patient has a risk of death or a probability of survival is a detection reagent prepared with the biomarker as the detection target, such as a sample pretreatment reagent, an antigen or an antibody, and other biological reagents and kits suitable for the detection of the biomarker; it can also be developed into a standardized reagent or kit suitable for LC-UV or LC-MS detection of the biomarker.

[0022] Further, the biomarkers include colicolic acid, NAAG and ureidopropionic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid and methylaminoglutamic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid and γ-CEHC; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC and γ-3 glutamyl dipeptide; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide and 3-hydroxybutylcarnitine; or, The biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine and estrone sulfate; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate and hydroxyproline; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline and bilirubin; or, the biomarkers include colicolic acid , NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, and 2-hydroxydipropionic acid; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, and methyl sulfide; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline , bilirubin, 2-hydroxydipropionic acid, methyl sulfide, and hexadecanedioic acid; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, and quinolinic acid; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, quinolinic acid, and thyroxine.

[0023] Furthermore, the biomarkers include one or more of the following: colic acid, NAAG, ureidopropionic acid, and methylaminoglutamate.

[0024] Furthermore, the biomarkers include one or more of the following: colic acid, NAAG, and ureidopropionic acid.

[0025] By examining the concentration differences, AUC value rankings, and significance rankings of biomarkers in serum samples of the survivor group and the non-survivor group at the time of admission, individual biomarkers that can significantly distinguish between the survivor group and the non-survivor group were further selected from the 15 biomarkers, which can be used to more effectively predict whether ACLF patients have a risk of death in the short term, or to build a diagnostic model for predicting the short-term risk of death in ACLF patients.

[0026] Furthermore, the reagent is used to detect biomarkers in blood, serum, plasma or whole blood samples of ACLF patients.

[0027] Further, the reagent is used to detect the presence or absence of biomarkers in the sample, or the relative abundance or relative concentration, or the degree or amplitude of the change in the content or quantity of the marker. The presence or absence of the marker or the level of the content here is a relative concept. For example, for patients with ACLF death and surviving patients, the content of these specific markers is compared with the ACLF death or surviving patients as a benchmark. It may be that for some markers, the content in patients with ALCF death or high probability of death is higher than that in ACLF surviving patients, and this high has a statistical difference, such as a significant or extremely significant increase. When determining whether non-ALCF progresses or develops into ACLF, the content of these specific markers can be used as a comparison standard relative to one of non-ALCF or ACLF patients. Therefore, when these markers are judged, if it is a single marker, if the probability of the marker occurring at a certain risk increases, the content of the marker changes, and the change here may be a relative increase or a relative decrease. This relative increase or relative decrease has a significant difference, and of course it can also be an extremely significant difference. Therefore, no matter what means are used for detection, a predetermined value can be used as a standard (cut-off value). If the value is higher than this value, it is considered that the content has changed. Such results can be used as predictive or diagnostic value.

[0028] Therefore, in some aspects, the markers described in the present invention can be obtained by detecting the marker content in the sample by any existing known method, such as liquid chromatography, gas chromatography, mass spectrometry, LC-MS, GC-MS, CC-MS, LC-MS-MS, NMR, immunochromatography test paper, immunoreaction chip, capillary electrophoresis, infrared spectroscopy, etc. As long as they can be used to detect the marker content in the sample, they can be used to predict whether ACLF patients have a risk of death in the short term. As long as the content in the sample can be detected, it can be used to predict or diagnose the probability of a certain disease. It can be understood that the detection here is to detect the individual sample, and then compare it with the pre-set standard, and the comparison result is used to judge or predict the occurrence state of the disease, for example, it can be used to predict whether ACLF will die or the probability of death, whether ACLF will survive or the probability of survival, and whether non-ACLF will progress to ACLF. This prediction or diagnosis is whether it will occur within a certain period of time. Of course, such detection can be continuous detection, and the progress of the disease can be inferred as the content of certain substances changes.

[0029] In some embodiments, the relative abundance is the peak area of ​​the biomarker in the detection spectrum obtained by high performance liquid chromatography-tandem mass spectrometry. For example, if the average peak area of ​​a biomarker measured in the control sample (ACLF patient short-term survivor group) is 500, and the average peak area measured in the sample of the ACLF patient short-term non-survivor group is 3000, then the abundance of the biomarker in the sample is considered to be 6 times that in the control sample.

[0030] Furthermore, the new markers discovered by the present invention can be combined with the following factors for prediction or judgment, and the following factors can be some chemical substances or some indicators. Therefore, in some methods, one or more factors of age, INR (international normalized ratio), creatine (CR), sodium, albumin (ALB), neutrophil / lymphocyte count ratio (granulocyte / lymphocyte ratio, N / L ratio), overthepatic encephalopathy (overt HE), and TB (total bilirubin) can also be combined for diagnosis or prediction.

[0031] Furthermore, the predicting whether the ACLF patient has a risk of death or a probability of survival refers to predicting whether the ACLF patient has a risk of death or a probability of survival within 360 days.

[0032] Furthermore, the predicting whether an ACLF patient has a risk of death or a probability of survival refers to predicting whether an ACLF patient has a risk of death or a probability of survival within 90 days.

[0033] In yet another aspect, the present invention provides a kit for predicting whether an ACLF patient has a risk of death or a probability of survival, the kit comprising a detection reagent for the biomarker for the use described above.

[0034] Furthermore, the detection reagent of the biomarker includes a standard of the biomarker.

[0035] In another aspect, the present invention provides a biomarker combination for predicting whether an ACLF patient has a risk of death, comprising the following biomarkers: colicoagulase, NAAG, and ureidopropionic acid; or colicoagulase, NAAG, ureidopropionic acid, and methylaminoglutamic acid; or colicoagulase, NAAG, ureidopropionic acid, methylaminoglutamic acid, and γ-CEHC; or colicoagulase, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, and γ-3 glutamyl dipeptide; or colicoagulase, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide dipeptide and 3-hydroxybutylcarnitine; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine and estrone sulfate; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate and hydroxyproline; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline and bilirubin; or colicolic acid, NAA G, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, and 2-hydroxydipropionic acid; or colic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, and methyl sulfide; or colic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, and methyl sulfide. erythrin, 2-hydroxydipropionic acid, methyl sulfide and hexadecanedioic acid; or colic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid and quinolinic acid; or colic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, quinolinic acid and thyroxine.

[0036] Further, the biomarker combination includes the following biomarkers: colicosanoids, NAAG, ureidopropionic acid and methylaminoglutamic acid; or colicosanoids, NAAG and ureidopropionic acid.

[0037] Furthermore, the biomarker combination also includes any one or more of age, INR (international normalized ratio), creatine, Na (sodium), ALB (albumin), N / L ratio (granulocyte / lymphocyte ratio), overt HE (overt hepatic encephalopathy), and TB (total bilirubin).

[0038] On the other hand, the present invention provides a system for predicting whether an ACLF patient has a risk of death or a probability of survival, the system comprising a data analysis module; the data analysis module is used to analyze the detection value of a biomarker, wherein the biomarker comprises one or more of the following: 2-hydroxydipropionic acid, hexadecanedioic acid, hydroxyproline, methyl sulfide, 3-hydroxybutylcarnitine, colic acid, estrone sulfate, N-acetylamino-glutamic acid, quinolinic acid, ureidopropionic acid, γ-CEHC, thyroxine, methylaminoglutamic acid, γ-3 glutamyl dipeptide, and bilirubin.

[0039] Furthermore, the detection value of the biomarker is to detect the biomarker in the blood, serum, plasma or whole blood sample of ACLF patients, and is used to detect the presence or relative abundance or relative concentration of the biomarker in the sample, or the change in relative content.

[0040] Further, the biomarkers include colicolic acid, NAAG and ureidopropionic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid and methylaminoglutamic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid and γ-CEHC; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC and γ-3 glutamyl dipeptide; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide and 3-hydroxybutylcarnitine; or, The biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine and estrone sulfate; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate and hydroxyproline; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline and bilirubin; or, the biomarkers include colicolic acid , NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, and 2-hydroxydipropionic acid; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, and methyl sulfide; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline , bilirubin, 2-hydroxydipropionic acid, methyl sulfide, and hexadecanedioic acid; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, and quinolinic acid; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, quinolinic acid, and thyroxine.

[0041] Furthermore, the data analysis module calculates a predicted value for predicting whether an ACLF patient has a risk of death by substituting the detection value of the biomarker into a regression equation, thereby evaluating the risk of death of the ACLF patient.

[0042] The present invention can analyze the detection value of the biomarker by any method (such as by computer software, or manual experience evaluation, etc.), obtain certain equations, and then bring the test results into the equation to automatically obtain the predicted result value. It is used to calculate the predicted value of whether ACLF patients have a risk of death or predict whether non-ACLF patients will progress to ACLF. These equations or computer programs are generally generated by analyzing between certain groups, and then verifying the accuracy of the equation operation results between the determined groups. For example, select the death group and the survivor group of ACLF patients for equation operation training, and then continue to select another death group and survivor group of ACLF patients for verification to see the accuracy of the results, so as to obtain the ideal equation for operation. It can be understood that the more data obtained, the more the equation can be continuously revised or the AI ​​can self-learn and self-adjust, so that the prediction accuracy is higher and more in line with the actual situation. Therefore, no matter which model or equation can be applied to the invention to automatically output the prediction results. Therefore, there are many computer algorithms to choose from, such as mixed discriminant analysis (mda); gradient boosting (gbm); logistic regression (plr); generalized linear model (glmnet); random forest (ranger); naive Bayes ( _bayes); neural network (avNNet); support vector machine (svmRadial) and other computing methods, these computing methods can be used for the markers or combinations of markers of the present invention to make predictions or diagnoses.

[0043] Furthermore, predicting whether an ACLF patient has a risk of death refers to predicting whether an ACLF patient has a risk of death in the short term.

[0044] The "short term" here can refer to whether the ACLF patient has a risk of death within 90 days, 180 days, 360 days, etc., and of course it can also be less than 90 days, such as 80 days, 70 days, 60 days or shorter, or it can be within 1 year, 11 months, 10 months, 9 months, 8 months, 7 months, 6 months, 5 months, 4 months, 3 months, 2 months, 1 month, etc.

[0045] Because the short-term here is difficult to predict in advance, it is a time node summarized based on current clinical experience. With the passage of time or the improvement of other treatment methods, this time node or the length of time may change. It may also change according to the length of the patient's hospitalization.

[0046] Furthermore, the regression equation is a logistic regression equation.

[0047] Furthermore, the logistic regression equation is constructed and trained by the detection values ​​of biomarkers of known samples.

[0048] Furthermore, the training is performed by constructing a training set and using a training method to train the training set so as to obtain a logistic regression equation; the training set is serum samples of ACLF patients.

[0049] Furthermore, the training method is to divide the training set into a survivor group and a non-survivor group according to whether they survived in the short term, detect the biomarker levels of the survivor group and the non-survivor group respectively, obtain the detection values, perform logistic regression based on the detection values ​​of the biomarkers, analyze the relationship between the detection values ​​of the survivor group and the non-survivor group, and construct a logistic regression equation.

[0050] The training set described in the present invention can be serum samples of ACLF patients collected under any circumstances, that is, it can be serum samples of ACLF patients collected at different times, in different regions, in different numbers, different genders, different ages, and so on.

[0051] Furthermore, the short term includes any one or more of within 90 days, 180 days, 360 days, 80 days, 70 days, 60 days, three months, six months, one year, two months, and one month.

[0052] Furthermore, the data analysis module substitutes the detection value of the biomarker into the logistic regression equation to calculate the predicted value of the ACLF patient's risk of death within 360 days, thereby evaluating the ACLF patient's risk of death within 360 days.

[0053] Furthermore, the logistic regression equation is:

[0054] P = -20.584 + 0.846 * colic acid + 1.254 * NAAG + 0.454 * ureidopropionic acid - 0.468 * aminoglutamic acid - 0.385 * γ-CEHC + 0.485 * γ-3 glutamyl dipeptide + 0.447 * 3-hydroxybutylcarnitine + 0.081 * estrone sulfate + 0.524 * hydroxyproline + 0.079 * bilirubin - 0.066 * 2-hydroxydipropionic acid - 0.015 * methyl sulfide - 0.089 * hexadecanedioic acid - 0.219 * quinolinic acid + 0.942 * thyroxine;

[0055] or P = -14.286 + 0.804 * tubulocolic acid + 0.791 * NAAG + 0.448 * ureidopropionic acid - 0.396 * methylaminoglutamic acid - 0.338 * γ-CEHC + 0.434 * γ-3 glutamyl dipeptide + 0.425 * 3-hydroxybutylcarnitine + 0.107 * estrone sulfate + 0.356 * hydroxyproline + 0.091 * bilirubin - 0.082 * 2-hydroxydipropionic acid + 0.015 * methyl sulfide - 0.066 * hexadecanedioic acid - 0.154 * quinolinic acid;

[0056] or P = -14.392 + 0.799 * tubulocolic acid + 0.792 * NAAG + 0.383 * ureidopropionic acid - 0.403 * methylaminoglutamic acid - 0.352 * γ-CEHC + 0.428 * γ-3 glutamyl dipeptide + 0.388 * 3-hydroxybutylcarnitine + 0.121 * estrone sulfate + 0.353 * hydroxyproline + 0.095 * bilirubin - 0.153 * 2-hydroxydipropionic acid + 0.009 * methyl sulfide - 0.025 * hexadecanedioic acid;

[0057] or P = -14.5 + 0.794 * tubulocolic acid + 0.806 * NAAG + 0.376 * ureidopropionic acid - 0.403 * methylaminoglutamic acid - 0.352 * γ-CEHC + 0.429 * γ-3 glutamyl dipeptide + 0.378 * 3-hydroxybutylcarnitine + 0.114 * estrone sulfate + 0.362 * hydroxyproline + 0.095 * bilirubin - 0.16 * 2-hydroxydipropionic acid + 0.007 * methyl sulfide;

[0058] or P = -14.476 + 0.795 * colic acid + 0.807 * NAAG + 0.376 * ureidopropionic acid - 0.404 * aminoglutamic acid - 0.353 * γ-CEHC + 0.43 * γ-3 glutamyl dipeptide + 0.375 * '3-hydroxybutyrate carnitine' + 0.115 * estrone sulfate + 0.364 * hydroxyproline + 0.096 * bilirubin - 0.157 * 2-hydroxydipropionic acid;

[0059] or P = -14.315 + 0.777*colic acid + 0.77*NAAG + 0.342*ureidopropionic acid - 0.411*methylaminoglutamic acid - 0.352*γ-CEHC + 0.432*γ-3 glutamyl dipeptide + 0.325*3-hydroxybutyrate carnitine + 0.109*estrone sulfate + 0.357*hydroxyproline + 0.1*bilirubin;

[0060] or P = -13.898 + 0.825 * colic acid + 0.744 * NAAG + 0.322 * ureidopropionic acid + -0.391 * aminoglutamic acid + -0.269 * γ-CEHC + 0.424 * γ-3 glutamyl dipeptide + 0.345 * 3-hydroxybutyrate carnitine + 0.162 * estrone sulfate + 0.276 * hydroxyproline;

[0061] or P = -12.473 + 0.857 * tubulocolic acid + 0.781 * NAAG + 0.415 * ureidopropionic acid - 0.405 * aminoglutamic acid - 0.257 * γ-CEHC + 0.42 * γ-3 glutamyl dipeptide + 0.343 * 3-hydroxybutyrate carnitine + 0.14 * estrone sulfate;

[0062] or P = -13.147 + 0.947 * tubulocolic acid + 0.832 * NAAG + 0.435 * ureidopropionic acid - 0.407 * methylaminoglutamic acid - 0.292 * γ-CEHC + 0.428 * γ-3 glutamyl dipeptide + 0.334 * 3-hydroxybutylcarnitine;

[0063] or P = -12.838 + 0.839 * tubulocolic acid + 1.125 * NAAG + 0.49 * ureidopropionic acid - 0.39 * methylaminoglutamic acid - 0.263 * γ-CEHC + 0.39 * γ-3 glutamyl dipeptide;

[0064] or P = -10.874 + 0.685 * tubulocolic acid + 1.105 * NAAG + 0.51 * ureidopropionic acid - 0.214 * aminoglutamic acid - 0.22 * γ-CEHC;

[0065] Or p = -11.911 + 0.81 * tubulocolic acid + 0.992 * NAAG + 0.559 * ureidopropionic acid - 0.2 * methylaminoglutamic acid;

[0066] or p = -11.25 + 0.73*colic acid + 0.917*NAAG + 0.435*ureidopropionic acid;

[0067] Wherein, p represents the predicted value of the risk of death within 360 days in ACLF patients.

[0068] Furthermore, the logistic regression equation is:

[0069] p = -11.911 + 0.81 * tubulocholic acid + 0.992 * NAAG + 0.559 * ureidopropionic acid - 0.2 * methylaminoglutamic acid; or p = -11.25 + 0.73 * tubulocholic acid + 0.917 * NAAG + 0.435 * ureidopropionic acid.

[0070] Furthermore, p represents the predicted value of the risk of death within 90 days in ACLF patients.

[0071] Furthermore, when the logistic regression equation is:

[0072] When p=-11.911+0.81*colic acid+0.992*NAAG+0.559*ureidopropionic acid-0.2*aminoglutamic acid, P is greater than 0.331, which predicts that ACLF patients have a high risk of death within 90 days; when p is less than 0.331, it predicts that ACLF patients have a low risk of death within 90 days;

[0073] When the logistic regression equation is:

[0074] When p=-11.25+0.73*colic acid+0.917*NAAG+0.435*ureapropionic acid, P is greater than 0.320, which predicts that the risk of death within 90 days in ACLF patients is high; when p is less than 0.320, it predicts that the risk of death within 90 days in ACLF patients is low.

[0075] Furthermore, the logistic regression equation also includes one or more clinical indicators of age, INR (international normalized ratio), creatine, Na (sodium), ALB (albumin), N / L ratio (granulocyte / lymphocyte ratio), overt HE (overt hepatic encephalopathy), and TB (total bilirubin).

[0076] Furthermore, the system also includes a data storage module, a data input interface and a data output interface; the data storage module is used to store the detection values ​​of biomarkers; the data input interface is used to input the detection values ​​of biomarkers, and the data output interface is used to output the prediction results.

[0077] On the other hand, the present invention provides a method for predicting whether an ACLF patient has a risk of death, by analyzing the detection value of a biomarker, for predicting whether an ACLF patient has a risk of death; the biomarker includes one or more of the following: 2-hydroxydipropionic acid, hexadecanedioic acid, hydroxyproline, methyl sulfide, 3-hydroxybutylcarnitine, colic acid, estrone sulfate, N-acetylamino-glutamic acid, quinolinic acid, ureidopropionic acid, γ-CEHC, thyroxine, methylaminoglutamic acid, γ-3 glutamyl dipeptide, bilirubin.

[0078] Furthermore, the detection value of the biomarker is to detect the biomarker in the blood, serum, plasma or whole blood sample of ACLF patients, and is used to detect the presence or relative abundance or concentration of the biomarker in the sample.

[0079] Further, the biomarkers include colicolic acid, NAAG and ureidopropionic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid and methylaminoglutamic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid and γ-CEHC; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC and γ-3 glutamyl dipeptide; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide and 3-hydroxybutylcarnitine; or, The biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine and estrone sulfate; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate and hydroxyproline; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline and bilirubin; or, the biomarkers include colicolic acid , NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, and 2-hydroxydipropionic acid; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, and methyl sulfide; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline , bilirubin, 2-hydroxydipropionic acid, methyl sulfide, and hexadecanedioic acid; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, and quinolinic acid; or, the biomarkers include colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, quinolinic acid, and thyroxine.

[0080] Furthermore, the method calculates the predicted value for predicting whether an ACLF patient has a risk of death by substituting the detection value of the biomarker into a regression equation, thereby evaluating the risk of death in ACLF patients.

[0081] Furthermore, predicting whether an ACLF patient has a risk of death refers to predicting whether an ACLF patient has a risk of death in the short term.

[0082] The "short term" here can refer to whether the ACLF patient has a risk of death within 90 days, 180 days, 360 days, etc., and of course it can also be less than 90 days, such as 80 days, 70 days, 60 days or shorter, or it can be within 1 year, 11 months, 10 months, 9 months, 8 months, 7 months, 6 months, 5 months, 4 months, 3 months, 2 months, 1 month, etc.

[0083] Because the short-term here is difficult to predict in advance, it is a time node summarized based on current clinical experience. With the passage of time or the improvement of other treatment methods, this time node or the length of time may change. It may also change according to the length of the patient's hospitalization.

[0084] Furthermore, the regression equation is a logistic regression equation.

[0085] Furthermore, the logistic regression equation is constructed and trained by the detection values ​​of biomarkers of known samples.

[0086] Furthermore, the training is performed by constructing a training set and using a training method to train the training set so as to obtain a logistic regression equation; the training set is serum samples of ACLF patients.

[0087] Furthermore, the training method is to divide the training set into a survivor group and a non-survivor group according to whether the patients survived in the short term, detect the biomarker levels of the survivor group and the non-survivor group respectively, obtain the detection values, perform logistic regression based on the detection values ​​of the biomarkers, analyze the relationship between the detection values ​​of the survivor group and the non-survivor group, and construct a logistic regression equation.

[0088] Furthermore, the short term includes any one or more of within 90 days, 180 days, 360 days, 80 days, 70 days, 60 days, three months, six months, one year, two months, and one month.

[0089] Furthermore, the data analysis module substitutes the detection value of the biomarker into the logistic regression equation to calculate the predicted value of the ACLF patient's risk of death within 360 days, thereby evaluating the ACLF patient's risk of death within 360 days.

[0090] The acute-on-chronic liver failure prognosis prediction system provided by the present invention has the following beneficial effects:

[0091] 1. Screened 15 new metabolic biomarkers that can predict whether ACLF patients have the risk of death at an early stage; especially metabolic biomarkers that predict the risk of death in ACLF patients in the short term;

[0092] 2. Four biomarkers were screened and a prediction model for the short-term mortality risk of ACLF patients was constructed. The prediction performance was excellent, with the highest AUC value reaching 0.79. Further combined with clinical indicators such as age, INR, CR, Na, ALB, N / Lratio, overt HE, and TB, the AUC value can be further improved, and basically can reach above 0.8.

[0093] 3. Achieve convenient and efficient prediction, grasp the golden window of intervention for high-risk groups, reduce the economic burden for low-risk patients, and meet clinical needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 This is a flow chart of the screening, model building and validation of ACLF-related metabolomics biomarkers in Example 1;

[0095] Figure 2 Schematic diagram of significant differences in metabolites between the ACLF group and the non-ACLF group in Example 1;

[0096] Figure 3 Schematic diagram of the metabolite categories with significant effects in Example 1;

[0097] Figure 4 This is a schematic diagram of changes in liver metabolic pathways in ACLF patients in Example 1;

[0098] Figure 5 Schematic diagram of significant differences in metabolites between non-survivor group and survivor group of ACLF patients in Example 2;

[0099] Figure 6 This is a schematic diagram of the random forest ranking results of metabolites used to distinguish the non-survivor group and the survivor group of ACLF patients in Example 2;

[0100] Figure 7 Schematic diagram of significant differences in metabolites between the pre-ACLF group and the non-ACLF group in Example 3;

[0101] Figure 8 This is a schematic diagram of the random forest ranking results of metabolites that distinguish the pre-ACLF group and the non-ACLF group in Example 3;

[0102] Fig. 9 The ROC curve obtained from the detection results of the four metabolites in the discovery set in Example 5;

[0103] Fig.10 The ROC curve obtained from the detection results of the four metabolites in the validation set in Example 5;

[0104] Fig.11 The ROC curves obtained from the detection results of the three metabolites in the discovery set in Example 6;

[0105] Fig.12 The ROC curves obtained from the detection results of the three metabolites in the validation set in Example 6;

[0106] Fig.13 The evaluation performance comparison of the four metabolite models in Example 7 for predicting mortality within 90 days of ACLF in the discovery set;

[0107] Fig.14 The comparison of the evaluation performance of the four metabolite models in Example 7 in predicting the mortality rate within 90 days of ACLF in the validation set;

[0108] Fig.15 The comparison of the evaluation performance of the three metabolite models in Example 7 in predicting the risk of progression from non-ACLF to ACLF within 28 days in the discovery set;

[0109] Fig.16 The evaluation performance of the three metabolite models in Example 7 in predicting the risk of progression of non-ACLF to ACLF within 28 days in the validation set is compared.

[0110] Detailed description

[0111] (1) Diagnosis or testing

[0112] The diagnosis or detection here, prediction refers to the detection or analysis of biomarkers in the sample, or the content of the target biomarker, such as the absolute content or relative content, and then the presence or amount of the target marker is used to explain whether the individual providing the sample may have or suffer from a certain disease, or the possibility of having a certain disease. The meanings of diagnosis and detection here are interchangeable. The result of such a test or diagnosis cannot be directly used as a direct result of illness, but is an intermediate result. If a direct result is obtained, other auxiliary means such as pathology or anatomy are required to confirm that a certain disease is present. For example, the present invention provides a variety of new biomarkers that are associated with the death or survival of ACLF patients, or ALCF patients. The changes in the content of these markers are directly associated with suffering from ACL or suffering from ACLF and death.

[0113] (2) Association between markers or biomarkers and ACLF

[0114] Markers and biomarkers have the same meaning in the present invention. The connection here means that the appearance or change in the content of a certain biomarker in a sample has a direct correlation with a specific disease or the progression of the disease, such as a relative increase or decrease in the content, indicating that the possibility of suffering from the disease is higher than that of healthy people, or the progression of the disease has become more serious or has progressed from one stage to another. For example, a single marker or a combination of marker substances of the multiple new markers of the present invention can be used to predict whether a non-ACLF patient will progress to an ACLF patient, or if an ALCF patient is already a patient, whether death or non-death will occur.

[0115] If multiple different markers appear in a sample at the same time or the relative changes in their contents indicate that the probability of the person suffering from the disease is higher than that of a healthy person. That is to say, among the types of markers, some markers are strongly associated with the disease, some markers are weakly associated with the disease, or some are even not associated with a specific disease. One or more of those markers with strong correlations can be used as markers for diagnosing the disease, and those markers with weak correlations can be combined with strong markers to diagnose a disease, thereby increasing the accuracy of the test results. The disease here can be the process or progression of the disease, such as the development from a relatively good stage of a disease to a more malignant or serious stage, or even death.

[0116] For the numerous biomarkers in serum discovered by the present invention, these markers can be used to distinguish non-ALCF patients from ACLF patients, or predict whether non-ACLF will develop into ACLF; they can also be used to diagnose or predict the probability or possibility of whether ALCF patients will develop or die. The markers here can be used as a single marker for direct detection or diagnosis. The selection of such markers indicates that the relative change in the content of the marker is highly correlated with the development of non-ACLF patients into ACLF patients, or whether ALCF patients develop or die. Of course, it is understandable that one or more markers related to whether ALCF patients develop or die, or non-ACLF develops into ACLF patients, can be selected for simultaneous detection. It is normal to understand that in some ways, selecting biomarkers with strong correlation for detection or diagnosis can achieve a certain standard of accuracy, such as 60%, 65%, 70%, 80%, 85%, 90% or 95% accuracy, which can indicate that these markers can obtain the intermediate value of diagnosing a certain disease, but it does not mean that a certain disease can be directly confirmed. .

[0117] Of course, the difference marker with the larger ROC value can also be selected as the diagnostic marker. The so-called strong and weak are generally calculated and confirmed by some algorithms, such as the contribution rate or weight analysis of the marker and the probability of ACLF or ALCF death. Such calculation methods can be significance analysis (p value or FDR value) and fold change. Multivariate statistical analysis mainly includes principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA) and orthogonal partial least squares discriminant analysis (OPLS-DA).

[0118] (3) ACLF patients

[0119] Acute-on-chronic liver failure (ACLF) is a rapidly progressive disease with rapid clinical deterioration and high mortality, manifested as acute liver dysfunction in patients with chronic liver disease or cirrhosis, characterized by organ failure and high short-term mortality, which occurs in more than 25% of hospitalized patients with cirrhosis. ACLF is known to be associated with acute injuries such as alcohol, viral infection, extrahepatic organ failure, and systemic inflammation, and its pathophysiology is related to persistent inflammation, immune dysregulation with initial widespread immune activation, a state of systemic inflammatory response syndrome, and subsequent sepsis due to immunosuppression.

[0120] The so-called "chronic liver disease" mainly refers to liver disease (as opposed to health), such as hepatitis patients, alcoholic liver disease, or patients after liver transplantation, or patients with cirrhosis, or other patients with liver lesions. Chronic liver disease is a general term for a class of diseases. The causes are different, the clinical symptoms are not the same, and the treatment methods and prognosis are also different. The common characteristics of chronic liver disease are varying degrees of liver tissue necrosis and inflammatory response. Liver fibrosis may occur and eventually develop into cirrhosis. The degree of chronic liver damage can be assessed by liver tissue biopsy. Cirrhosis is also a common chronic progressive liver disease in clinical practice. It is caused by the long-term or repeated action of one or more causes, forming diffuse liver damage. The more common cause is hepatitis. It may also be related to alcoholic cirrhosis and schistosomiasis cirrhosis.

[0121] The diagnosis of chronic liver disease must meet at least one of the following conditions:

[0122] 1) Liver enzyme abnormalities that exceed a certain period of time (such as more than 1-3 months, more than 6 months, etc.); the liver enzymes include any one or more of alanine aminotransferase, aspartate aminotransferase, alkaline phosphatase or γ-transpeptidase; of course, it is understood that there are other abnormalities of enzymes related to liver enzymes. The so-called abnormality is a value that exceeds the upper limit of the normal value. The abnormality here can be a frequent abnormality or a high frequency of abnormality in a certain period of time, for example, within 3 months, there are 2 months of abnormality, or 80% of the time is abnormal.

[0123] 2) Liver elasticity ultrasound (a method of measuring liver stiffness, including but not limited to Fibroscan or Fibrotouch) indicates the presence of liver fibrosis;

[0124] 3) Liver fibrosis found in liver biopsy;

[0125] 4) Non-invasive blood index tests (including but not limited to FIB-4) indicating the presence of liver fibrosis;

[0126] 5) Clinical, imaging, digestive endoscopy or liver tissue pathology suggest the presence of cirrhosis.

[0127] (4) Non-ACLF patients

[0128] Non-ACLF patients here refer to patients who do not have acute-on-chronic liver failure, or who have clinical symptoms but do not have acute-on-chronic liver failure (ACLF), or patients with chronic liver disease who do not develop liver failure. In other words, a certain type of patients are chronic liver disease patients, some of whom may develop acute-on-chronic liver failure (ACLF), while others will not. In this way, the time from non-ACLF to ACLF can be any time, any time period within 360 days. DETAILED DESCRIPTION

[0129] The present invention is further described in detail below in conjunction with the accompanying drawings and examples. It should be noted that the following examples are intended to facilitate understanding of the present invention and do not have any limiting effect on it. The reagents used in this example are all known products and were obtained by purchasing commercially available products.

[0130] Example 1: Preliminary screening of biomarkers using metabolomics

[0131] This example established a research alliance consisting of 14 hospitals and collected 1,300 ACLF-related samples in three years to screen metabolomics biomarkers, build prediction models and verify them. The specific process is detailed in Figure 1 .

[0132] The key clinical parameters of ACLF and non-ACLF patients at admission are shown in Table 1. It can be seen that hepatitis B virus (HBV) is the main cause, accounting for about 75% of cases, and the rest are alcohol and other causes; ACLF patients have a significantly increased mortality rate without liver transplantation within 90 days, an increased proportion of organ failure, especially liver failure and coagulation failure, increased TB, ALT, AST, creatine, INR, C-reactive protein (CRP), and increased MELD-Na and CLIF-C ACLF scores (p<0.05).

[0133] Table 1. Baseline clinical parameters of ACLF and non-ACLF groups

[0134]

[0135]

[0136] In this example, 1,300 samples (including 420 ACLF and 880 non-ACLF) were randomly divided into two equal halves as a discovery set (650 samples, including 210 ACLF and 440 non-ACLF) and a validation set (650 samples, including 210 ACLF and 440 non-ACLF). The baseline serum samples of the discovery set at admission were analyzed using non-targeted metabolomics by high-performance liquid chromatography-tandem mass spectrometry.

[0137] A total of 1,072 metabolites with known structures were identified by the untargeted metabolomics platform, covering a wide range of biochemical pathways, including amino acid, lipid, carbohydrate, nucleotide, xenobiotic, and gut microbiome metabolism. After removing metabolites detected in less than 20% of the samples, the remaining 896 metabolites were analyzed. We observed that ACLF had a profound effect on the serum metabolome, with 626 of the 896 metabolites (69.9%) showing significantly changed levels between the ACLF and non-ACLF groups (FDR < 0.05, Figure 2 ), affecting almost all metabolite classes ( Figure 3 After mapping the metabolites to typical biochemical pathways, it became apparent that most changes were in pathways associated with numerous metabolic and physiological functions of the liver ( Figure 4 The main observations are as follows:

[0138] 1. Biosynthesis of membrane lipids: Among the top 100 most important metabolites ranked by FDR (FDR < 1.41*10-14), 34 are essential lipids for plasma lipoprotein particles and cell membrane structures, including phospholipids, lysophospholipids, and sphingomyelin. These lipid species were significantly reduced in the ACLF group. The liver is the main organ for complex lipid synthesis and plays a central role in lipoprotein synthesis. The reduction in lipid species suggests that lipoprotein biosynthesis may be affected by ACLF. The results of the study demonstrated that reduced levels of high-density lipoprotein and low-density lipoprotein were associated with liver failure and ACLF-related mortality.

[0139] 2. Oxidative stress and inflammation: One of the characteristics of ACLF is systemic inflammation, which is manifested by elevated white blood cell counts, CRP (C-reactive protein), proinflammatory cytokines, and damage / danger associated molecular patterns (DAMPs). The oxidative stress marker methionine sulfoxide was significantly increased (FDR = 2.54*10 -27 ); antioxidant vitamin E and vitamin E metabolite α-tocopherol (FDR = 3.43*10-17 ), γ-tocopherol / β-tocopherol (FDR = 6.32*10 -9 ), γ-carboxyethyl hydroxychromium (γ-CEHC) (FDR = 1.27*10 -23 ) and δ-CEHC (FDR = 8.36*10 -18 ) were decreased; glutathione metabolism, and the levels of metabolites in the pentose phosphate pathway increased rapidly.

[0140] 3. Thyroxine and steroid hormones: The level of thyroxine in the ACLF group was decreased (FDR = 1.94*10 -39 ) was one of the most significant changes. We found that the ACLF group had a significant decrease in cortisol and corticosterone (FDR = 1.11*10 -5 ), combined androgen and estrogen metabolites were significantly reduced in the ACLF group; the only exception was estrone 3-sulfate, which was significantly increased (FDR = 2.13*10 -23 ).

[0141] 4. Energy metabolism: The liver is the center for maintaining energy homeostasis in the body. The levels of amino acids, amino acid metabolites, urea cycle metabolites, and polyamines in the ACLF group increased, indicating that amino acid catabolism accelerated; free fatty acids in the ACLF group increased; however, 3-hydroxybutyrate, the main product of fatty acid β-oxidation, did not change. At the same time, in the ACLF group, ω-oxidation products such as dicarboxylic acids, dicarboxylic acid carnitines, hydroxy fatty acids, and hydroxy fatty acid carnitines increased, because fatty acid ω-oxidation is usually activated by impaired fatty acid β-oxidation or mitochondrial dysfunction.

[0142] 5. Gut microbiome: We observed that gut microbiome-derived metabolites were severely disrupted in serum samples, and 44 of a total of 63 gut microbiome-derived metabolites were significantly changed between the ACLF and non-ACLF groups. The affected pathways included tryptophan and tyrosine metabolism, secondary bile acids, trimethylamine oxide (TMAO) biosynthesis, and xenobiotic metabolism.

[0143] 6. Bilirubin and bile acid: Another manifestation of ACLF is elevated bilirubin (FDR = 2.14*10 -31 ) and its precursor biliverdin (FDR = 1.05*10 -28 ). Under healthy conditions, the liver removes bilirubin by glucuronidation in hepatocytes, and elevated bilirubin is a sign of liver dysfunction. In addition, taurodeoxycholate, the major bile acid in the ACLF group (FDR = 1.98*10 -15 ), chenodeoxycholic acid (FDR = 4.53*10 -17 ), glycocholate (FOR = 0.008) and taurocholate (FRR = 5.75*10-5 ) is increased. Primary bile acids are synthesized in the liver, stored in the gallbladder, and secreted into the intestine for food digestion, and then actively circulate back to the liver. Elevated primary bile acids are associated with a variety of liver diseases.

[0144] Example 2: Development and validation of metabolic biomarkers for predicting short-term mortality in ACLF patients

[0145] In this example, based on the metabolic biomarkers obtained in the initial screening of Example 1, 420 ACLF patient samples were studied to screen and verify metabolic biomarkers that can predict the mortality rate of ACLF within 90 days. Among the 420 patients diagnosed with ACLF at admission, 151 died within 90 days. We randomly divided the samples into two equal halves as a discovery set (134 survivors and 76 non-survivors) and a validation set (135 survivors and 75 non-survivors). The discovery set was used to screen markers, and the validation set was used for validation.

[0146] According to the key clinical parameters of the two groups at admission, non-survivors had significantly higher proportions of liver failure and coagulation failure, increased TB, INR, white blood cell count, neutrophil / lymphocyte ratio, and higher MELD-Na and CLIF-C ACLF scores (p<0.05).

[0147] The test results showed that 234 metabolites were significantly changed between the non-survivor group and the survivor group (FDR<0.05, Figure 5 ), the metabolites with significant differences were mainly related to amino acid degradation, fatty acid ω-oxidation, oxidative stress pathway and hormone pathway.

[0148] We used random forest to rank the metabolites with the greatest ability to distinguish between the non-survivor and survivor groups, and the ranking results are shown in Figure 6 . The top 100 metabolites were evaluated as potential biomarkers by t-test and random forest ranking. We focused on metabolites with commercially available chemical standards and isotope standards because they are more suitable for targeted analysis development. The top 15 metabolites are shown in Table 2. In order to identify different metabolites, the fold change, p-value and estimated false discovery rate (FDR) of each metabolite were calculated. The two-sided Wilcoxon rank sum test was applied, and the p-value was adjusted by the Benjamini Hochberg (BH) multiple comparison test to calculate the estimated FDR. The 15 metabolites were analyzed to be significantly correlated with the mortality rate within 90 days of ACLF. The analysis results are shown in Table 3.

[0149] Table 2. 15 metabolic biomarkers associated with mortality within 90 days of ACLF

[0150]

[0151]

[0152] Table 3. Comparison of the correlation test results between 15 metabolites and mortality within 90 days of ACLF

[0153] Serial number Biomarkers RI MASS p-value FDR Fold Change 1 2-Hydroxydipropionic acid 3000 161.0455 0.00043 0.005931 2.106265 2 Hexadecanedioic acid 4615 285.2071 0.000324 0.005023 1.699285 3 Hydroxyproline 1064 132.0655 0.051648837 0.124067983 1.155849411 4 Methyl sulfide 1272 166.0533 2.49467E-06 0.00319318 2.006169391 5 3-Hydroxybutylcarnitine 2400 248.1493 0.000168 0.003602 1.980332 6 Colonic acid 2200 130.0863 7.15E-05 0.002464 1.593726 7 Estrone sulfate 4417.3 349.1115 0.000103749 0.002789889 2.722069235 8 N-Acetylamino-Glutamic Acid 1035 305.098 1.72325E-05 0.00110288 2.119413468 9 Quinolinic acid 845 168.0291 0.00073 0.007879 1.585904 10 Urea propionic acid 875 133.0608 1.15E-06 0.000257 1.717584 11 γ-CEHC 3843 263.1289 3.90293E-05 0.001520447 0.357239443 12 Thyroxine 4808 775.6794 0.028389914 0.081248518 0.737049142 13 Methylaminoglutamate 1880 175.0713 0.000873 0.009097 1.824112 14 Gamma-3 Glutamyl Dipeptide 1825 293.1143 8.13E-05 0.002602 1.562759 15 Bilirubin 1840 585.2708 0.0099778203 0.198669489 1.200212292

[0154] The ROC curve method was used to evaluate the efficacy of 15 metabolic biomarkers in predicting 90-day mortality in ACLF in the discovery set and validation set (including 269 survivors and 151 non-survivors). The results are shown in Table 4.

[0155] Table 4. ROC analysis results of 15 metabolites in predicting mortality within 90 days of ACLF

[0156]

[0157]

[0158] The correlation between the concentration changes of the 15 biomarkers and whether or not the patient survived within 90 days of ACLF can be distinguished by the RI, MASS, p-value, FDR, Fold Change, etc. in Table 3, or by the AUC value in Table 4, among which the FDR value and AUC value are the most intuitive and obvious. The lower the FDR value, the greater the impact of the non-survivors of ACLF within 90 days on this indicator relative to the survivors of ACLF within 90 days. The higher the AUC value, the more accurately the biomarker can distinguish between the non-survivors and survivors of ACLF within 90 days.

[0159] As can be seen from Table 3, the concentration changes of the 15 biomarkers were significantly correlated with the survival of ACLF within 90 days, among which ureidopropionic acid had the highest correlation, with an FDR value of 0.000257, followed by N-acetylamino-glutamic acid, with an FDR value of 0.00110288.

[0160] As can be seen from Table 4, the concentration change of any one of the 15 biomarkers can be used to distinguish between non-survivors and survivors of ACLF within 90 days, and the AUC value can reach above 0.6, which has a high accuracy. Among them, ureidopropionic acid has the highest AUC value of 0.719, followed by colicoline acid and estrone sulfate, with AUC values ​​of 0.694.

[0161] At the same time, the 15 biomarkers obtained by screening in this example are used to predict whether ACLF has a risk of death within 360 days, and good results can also be obtained, with AUC values ​​reaching above 0.5.

[0162] Example 3: Development and validation of metabolic biomarkers for predicting the risk of progression to ACLF in non-ACLF patients in the short term

[0163] In this example, based on the metabolic biomarkers obtained in the initial screening of Example 1, 880 non-ACLF patient samples were studied to screen and verify metabolic biomarkers that can predict the risk of progression from non-ACLF to ACLF within 28 days. The non-ACLF group consisted of 880 patients, of which 108 patients developed ACLF (pre-ACLF) within 28 days after being diagnosed with non-ACLF. We randomly divided the samples into two equal halves as discovery and validation sets (54 pre-ACLF and 386 non-ACLF in each group). The discovery set was used to screen markers, and the validation set was used for validation.

[0164] Patients who progressed to ACLF had significantly higher TB, INR, neutrophil / lymphocyte ratio, and higher MELD-Na and CLIF-C ACLF-D scores compared with patients whose disease was stable during this period (p<0.05).

[0165] A total of 89 metabolites showed significant changes in levels between the pre-ACLF group and the non-ACLF group (FDR < 0.05, Figure 7 Metabolites related to fatty acid ω-oxidation, oxidative stress pathways, bilirubin metabolism, bile acid metabolism, and steroid hormone pathways comprised the majority of significant changes. Using a similar approach to ACLF mortality biomarker identification, we used random forests to rank metabolites to distinguish between the pre-ACLF and non-ACLF groups ( Figure 8 ). The top 89 metabolites ranked by t-test and / or random forest were further evaluated as potential biomarkers, and metabolites without commercially available chemical standards and isotope standards were excluded. The top 15 metabolites were selected, and these 15 metabolites were consistent with the metabolites that predicted mortality within 90 days of ACLF. In order to identify different metabolites, the fold change, p-value and estimated false discovery rate (FDR) of each metabolite were calculated. The two-sided Wilcoxon rank sum test was applied, and the p-value was adjusted by the Benjamini Hochberg (BH) multiple comparison test to calculate the estimated FDR. The 15 metabolites were analyzed to be significantly associated with the risk of progression from non-ACLF to ACLF within 28 days. The analysis results are shown in Table 5.

[0166] Table 5. Comparison of the correlation test results between 15 metabolites and the risk of progression from non-ACLF to ACLF within 28 days

[0167] Serial number Biomarkers RI MASS p-value FDR Fold Change 1 2-Hydroxydipropionic acid 3000 161.0455 0.015924 0.092754 1.663287 2 Hexadecanedioic acid 4615 285.2071 0.018446 0.102023 1.473388 3 Hydroxyproline 1064 132.0655 0.751630289 0.867244843 1.021273795 4 Methyl sulfide 1272 166.0533 0.000641548 0.01957662 1.357941699 5 3-Hydroxybutylcarnitine 2400 248.1493 0.33344 0.563702 1.160571 6 Colonic acid 2200 130.0863 0.148843 0.353749 1.190887 7 Estrone sulfate 4417.3 349.1115 0.005572803 0.052845904 2.059286596 8 N-Acetylamino-Glutamic Acid 1035 305.098 0.913227761 0.959263862 0.999207206 9 Quinolinic acid 845 168.0291 0.875164 0.939098 1.039392 10 Urea propionic acid 875 133.0608 1.07E-05 0.001914 1.556914 11 γ-CEHC 3843 263.1289 0.003324966 0.037239622 0.465725777 12 Thyroxine 4808 775.6794 2.1074E-05 0.0027813 0.692212526 13 Methylaminoglutamate 1880 175.0713 0.185424 0.397093 1.213627 14 Gamma-3 Glutamyl Dipeptide 1825 293.1143 0.044516 0.16392 1.10889 15 Bilirubin 1840 585.2708 0.000184722 0.008711105 1.409553609

[0168] The ROC curve method was used to evaluate the efficacy of 15 metabolic biomarkers in predicting the risk of progression from non-ACLF to ACLF within 28 days in the discovery set and validation set (including 108 pre-ACLF (ACLF developed within 28 days after the diagnosis of non-ACLF) and 772 non-ACLF (no ACLF developed within 28 days after the diagnosis of non-ACLF). The results are shown in Table 6.

[0169] Table 6. Analysis results of 15 metabolites ROC predicting the risk of progression from non-ACLF to ACLF within 28 days

[0170] Serial number Biomarkers AUC value 1 2-Hydroxydipropionic acid 0.682 2 Hexadecanedioic acid 0.649 3 Hydroxyproline 0.683 4 Methyl sulfide 0.663 5 3-Hydroxybutylcarnitine 0.685 6 Colonic acid 0.684 7 Estrone sulfate 0.649 8 N-Acetylamino-Glutamic Acid 0.601 9 Quinolinic acid 0.644 10 Urea propionic acid 0.723 11 γ-CEHC 0.724 12 Thyroxine 0.689 13 Methylaminoglutamate 0.695 14 Gamma-3 Glutamyl Dipeptide 0.614 15 Bilirubin 0.668

[0171] The correlation between the concentration changes of the 15 biomarkers and whether non-ACLF progresses to ACLF within 28 days can be distinguished by RI, MASS, p-value, FDR, Fold Change, etc. in Table 5, and by AUC values ​​in Table 6, among which FDR and AUC values ​​are the most intuitive and obvious. The lower the FDR value, the greater the impact of the non-ACLF population that progresses to ACLF within 28 days on the indicator relative to the non-ACLF population that does not progress to ACLF within 28 days. The higher the AUC value, the more accurately the biomarker can distinguish whether non-ACLF progresses to ACLF within 28 days.

[0172] As can be seen from Table 5, the concentration changes of the 15 biomarkers were significantly correlated with whether non-ACLF progressed to ACLF within 28 days, among which ureidopropionic acid had the highest correlation, with an FDR value of 0.001914, followed by thyroxine, with an FDR value of 0.0027813.

[0173] As can be seen from Table 6, the concentration change of any one of the 15 biomarkers can be used to distinguish whether non-ACLF progresses to ACLF within 28 days, and its AUC value can reach above 0.6, which has high accuracy. Among them, γ-CEHC has the highest AUC value of 0.724, followed by ureidopropionic acid, with an AUC value of 0.723.

[0174] At the same time, the 15 biomarkers screened in this example are used to predict the risk of non-ACLF progressing to ACLF within 180 days, and good results can also be obtained, with AUC values ​​reaching above 0.5.

[0175] Example 4: Comparison of targeted and non-targeted detection performance of 15 metabolite combinations

[0176] To translate metabolomics findings into actual clinical applications, we used LCMS to perform non-targeted and targeted detection of 15 biomarkers (predicting mortality within 90 days of ACLF and the risk of progression to ACLF within 28 days of non-ACLF): γ-CEHC, N-acetylamino-glutamate (NAAG), colicoline, thyroxine, and ureidopropionic acid. The accuracy, precision, stability, and predictive performance of the assays of these 15 biomarkers were further evaluated.

[0177] Targeted detection uses standards for absolute qualitative and quantitative analysis of the substances to be tested. The specific analysis method is as follows: the plasma sample is added with an internal standard solution containing a stable isotope-labeled analyte analog, and protein precipitation is performed with acidified methanol (1% formic acid in methanol). After centrifugation, the supernatant is removed, evaporated and reconstituted. An aliquot of the extract is injected into the SciexExionLC / Sciex 5500+ triple quadrupole mass spectrometer LC-MS / MS system, and C18 reverse phase chromatography is used to analyze 3-ureidopropionate, methylaminoglutamate, NAAG, thyroxine and γ-CEHC in a water (0.1% formic acid) / acetonitrile gradient within 6.0 minutes, and the mass spectrometer is operated in negative mode; the second aliquot is injected into the same LC-MS / MS system for the analysis of pipecolic acid. C18 reverse phase ion pair chromatography, water (0.05% non-fluorovaleric acid) / acetonitrile (0.05% non-fluorovaleric acid) gradient 4.0 minutes, the mass spectrometer is operated in positive mode.

[0178] Targeted Detection The assay has a short run time on LCMS and is suitable for rapid data turnaround and large-scale detection for clinical testing.

[0179] During targeted analysis, each sample was spiked with authentic chemical and isotopic standards, and absolute quantification of metabolites in the sample was achieved by comparison with calibration curves of metabolite standards at eight concentrations. The eight concentrations were one at the beginning and end of the analytical run, and six single-concentration quality control samples were added to monitor analytical performance. The peak areas of the corresponding parent ion to product ion transitions were measured against the peak areas of the parent ion to product ion transitions of the corresponding internal standards, and quantification was performed using weighted linear least squares regression analysis.

[0180] In the initial screening of metabolic biomarkers, non-targeted determination is used for screening. In this embodiment, targeted and non-targeted detection are performed simultaneously to verify the correlation of changes in the two detection methods, proving that the overall trends of non-targeted detection and targeted detection are consistent, and specific detection results can be converted through correlation coefficients. Therefore, in the subsequent embodiments, targeted detection is used for more accurate quantification in the model construction and analysis of metabolic biomarkers.

[0181] Quantitative data generated by targeted assays showed correlation of changes with untargeted metabolomics data, with correlation coefficients ranging from 0.502 (aminoglutamate) to 0.725 (colic acid).

[0182] Example 5: Development and validation of a model for predicting mortality within 90 days of ACLF

[0183] In this example, the 15 metabolites screened in Example 2 were combined and logistic regression was used to construct a model for predicting the mortality rate of ACLF within 90 days, hoping to construct the model with as few metabolites as possible and have a relatively ideal prediction performance.

[0184] (1) Model of 15 metabolite combinations

[0185] Firstly, 15 metabolites were combined to construct a model for predicting mortality within 90 days of ACLF, and its predictive performance was examined.

[0186] 15 metabolites: endotoxin, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, quinolinic acid, thyroxine

[0187] The model formula is:

[0188] P=-20.584+0.846*colic acid+1.254*NAAG+0.454*ureidopropionic acid-0.468*methylaminoglutamic acid-0.385*γ-CEHC+0.485*γ-3-glutamyl dipeptide+0.447*3-hydroxybutylcarnitine+0.081*estrone sulfate+0.524*hydroxyproline+0.079*bilirubin-0.066*2-hydroxydipropionic acid-0.015*methyl sulfide-0.089*hexadecanediol acid-0.219*quinolinic acid+0.942*thyroxine

[0189] In the discovery set, the model constructed with 15 metabolites had an AUC of 0.829 for predicting the 90-day mortality of ACLF patients; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.734.

[0190] (2) Model of 14 metabolite combinations

[0191] Based on the 15 metabolites, one metabolite was randomly eliminated, and there were 15 elimination methods in total. The remaining 14 metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the 15 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 14 metabolite combinations:

[0192] Colonic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, quinolinic acid

[0193] The model formula is:

[0194] P=-14.286+0.804*colic acid+0.791*NAAG+0.448*ureidopropionic acid-0.396*methylaminoglutamic acid-0.338*γ-CEHC+0.434*γ-3-glutamyl dipeptide+0.425*3-hydroxybutylcarnitine+0.107*estrone sulfate+0.356*hydroxyproline+0.091*bilirubin-0.082*2-hydroxydipropionic acid+0.015*methyl sulfide-0.066*hexadecanedioic acid-0.154*quinolinic acid

[0195] In the discovery set, the AUC of the model constructed with these 14 metabolites for predicting the 90-day mortality of ACLF patients was 0.822; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.760.

[0196] (3) Model of 13 metabolite combinations

[0197] Based on the 14 metabolites, one metabolite was randomly eliminated, and there were 14 elimination methods in total. The remaining 13 metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the 14 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 13 metabolite combinations:

[0198] Colonic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid

[0199] The model formula is:

[0200] P=-14.392+0.799*colic acid+0.792*NAAG+0.383*ureidopropionic acid-0.403*methylaminoglutamic acid-0.352*γ-CEHC+0.428*γ-3 glutamyl dipeptide+0.388*3-hydroxybutylcarnitine+0.121*estrone sulfate+0.353*hydroxyproline+0.095*bilirubin-0.153*2-hydroxydipropionic acid+0.009*methyl sulfide-0.025*hexadecanediol acid

[0201] In the discovery set, the AUC of the model constructed with these 13 metabolites for predicting the 90-day mortality of ACLF patients was 0.819; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.726.

[0202] (4) Model of 12 metabolite combinations

[0203] Based on the 13 metabolites, one metabolite was randomly eliminated, and there were 13 elimination methods in total. The remaining 12 metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the 13 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 12 metabolite combinations:

[0204] Colonic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide

[0205] The model formula is:

[0206] P=-14.5+0.794*colic acid+0.806*NAAG+0.376*ureidopropionic acid-0.403*methylaminoglutamic acid-0.352*γ-CEHC+0.429*γ-3-glutamyl dipeptide+0.378*3-hydroxybutylcarnitine+0.114*estrone sulfate+0.362*hydroxyproline+0.095*bilirubin-0.16*2-hydroxydipropionic acid+0.007*methyl sulfide

[0207] In the discovery set, the AUC of the model constructed with these 12 metabolites for predicting the 90-day mortality of ACLF patients was 0.819; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.763.

[0208] (5) Model of 11 metabolite combinations

[0209] Based on the 12 metabolites, one metabolite was randomly eliminated, and there were 12 elimination methods in total. The remaining 11 metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the 12 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 11 metabolite combinations:

[0210] Colonic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid

[0211] The model formula is:

[0212] P=-14.476+0.795*colic acid+0.807*NAAG+0.376*ureidopropionic acid-0.404*aminoglutamic acid-0.353*γ-CEHC+0.43*γ-3 glutamyl dipeptide+0.375*'3-hydroxybutyrate carnitine'+0.115*estrone sulfate+0.364*hydroxyproline+0.096*bilirubin-0.157*2-hydroxydipropionic acid

[0213] In the discovery set, the AUC of the model constructed with these 11 metabolites for predicting the 90-day mortality of ACLF patients was 0.819; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.763.

[0214] (6) Model of 10 metabolite combinations

[0215] Based on the 11 metabolites, one metabolite was randomly eliminated. There were 11 elimination methods in total. The remaining 10 metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the 11 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 10 metabolite combinations:

[0216] Colonic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin

[0217] The model formula is:

[0218] P=-14.315+0.777*colic acid+0.77*NAAG+0.342*ureidopropionic acid-0.411*aminoglutamic acid-0.352*γ-CEHC+0.432*γ-3 glutamyl dipeptide+0.325*3-hydroxybutyrate carnitine+0.109*estrone sulfate+0.357*hydroxyproline+0.1*bilirubin

[0219] In the discovery set, the AUC of the model constructed with these 10 metabolites for predicting the 90-day mortality of ACLF patients was 0.817; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.763.

[0220] (7) Model of 9 metabolite combinations

[0221] Based on the 10 metabolites, one metabolite was randomly eliminated, and there were 10 elimination methods in total. The remaining 9 metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the 10 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 9 metabolite combinations:

[0222] Colonic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline

[0223] The model formula is:

[0224] P=-13.898+0.825*colic acid+0.744*NAAG+0.322*ureidopropionic acid+-0.391*aminoglutamic acid+-0.269*γ-CEHC+0.424*γ-3 glutamyl dipeptide+0.345*3-hydroxybutyrate carnitine+0.162*estrone sulfate+0.276*hydroxyproline

[0225] In the discovery set, the AUC of the model constructed with these 9 metabolites for predicting the 90-day mortality of ACLF patients was 0.815; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.763.

[0226] (8) Model of 8 metabolite combinations

[0227] Based on the 9 metabolites, one metabolite was randomly eliminated, and there were 9 elimination methods in total. The remaining 8 metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the 9 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 8 metabolite combinations:

[0228] Colonic acid, NAAG, ureidopropionic acid, aminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate

[0229] The model formula is:

[0230] P=-12.473+0.857*colic acid+0.781*NAAG+0.415*ureidopropionic acid-0.405*aminoglutamic acid-0.257*γ-CEHC+0.42*γ-3 glutamyl dipeptide+0.343*3-hydroxybutyric acid carnitine+0.14*estrogen sulfate

[0231] In the discovery set, the AUC of the model constructed with these 8 metabolites for predicting the 90-day mortality of ACLF patients was 0.813; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.759.

[0232] (9) Model of 7 metabolite combinations

[0233] Based on the 8 metabolites, one metabolite was randomly eliminated, and there were 8 elimination methods in total. The remaining 7 metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the 8 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the 7 metabolite combinations shown below:

[0234] Colonic acid, NAAG, ureidopropionic acid, aminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine

[0235] The model formula is:

[0236] P = -13.147 + 0.947 * colic acid + 0.832 * NAAG + 0.435 * ureidopropionic acid - 0.407 * aminoglutamic acid - 0.292 * γ-CEHC + 0.428 * γ-3 glutamyl dipeptide + 0.334 * 3-hydroxybutylcarnitine

[0237] In the discovery set, the AUC of the model constructed with these 7 metabolites for predicting the 90-day mortality of ACLF patients was 0.812; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.755.

[0238] (10) Model of 6 metabolite combinations

[0239] Based on the 7 metabolites, 1 metabolite was randomly eliminated, and there were 7 elimination methods in total. The remaining 6 metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the 7 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 6 metabolite combinations:

[0240] Colonic acid, NAAG, ureidopropionic acid, aminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide

[0241] The model formula is:

[0242] P=-12.838+0.839*colic acid+1.125*NAAG+0.49*ureidopropionic acid-0.39*aminoglutamic acid-0.263*γ-CEHC+0.39*γ-3 glutamyl dipeptide

[0243] In the discovery set, the AUC of the model constructed with these six metabolites for predicting the 90-day mortality of ACLF patients was 0.802; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.746.

[0244] (11) Model of five metabolite combinations

[0245] Based on the six metabolites, one metabolite was randomly eliminated, and there were six elimination methods in total. The remaining five metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the six elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following five metabolite combinations:

[0246] Colonic acid, NAAG, ureidopropionic acid, aminoglutamic acid, γ-CEHC

[0247] The model formula is:

[0248] P=-10.874+0.685*colic acid+1.105*NAAG+0.51*ureidopropionic acid-0.214*aminoglutamic acid-0.22*γ-CEHC

[0249] In the discovery set, the AUC of the model constructed with these five metabolites for predicting the 90-day mortality of ACLF patients was 0.793; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.736.

[0250] (12) Model of four metabolite combinations

[0251] Based on the five metabolites, one metabolite was randomly eliminated. There were five elimination methods in total. The remaining four metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the five elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following four metabolite combinations:

[0252] Colonic acid, NAAG, urea propionic acid, methylaminoglutamic acid

[0253] The model formula is:

[0254] p=-11.911+0.81*colic acid+0.992*NAAG+0.559*ureidopropionic acid-0.2*methylaminoglutamic acid

[0255] In the discovery set, the AUC of the model constructed with these four metabolites for predicting the 90-day mortality of ACLF patients was 0.790; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.724.

[0256] (13) Model of three metabolite combinations

[0257] Based on the four metabolites, one metabolite was randomly eliminated. There were four elimination methods in total. The remaining three metabolites could be used to construct a model for predicting the mortality rate of ACLF within 90 days. The prediction models constructed by the four elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the combination of three metabolites as shown below:

[0258] Colonic acid, NAAG, ureidopropionic acid

[0259] The model formula is:

[0260] p=-11.25+0.73*colic acid+0.917*NAAG+0.435*ureidopropionic acid

[0261] In the discovery set, the AUC of the model constructed with these three metabolites for predicting the 90-day mortality of ACLF patients was 0.775; in the validation set, the AUC for predicting the 90-day mortality of ACLF patients was 0.722.

[0262] It can be seen that the model constructed by these three metabolites for predicting the mortality rate of ACLF patients within 90 days can also achieve good prediction results, and the required metabolites are few and the cost is low.

[0263] However, when one metabolite was further eliminated and two metabolites remained, the AUC for predicting the 90-day mortality of ACLF patients in both the discovery and validation sets decreased significantly, both to below 0.7. Therefore, it is not recommended to use a combination of two metabolites to construct a model for predicting the 90-day mortality of ACLF patients.

[0264] In summary, as the number of metabolite types decreases, the AUC values ​​in the discovery set and validation set will decrease slightly when the constructed prediction model predicts the 90-day mortality of ACLF patients. However, too many types of metabolites will add a burden to the detection process. After weighing the pros and cons, the models constructed with 3 and 4 metabolites are preferred.

[0265] When the model was constructed using four indicators, namely, colic acid, N-acetyl-aspartate glutamate (NAAG), ureidopropionic acid, and aminoglutamate, it could still maintain a high AUC value in the discovery set and validation set. In the discovery set, the AUC for predicting the mortality rate of ACLF patients within 90 days was 0.79 ( Fig. 9 ), which was accurately reproduced in the validation set, with an AUC of 0.724 ( Fig.10 ). The study showed that for the model constructed with the four metabolites, the prediction value P was greater than 0.331, which predicted that ACLF patients had a high risk of death within 90 days; p was less than 0.331, which predicted that ACLF patients had a low risk of death within 90 days.

[0266] When the model constructed by using three indicators, colic acid, NAAG, and urea propionic acid, can still maintain a high AUC value in the discovery set and validation set. In the discovery set, the AUC for predicting the mortality rate of ACLF patients within 90 days was 0.775. This result was accurately reproduced in the validation set, with an AUC of 0.722 in the validation set. The study showed that for the model constructed by the three metabolites, the prediction value P was greater than 0.320, predicting a high risk of death within 90 days for ACLF patients; p was less than 0.320, predicting a low risk of death within 90 days for ACLF patients.

[0267] At the same time, this example uses the three metabolite models obtained by screening (p=-11.25+0.73*colic acid+0.917*NAAG+0.435*ureidopropionic acid) and the four metabolite models (p=-11.911+0.81*colic acid+0.992*NAAG+0.559*ureidopropionic acid-0.2*methylaminoglutamic acid) to predict whether ACLF has a risk of death within 360 days, and also obtains good results, and the AUC values ​​can reach above 0.6.

[0268] Example 6. Development and validation of a risk model for predicting progression of non-ACLF to ACLF within 28 days

[0269] In this example, the 15 metabolites screened in Example 3 were combined and logistic regression was used to construct a risk model for predicting progression of non-ACLF to ACLF within 28 days. It is hoped that the model can be constructed with as few metabolites as possible and have a relatively ideal prediction performance.

[0270] (1) Model of 15 metabolite combinations

[0271] Firstly, 15 metabolites were combined to construct a model for predicting the progression of non-ACLF to ACLF within 28 days, and its predictive performance was investigated.

[0272] 15 metabolites: ureidopropionic acid, γ-CEHC, thyroxine, guanylic acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid, estrone sulfate, hydroxyproline, quinolinic acid, γ-3 glutamyl dipeptide, methylaminoglutamic acid, bilirubin, hexadecanedioic acid, NAAG, methyl sulfide

[0273] The model formula is:

[0274] P=-3.271+0.716*ureidopropionic acid-0.512*γ-CEHC-1.1*thyroxine+0.357*colic acid+0.134*3-hydroxybutylcarnitine-1.225*2-hydroxydipropionic acid+0.029*estrone sulfate+0.168*hydroxyproline-0.04*quinolinic acid+0.078*γ-3 glutamyl dipeptide-0.014*methylaminoglutamic acid-0.03*bilirubin+0.519*hexadecanediol acid-0.139*methyl sulfide+0.653*NAAG

[0275] In the discovery set, the model constructed with 15 metabolites had an AUC of 0.838 for predicting progression of non-ACLF to ACLF within 28 days; in the validation set, the AUC for predicting progression of non-ACLF to ACLF within 28 days was 0.741.

[0276] (2) Model of 14 metabolite combinations

[0277] Based on the 15 metabolites, one metabolite was randomly eliminated, and there were 15 elimination methods in total. The remaining 14 metabolites could be used to construct a risk model for predicting the progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the 15 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 14 metabolite combinations:

[0278] Urea propionic acid, γ-CEHC, thyroxine, colic acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid, estrone sulfate, hydroxyproline, quinolinic acid, γ-3 glutamyl dipeptide, methylaminoglutamic acid, methyl sulfide, bilirubin, hexadecanedioic acid

[0279] The model formula is:

[0280] P=-1.518+0.724*ureidopropionic acid-0.494*γ-CEHC-1.216*thyroxine+0.409*colic acid+0.157*3-hydroxybutylcarnitine-1.131*2-hydroxydipropionic acid+0.051*estrone sulfate+0.202*hydroxyproline-0.024*quinolinic acid+0.11*γ-3 glutamyl dipeptide-0.007*methylaminoglutamic acid+0.009*bilirubin+0.479*hexadecanediol acid-0.17*methyl sulfide

[0281] In the discovery set, the AUC of the model constructed with these 14 metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.830; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.746.

[0282] (3) Model of 13 metabolite combinations

[0283] Based on the 14 metabolites, one metabolite was randomly eliminated, and there were 14 elimination methods in total. The remaining 13 metabolites could be used to construct a risk model for predicting the progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the 14 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 13 metabolite combinations:

[0284] Urea propionic acid, γ-CEHC, thyroxine, colonic acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid, estrone sulfate, hydroxyproline, quinolinic acid, γ-3 glutamyl dipeptide, methylaminoglutamic acid, bilirubin, hexadecanedioic acid

[0285] The model formula is:

[0286] P=-2.436+0.726*ureidopropionic acid+-0.498*γ-CEHC-1.18*thyroxine+0.382*colic acid+0.195*3-hydroxybutylcarnitine-1.121*2-hydroxydipropionic acid+0.044*estrone sulfate+0.199*hydroxyproline-0.009*quinolinic acid+0.046*γ-3 glutamyl dipeptide-0.014*methylaminoglutamic acid-0.012*bilirubin+0.446*hexadecanediol acid

[0287] In the discovery set, the AUC of the model constructed with these 13 metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.833; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.748.

[0288] (4) Model of 12 metabolite combinations

[0289] Based on the 13 metabolites, one metabolite was randomly eliminated, and there were 13 elimination methods in total. The remaining 12 metabolites could be used to construct a risk model for predicting the progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the 13 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 12 metabolite combinations:

[0290] Urea propionic acid, γ-CEHC, thyroxine, colic acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid, estrone sulfate, hydroxyproline, quinolinic acid, γ-3 glutamyl dipeptide, methylaminoglutamic acid, bilirubin

[0291] The model formula is:

[0292] P=-2.837+0.819*ureidopropionic acid-0.514*γ-CEHC-1.019*thyroxine+0.399*colic acid+0.362*3-hydroxybutylcarnitine-0.932*2-hydroxydipropionic acid+0.173*estrone sulfate+0.235*hydroxyproline-0.08*quinolinic acid+0.074*γ-3 glutamyl dipeptide+0.001*methylaminoglutamic acid-0.02*bilirubin

[0293] In the discovery set, the AUC of the model constructed with these 12 metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.824; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.751.

[0294] (5) Model of 11 metabolite combinations

[0295] Based on the 12 metabolites, one metabolite was randomly eliminated, and there were 12 elimination methods in total. The remaining 11 metabolites could be used to construct a risk model for predicting the progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the 12 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 11 metabolite combinations:

[0296] Urea propionic acid, γ-CEHC, thyroxine, colic acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid, estrone sulfate, hydroxyproline, quinolinic acid, γ-3 glutamyl dipeptide, methylaminoglutamic acid

[0297] The model formula is:

[0298] P=-2.894+0.816*ureidopropionic acid-0.512*γ-CEHC-1.016*thyroxine+0.395*colic acid+0.362*3-hydroxybutylcarnitine-0.93*2-hydroxydipropionic acid+0.166*estrone sulfate+0.238*hydroxyproline-0.076*quinolinic acid+0.071*γ-3 glutamyl dipeptide+0.001*methylaminoglutamic acid

[0299] In the discovery set, the AUC of the model constructed with these 11 metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.824; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.751.

[0300] (6) Model of 10 metabolite combinations

[0301] Based on the 11 metabolites, one metabolite was randomly eliminated, and there were 11 elimination methods in total. The remaining 10 metabolites could be used to construct a risk model for predicting the progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the 11 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 10 metabolite combinations:

[0302] Urea propionic acid, γ-CEHC, thyroxine, colic acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid, estrone sulfate, hydroxyproline, quinolinic acid, γ-3 glutamyl dipeptide

[0303] The model formula is:

[0304] P=-2.899+0.816*ureidopropionic acid-0.512*γ-CEHC-1.015*thyroxine+0.396*colic acid+0.362*3-hydroxybutylcarnitine-0.929*2-hydroxydipropionic acid+0.166*estrone sulfate+0.238*hydroxyproline-0.076*quinolinic acid+0.071*γ-3 glutamyl dipeptide

[0305] In the discovery set, the AUC of the model constructed with these 10 metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.824; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.751.

[0306] (7) Model of 9 metabolite combinations

[0307] Based on the 10 metabolites, one metabolite was randomly eliminated, and there were 10 elimination methods in total. The remaining 9 metabolites could be used to construct a risk model for predicting the progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the 10 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 9 metabolite combinations:

[0308] Urea propionic acid, γ-CEHC, thyroxine, guanine acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid, estrone sulfate, hydroxyproline, quinolinic acid

[0309] The model formula is:

[0310] P=-2.647+0.828*ureidopropionic acid-0.512*γ-CEHC-1.039*thyroxine+0.388*colic acid+0.347*3-hydroxybutylcarnitine-0.926*2-hydroxydipropionic acid+0.163*estrone sulfate+0.241*hydroxyproline-0.047*quinolinic acid

[0311] In the discovery set, the AUC of the model constructed with these 9 metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.825; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.750.

[0312] (8) Model of 8 metabolite combinations

[0313] Based on the 9 metabolites, one metabolite was randomly eliminated, and there were 9 elimination methods in total. The remaining 8 metabolites could be used to construct a risk model for predicting the progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the 9 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 8 metabolite combinations:

[0314] Urea propionic acid, γ-CEHC, thyroxine, guanylic acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid, estrone sulfate, hydroxyproline

[0315] The model formula is:

[0316] P=-2.647+0.816*ureidopropionic acid-0.513*γ-CEHC-1.049*thyroxine+0.393*colic acid+0.34*3-hydroxybutylcarnitine-0.939*2-hydroxydipropionic acid+0.164*estrone sulfate+0.23*hydroxyproline

[0317] In the discovery set, the AUC of the model constructed with these 8 metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.825; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.750.

[0318] (9) Model of 7 metabolite combinations

[0319] Based on the 8 metabolites, one metabolite was randomly eliminated, and there were 8 elimination methods in total. The remaining 7 metabolites could be used to construct a risk model for predicting the progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the 8 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 7 metabolite combinations:

[0320] Urea propionic acid, γ-CEHC, thyroxine, guanylic acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid, estrone sulfate

[0321] The model formula is:

[0322] P=-0.998+0.856*ureidopropionic acid-0.502*γ-CEHC-1.104*thyroxine+0.408*colic acid+0.325*3-hydroxybutylcarnitine+-0.916*2-hydroxydipropionic acid+0.154*estrone sulfate

[0323] In the discovery set, the AUC of the model constructed with these 7 metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.827; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.748.

[0324] (10) Model of 6 metabolite combinations

[0325] Based on the 7 metabolites, one metabolite was randomly eliminated, and there were 7 elimination methods in total. The remaining 6 metabolites could be used to construct a risk model for predicting the progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the 7 elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following 6 metabolite combinations:

[0326] Urea propionic acid, γ-CEHC, thyroxine, guanine acid, 3-hydroxybutylcarnitine, 2-hydroxydipropionic acid

[0327] The model formula is:

[0328] P=-1.935+0.919*ureidopropionic acid-0.537*γ-CEHC-1.018*thyroxine+0.464*colic acid+0.296*3-hydroxybutylcarnitine-0.867*2-hydroxydipropionic acid

[0329] In the discovery set, the AUC of the model constructed with these six metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.823; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.746.

[0330] (11) Model of five metabolite combinations

[0331] Based on the six metabolites, one metabolite was randomly eliminated, and there were six elimination methods in total. The remaining five metabolites could be used to construct a risk model for predicting progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the six elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following five metabolite combinations:

[0332] Urea propionic acid, γ-CEHC, thyroxine, guanylic acid, 3-hydroxybutylcarnitine

[0333] The model formula is:

[0334] P = -2.077 + 0.745 * ureidopropionic acid - 0.539 * γ-CEHC - 0.806 * thyroxine + 0.191 * colic acid + 0.088 * 3-hydroxybutylcarnitine

[0335] In the discovery set, the AUC of the model constructed with these five metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.803; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.743.

[0336] (12) Model of four metabolite combinations

[0337] Based on the five metabolites, one metabolite was randomly eliminated, and there were five elimination methods in total. The remaining four metabolites could be used to construct a risk model for predicting progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the five elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the following four metabolite combinations:

[0338] Urea propionic acid, γ-CEHC, thyroxine, guanidine acid

[0339] The model formula is:

[0340] P=-1.878+0.77*ureidopropionic acid-0.54*γ-CEHC-0.827*thyroxine+0.188*colic acid

[0341] In the discovery set, the AUC of the model constructed with these four metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.801; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.740.

[0342] (13) Model of three metabolite combinations

[0343] Based on the four metabolites, one metabolite was randomly eliminated. There were four elimination methods in total. The remaining three metabolites could be used to construct a risk model for predicting progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the four elimination methods were compared, and the two prediction models with the best combination performance and the highest AUC were selected to obtain two combinations of three metabolites as shown below:

[0344] ①Uredopropionic acid, γ-CEHC, thyroxine

[0345] The model formula is:

[0346] P=-0.983+0.845*ureidopropionic acid-0.565*γ-CEHC-0.835*thyroxine

[0347] In the discovery set, the AUC of the model constructed with these three metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.800; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.734.

[0348] ②Urea propionic acid, γ-CEHC, tubulocholine

[0349] The model formula is:

[0350] P = -4.744 + 1.015 * ureidopropionic acid - 0.584 * γ-CEHC + 0.122 * tubulocolic acid;

[0351] In the discovery set, the AUC of the model constructed with these three metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.789; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.742.

[0352] (14) Model of two metabolite combinations

[0353] Based on the three metabolites, one metabolite was randomly eliminated. There were three elimination methods in total. The remaining two metabolites could be used to construct a risk model for predicting progression of non-ACLF to ACLF within 28 days. The prediction models constructed by the three elimination methods were compared, and the prediction model with the best combination performance and the highest AUC was selected to obtain the combination of two metabolites as shown below:

[0354] Urea propionic acid, γ-CEHC

[0355] The model formula is:

[0356] P=-4.803+0.930*ureidopropionic acid-0.563*γ-CEHC

[0357] In the discovery set, the AUC of the model constructed with these two metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.786; in the validation set, the AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days was 0.745.

[0358] However, compared with one metabolite, the model constructed with two metabolites for predicting the risk of progression from non-ACLF to ACLF within 28 days can indeed significantly improve the prediction performance, and it is the best combination of two among the 15 metabolites. The AUC for predicting the risk of progression from non-ACLF to ACLF within 28 days in both the discovery set and the validation set reached above 0.7.

[0359] In summary, as the number of metabolite types decreases, the AUC values ​​in the discovery set and validation set will decrease slightly when the constructed prediction model predicts the risk of progression from non-ACLF to ACLF within 28 days. However, too many types of metabolites will add a burden to the detection process. After weighing the pros and cons, the model constructed with 3 or 2 metabolites is preferred. At this time, only ureidopropionic acid, γ-CEHC and thyroxine, or ureidopropionic acid, γ-CEHC and colic acid need to be detected, or only ureidopropionic acid and γ-CEHC need to be detected.

[0360] Model ① detected three indicators, ureidopropionic acid, γ-CEHC and thyroxine, and maintained a high AUC value in the discovery set and validation set. In the discovery set, the AUC for predicting the risk of progression of non-ACLF to ACLF within 28 days was 0.800 ( Fig.11 ), which was accurately reproduced in the validation set, with an AUC of 0.734 ( Fig.12 ). The study showed that for the model constructed with the three metabolites①, the prediction value P was greater than 0.150, which predicted a high risk of progression from non-ACLF to ACLF within 28 days; p was less than 0.150, which predicted a low risk of progression from non-ACLF to ACLF within 28 days.

[0361] For the model constructed with two metabolites (ureidopropionic acid and γ-CEHC), the prediction value P was greater than 0.123, which predicted a high risk of progression from non-ACLF to ACLF within 28 days; p was less than 0.123, which predicted a low risk of progression from non-ACLF to ACLF within 28 days.

[0362] At the same time, this example uses two metabolite models (P=-4.803+0.930*ureidopropionic acid-0.563*γ-CEHC) and three metabolite models (P=-0.983+0.845*ureidopropionic acid-0.565*γ-CEHC-0.835*thyroxine, or P=-4.744+1.015*ureidopropionic acid-0.584*γ-CEHC+0.122*colic acid) obtained by screening to predict the risk of progression from non-ACLF to ACLF within 180 days, and also obtains good results, with AUC values ​​reaching above 0.6.

[0363] Example 7: Optimization of prediction model

[0364] (1) Optimization of a four-metabolite combination model for predicting mortality within 90 days of ACLF

[0365] Next, in the discovery set and validation set, for the four metabolite combinations in Example 5, we used 1. combined model (TB (total bilirubin), INR (international normalized ratio), N / L ratio (neutrophil / lymphocyte ratio) and age (age) added to the 4-metabolite group, the model formula was: P = 9.143 + 2.196*age + 1.944*INR-0.432*creatine-6.076*Na+0.052*ALB+0.441*N / L ratio+0.218*overt HE+0.687*pipecolic.acid+1.166*NAAG+0.143*ureidopropionic.acid-0.119*formiminoglutamic.acid); 2. CLIF-C ACLF; 3. MELD-Na; 4. The four metabolite models were re-evaluated based on the quantitative data generated by the targeted assay. Prognostic performance of the 4-metabolite model for mortality within 90 days, as shown in the following table Fig.13 and 14 shown.

[0366] The AUCs of the 4-metabolite model (colic acid, NAAG, ureidopropionic acid, and aminoglutamic acid) were 0.77 and 0.75, respectively. Fig.13 , Fig.14 ). This result was higher than the AUC based on untargeted metabolomics data, which may be due to the improved data quality of targeted analysis. However, these results were not significantly (p < 0.05) better than CLIF-C ACLF and MELD-Na scores. We then explored whether the metabolite model could be improved by incorporating clinical parameters. We added TB, INR, neutrophil / lymphocyte ratio, and age to the 4-metabolite group and used logistic regression analysis to find that the AUC of the 4-metabolite combined with TB, INR, neutrophil / lymphocyte ratio, and age increased to 0.81, which was significantly better than CLIF-C ACLF (AUC 0.76), MELD Na (AUC 0.43), and the 4-metabolite model alone ( Fig.13 The performance of the model is well verified in the validation set ( Fig.14 ).

[0367] (2) Optimization of a three-metabolite combination model for predicting the risk of progression from non-ACLF to ACLF within 28 days

[0368] For the diagnosis of pre-ACLF, in the discovery set and validation set, we used the three metabolite combinations (model ①) in Example 6. We used 1. combined model (TB (total bilirubin), INR (international normalized ratio), Cirrhosis (liver cirrhosis) added to the 3-metabolite group, the model formula is: P = -6.916 + 0.173 * TB + 1.619 * INR + 0.375 * Cirrhosis + 0.735 * ureidopropionic acid - 0.101 * CEHC + 0.099 * thyroxine); 2. CLIF-C ACLF; 3. MELD-Na; 4. 3 metabolite combination model. Based on the quantitative data generated by targeted determination, we re-evaluated the prognostic performance of the 3-metabolite model for the risk of progression from non-ACLF to ACLF within 28 days. The results are as follows Fig.15 and 16 shown.

[0369] The AUCs of the 3-metabolite model (ureidopropionic acid, γ-CEHC, and thyroxine) based on the target assay data were 0.78 and 0.75 in the discovery and validation sets, respectively. These results were similar to the CLIF-C ACLF progression score and MELD Na score. However, by incorporating TB, INR, and the presence of cirrhosis into the metabolic panel, the AUC of the discovery group increased to 0.85, which was significantly higher than that of CLIF-C ACLF (AUC 0.73), MELD-Na (AUC 0.8), and the 3-metabolite panel alone ( Fig.15 The performance of the model is well verified in the validation set ( Fig.16 ).

[0370] Example 8: Combining the newly discovered metabolic biomarkers with other clinical indicators to construct a model

[0371] In this example, appropriate indicators are selected from clinical indicators such as age, INR (international normalized ratio), creatine, sodium, albumin, N / L ratio (granulocyte / lymphocyte ratio), overt HE (overt hepatic encephalopathy), TB (total bilirubin), Cirrhosis (liver cirrhosis), etc., and models are constructed respectively with the 15 metabolites screened in Example 1. Among them, the combination of age, INR, CR, sodium, ALB, N / L ratio, HE.overt, or the combination of age, TB, INR, and N / L ratio is selected to construct a model for predicting the short-term mortality of ACLF, and the combination of TB, INR, and Cirrhosis is used to construct a model for predicting the risk of non-ACLF progressing to ACLF in the short term.

[0372] 1. Model for predicting 90-day mortality in ACLF

[0373] 1) The model for predicting the mortality within 90 days of ACLF constructed by combining age, INR, CR, sodium, ALB, N / L ratio, HE.overt and any one of the 15 metabolites is shown in Table 7.

[0374] Table 7. Model constructed by the combination of age, INR, CR, Sodium, ALB, N / L ratio, HE.overt and one metabolite

[0375]

[0376]

[0377] According to Table 7, it can be seen that the model obtained by combining any one of the 15 newly discovered metabolic biomarkers of the present invention with age, INR, CR, Sodium, ALB, N / L ratio, and HE.overt can effectively improve the AUC values ​​in the discovery group and the validation group compared with the model without metabolites. It can be seen that the 15 metabolites provided by the present invention are effective in predicting the short-term mortality of ACLF; in addition, it can be understood that on the basis of the combined model provided in Table 7, further increasing the number of metabolites can further improve the effect of predicting the short-term mortality of ACLF, and the AUC value will be further improved.

[0378] 2) The model for predicting the mortality within 90 days of ACLF constructed by combining age, TB, INR, N / L ratio with any one of the 15 metabolites is shown in Table 8.

[0379] Table 8. Model constructed by the combination of age, TB, INR, N / L ratio and one metabolite

[0380]

[0381] According to Table 8, it can be seen that the model obtained by combining any one of the 15 newly discovered metabolic biomarkers of the present invention with age, TB, INR, and N / L ratio can effectively improve the AUC values ​​in the discovery group and the validation group compared with the model without metabolites. It can be seen that the 15 metabolites provided by the present invention are effective in predicting the short-term mortality of ACLF; in addition, it can be understood that on the basis of the combined model provided in Table 8, further increasing the number of metabolites can further improve the effect of predicting the short-term mortality of ACLF, and the AUC value will be further improved.

[0382] 3) The model constructed by combining TB, INR, Cirrhosis and any one of the 15 metabolites to predict the risk of progression from non-ACLF to ACLF within 28 days is shown in Table 9.

[0383] Table 9. Model constructed by combining TB, INR, Cirrhosis and one metabolite

[0384]

[0385]

[0386] According to Table 9, it can be seen that the model obtained by combining any one of the 15 newly discovered metabolic biomarkers of the present invention with TB, INR, and Cirrhosis can effectively improve the AUC values ​​in the discovery group and the validation group compared with the model without metabolites. It can be seen that the 15 metabolites provided by the present invention are effective in predicting the risk of non-ACLF progressing to ACLF in a short period of time; in addition, it can be understood that on the basis of the combined model provided in Table 9, further increasing the number of metabolites can further improve the effect of the risk of non-ACLF progressing to ACLF in a short period of time, and the AUC value will be further improved.

[0387] All patents and publications mentioned in the present specification indicate that these are public technologies in the field and can be used in the present invention. All patents and publications cited here are also listed in the references, just as each publication is specifically cited separately. The present invention described here can be implemented in the absence of any one or more elements, one or more restrictions, and such restrictions are not specifically stated here. For example, in each example here, the terms "comprising", "substantially consisting of..." and "consisting of..." can be replaced by the other two terms of one of the two. The so-called "one" here only means "one", but does not exclude that it only includes one, and can also mean that it includes more than two. The terms and expressions used here are descriptive and not limited by them. There is no intention to indicate that these terms and explanations described in this book exclude any equivalent features, but it can be known that any appropriate changes or modifications can be made within the scope of the present invention and the claims. It can be understood that the embodiments described in the present invention are some preferred embodiments and features, and any person skilled in the art can make some changes and variations based on the essence of the present invention, and these changes and variations are also considered to belong to the scope of the present invention and the scope limited by the independent claims and the appended claims.

Claims

1. Use of a biomarker for preparing a reagent for predicting whether an ACLF patient has a risk of death or a probability of survival, characterized in that: The biomarkers include one or more of the following: 2-hydroxydipropionic acid, hexadecanedioic acid, hydroxyproline, methyl sulfide, 3-hydroxybutylcarnitine, colic acid, estrone sulfate, N-acetylamino-glutamic acid, quinolinic acid, ureidopropionic acid, γ-CEHC, thyroxine, methylaminoglutamic acid, γ-3 glutamyl dipeptide, and bilirubin.

2. The use according to claim 1, characterized in that The biomarkers include urocandin, NAAG and ureidopropionic acid; or, The biomarkers include colicolic acid, NAAG, ureidopropionic acid, and methylaminoglutamic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, and γ-CEHC; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, and γ-3 glutamyl dipeptide; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, and 3-hydroxybutylcarnitine; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, and 3-hydroxybutylcarnitine. The biomarkers include glutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine and estrone sulfate; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate and hydroxyproline; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, and hydroxyproline; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, and hydroxyproline; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, Acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, and 2-hydroxydipropionic acid; Alternatively, the biomarker includes colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, and methyl sulfide; Alternatively, the biomarker includes colic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, and methyl sulfide. or the biomarkers include 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid and quinolinic acid; or the biomarkers include 2-hydroxydipropionic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid and quinolinic acid; or the biomarkers include 2-hydroxydipropionic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, quinolinic acid and thyroxine.

3. The use according to claim 1, characterized in that The biomarkers include one or more of the following: colic acid, NAAG, ureidopropionic acid.

4. The use according to claim 1, characterized in that The reagent is used to detect biomarkers in blood, serum, plasma or whole blood samples of ACLF patients; the reagent is used to detect the presence or relative abundance or concentration of biomarkers in the sample.

5. The use according to claim 1, characterized in that The biomarkers also include one or more of age, INR, creatine, sodium, albumin, neutrophil / lymphocyte count ratio, overt hepatic encephalopathy, and TB.

6. The use according to claim 1, characterized in that The described uses include predicting whether an ACLF patient has a risk of death or a probability of survival within 360 days.

7. The use according to claim 1, characterized in that The described uses include predicting whether an ACLF patient has a risk of death or a probability of survival within 90 days.

8. A kit for predicting whether an ACLF patient has a risk of death or a probability of survival, characterized in that: The kit comprises a detection reagent for a biomarker for use according to any one of claims 1 to 7.

9. The kit according to claim 8, characterized in that The detection reagent of the biomarker includes a standard substance of the biomarker.

10. A biomarker combination for predicting whether an ACLF patient has a risk of death or a probability of survival, characterized in that: The following biomarkers are included: colicolic acid, NAAG, and ureidopropionic acid; or colicolic acid, NAAG, ureidopropionic acid, and methylaminoglutamic acid; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, and γ-CEHC; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, and γ-3 glutamyl dipeptide; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, and 3-hydroxybutylcarnitine; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, and 3-hydroxybutylcarnitine Glutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine and estrone sulfate; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate and hydroxyproline; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline and bilirubin; or colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline and bilirubin , γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, and 2-hydroxydipropionic acid; or colic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, and methyl sulfide; or colic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid , methyl sulfide and hexadecanedioic acid; or colic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid and quinolinic acid; or colic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, quinolinic acid and thyroxine.

11. The biomarker combination according to claim 10, characterized in that The following biomarkers were included: urogenital acid, NAAG, and ureidopropionic acid.

12. The biomarker combination according to claim 11, characterized in that It also includes one or more of age, INR, creatine, sodium, albumin, neutrophil / lymphocyte count ratio, overt hepaticencephalopathy, and TB.

13. A system for predicting whether an ACLF patient has a risk of death or a probability of survival, characterized in that: The system includes a data analysis module; the data analysis module is used to analyze the detection value of the biomarker, and the biomarker includes one or more of the following: 2-hydroxydipropionic acid, hexadecanedioic acid, hydroxyproline, methyl sulfide, 3-hydroxybutylcarnitine, colic acid, estrone sulfate, N-acetylamino-glutamic acid, quinolinic acid, ureidopropionic acid, γ-CEHC, thyroxine, methylaminoglutamic acid, γ-3 glutamyl dipeptide, and bilirubin.

14. The system of claim 13, wherein: The detection value of the biomarker is to detect the biomarker in the blood, serum, plasma or whole blood sample of ACLF patients, and is used to detect the presence or relative abundance or concentration of the biomarker in the sample.

15. The system of claim 14, wherein: The biomarkers include colicolic acid, NAAG, and ureidopropionic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, and methylaminoglutamic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, and γ-CEHC; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, and γ-3 glutamyl dipeptide; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, and 3-hydroxybutylcarnitine; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamic acid, γ-CEHC, γ-3 glutamyl dipeptide, and 3-hydroxybutylcarnitine, and estrone sulfate; or, the biomarkers include colicolic acid, NA AG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate and hydroxyproline; or, the biomarker includes colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline and bilirubin; or, the biomarker includes colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin and 2-hydroxydipropionic acid; or, the biomarker includes colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid and methyl sulfide; or, The biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, and hexadecanedioic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid and quinolinic acid; or, the biomarkers include colicolic acid, NAAG, ureidopropionic acid, methylaminoglutamate, γ-CEHC, γ-3 glutamyl dipeptide, 3-hydroxybutylcarnitine, estrone sulfate, hydroxyproline, bilirubin, 2-hydroxydipropionic acid, methyl sulfide, hexadecanedioic acid, quinolinic acid and thyroxine.

16. The system of claim 15, wherein: The data analysis module substitutes the detection value of the biomarker into the regression equation to calculate the predicted value for predicting whether the ACLF patient has the risk of death, thereby evaluating the risk of death of the ACLF patient.

17. The system of claim 16, wherein: The regression equation is a logistic regression equation.

18. The system of claim 17, wherein: The logistic regression equation is constructed and trained by the detection values ​​of biomarkers of known samples.

19. The system of claim 18, wherein: The training is performed by constructing a training set and using a training method to train the training set so as to obtain a logistic regression equation; the training set is serum samples of ACLF patients.

20. The system of claim 19, wherein: The training method is to divide the training set into a survivor group and a non-survivor group according to whether the patients survived in the short term, detect the biomarker contents of the survivor group and the non-survivor group respectively, obtain the detection values, perform logistic regression based on the detection values ​​of the biomarkers, analyze the relationship between the detection values ​​of the survivor group and the non-survivor group, and construct a logistic regression equation.

21. The system of claim 20, wherein: The short term includes any one or more of within 90 days, 180 days, 360 days, 80 days, 70 days, 60 days, three months, six months, one year, two months, and one month.

22. The system of claim 21, wherein: The data analysis module substitutes the detection value of the biomarker into the logistic regression equation to calculate the predicted value of the ACLF patient's risk of death within 360 days, thereby evaluating the ACLF patient's risk of death within 360 days.

23. The system of claim 22, wherein: The logistic regression equation is: P = -20.584 + 0.846 * colic acid + 1.254 * NAAG + 0.454 * ureidopropionic acid - 0.468 * aminoglutamic acid - 0.385 * γ-CEHC + 0.485 * γ-3 glutamyl dipeptide + 0.447 * 3-hydroxybutylcarnitine + 0.081 * estrone sulfate + 0.524 * hydroxyproline + 0.079 * bilirubin - 0.066 * 2-hydroxydipropionic acid - 0.015 * methyl sulfide - 0.089 * hexadecanedioic acid - 0.219 * quinolinic acid + 0.942 * thyroxine; or P = -14.286 + 0.804 * tubulocolic acid + 0.791 * NAAG + 0.448 * ureidopropionic acid - 0.396 * methylaminoglutamic acid - 0.338 * γ-CEHC + 0.434 * γ-3 glutamyl dipeptide + 0.425 * 3-hydroxybutylcarnitine + 0.107 * estrone sulfate + 0.356 * hydroxyproline + 0.091 * bilirubin - 0.082 * 2-hydroxydipropionic acid + 0.015 * methyl sulfide - 0.066 * hexadecanedioic acid - 0.154 * quinolinic acid; or P = -14.392 + 0.799 * tubulocolic acid + 0.792 * NAAG + 0.383 * ureidopropionic acid - 0.403 * methylaminoglutamic acid - 0.352 * γ-CEHC + 0.428 * γ-3 glutamyl dipeptide + 0.388 * 3-hydroxybutylcarnitine + 0.121 * estrone sulfate + 0.353 * hydroxyproline + 0.095 * bilirubin - 0.153 * 2-hydroxydipropionic acid + 0.009 * methyl sulfide - 0.025 * hexadecanedioic acid; or P = -14.5 + 0.794 * tubulocolic acid + 0.806 * NAAG + 0.376 * ureidopropionic acid - 0.403 * methylaminoglutamic acid - 0.352 * γ-CEHC + 0.429 * γ-3 glutamyl dipeptide + 0.378 * 3-hydroxybutylcarnitine + 0.114 * estrone sulfate + 0.362 * hydroxyproline + 0.095 * bilirubin - 0.16 * 2-hydroxydipropionic acid + 0.007 * methyl sulfide; or P = -14.476 + 0.795 * colic acid + 0.807 * NAAG + 0.376 * ureidopropionic acid - 0.404 * aminoglutamic acid - 0.353 * γ-CEHC + 0.43 * γ-3 glutamyl dipeptide + 0.375 * '3-hydroxybutyrate carnitine' + 0.115 * estrone sulfate + 0.364 * hydroxyproline + 0.096 * bilirubin - 0.157 * 2-hydroxydipropionic acid; or P = -14.315 + 0.777*colic acid + 0.77*NAAG + 0.342*ureidopropionic acid - 0.411*methylaminoglutamic acid - 0.352*γ-CEHC + 0.432*γ-3 glutamyl dipeptide + 0.325*3-hydroxybutyrate carnitine + 0.109*estrone sulfate + 0.357*hydroxyproline + 0.1*bilirubin; or P = -13.898 + 0.825 * colic acid + 0.744 * NAAG + 0.322 * ureidopropionic acid + -0.391 * aminoglutamic acid + -0.269 * γ-CEHC + 0.424 * γ-3 glutamyl dipeptide + 0.345 * 3-hydroxybutyric acid carnitine + 0.162 * estrone sulfate + 0.276 * hydroxyproline; or P = -12.473 + 0.857 * colic acid + 0.781 * NAAG + 0.415 * ureidopropionic acid -0.405 * aminoglutamic acid -0.257 * γ-CEHC + 0.42 * γ-3 glutamyl dipeptide + 0.343 * 3-hydroxybutyric acid carnitine + 0.14 * estrone sulfate; or P = -13.147 + 0.947 * tubulocolic acid + 0.832 * NAAG + 0.435 * ureidopropionic acid - 0.407 * methylaminoglutamic acid - 0.292 * γ-CEHC + 0.428 * γ-3 glutamyl dipeptide + 0.334 * 3-hydroxybutylcarnitine; or P = -12.838 + 0.839 * tubulocolic acid + 1.125 * NAAG + 0.49 * ureidopropionic acid - 0.39 * methylaminoglutamic acid - 0.263 * γ-CEHC + 0.39 * γ-3 glutamyl dipeptide; or P = -10.874 + 0.685 * tubulocolic acid + 1.105 * NAAG + 0.51 * ureidopropionic acid - 0.214 * aminoglutamic acid - 0.22 * γ-CEHC; Or p = -11.911 + 0.81 * tubulocolic acid + 0.992 * NAAG + 0.559 * ureidopropionic acid - 0.2 * methylaminoglutamic acid; or p = -11.25 + 0.73*colic acid + 0.917*NAAG + 0.435*ureidopropionic acid; Where p represents the predicted value of the mortality risk in ACLF patients.

24. The system of claim 23, wherein: The logistic regression equation also includes one or more clinical indicators of age, INR, creatine, sodium, albumin, neutrophil / lymphocyte count ratio, overt hepaticencephalopathy, and TB.

25. The system of claim 24, wherein: The logistic regression equation is: p=-11.25+0.73*colic acid+0.917*NAAG+0.435*urea propionic acid; p represents the predicted value of predicting the risk of death within 90 days in ACLF patients.

26. The system of claim 25, wherein: When P was greater than 0.320, it was predicted that ACLF patients had a high risk of death within 90 days; When p was less than 0.320, it was predicted that ACLF patients had a low risk of death within 90 days.

27. The system of claim 26, wherein: The system also includes a data storage module, a data input interface and a data output interface; the data storage module is used to store the detection values ​​of the biomarkers; the data input interface is used to input the detection values ​​of the biomarkers, and the data output interface is used to output the prediction results.

Images