Combined biomarker for drug-induced liver injury diagnosis, diagnosis model and application

Abstract

The invention discloses a combined biomarker for drug-induced liver injury diagnosis, a diagnosis model and application, plasma metabolites with abundance difference are obtained on the basis of plasma metabolite data related to intestinal flora, and the metabolites show excellent differential diagnosis value for distinguishing drug-induced liver injury from healthy people and have good application prospects. A high-specificity and high-accuracy diagnosis model is constructed by combining the plasma metabolite and baseline clinical information, the model can effectively and accurately diagnose the drug-induced liver injury and healthy people, and an auxiliary decision-making tool with conversion application value is provided for objective evaluation of the drug-induced liver injury in clinical practice.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of drug-induced liver injury diagnosis, in particular to a combined biomarker for drug-induced liver injury diagnosis, a diagnostic model and application. BACKGROUND

[0002] Drug-induced liver injury (DILI) refers to liver damage caused directly or indirectly by exogenous substances such as various chemical drugs, biological agents and traditional Chinese medicines, and their metabolites. It is one of the most common types of serious adverse drug reactions in clinical practice, and is also an important reason for the failure of new drug research and development and the withdrawal of marketed drugs. The clinical manifestations of DILI are complex, which can progress from asymptomatic liver enzyme elevation to acute liver failure. Its diagnosis and management remain a major challenge in the field of liver disease. In recent years, the rise of the "gut-liver axis" theory has provided a new perspective for understanding the pathogenesis of DILI. The structural and functional disorders of the intestinal flora can significantly regulate the susceptibility of individuals to drug hepatotoxicity through mechanisms such as regulating host immunity, changing bile acid metabolism profile, and affecting drug bioconversion processes. Metabolites derived from the flora, such as bile acids, short-chain fatty acids, and tryptophan metabolites, enter the portal circulation and directly participate in liver inflammation, oxidative stress and cell death, etc., forming a key molecular bridge connecting intestinal microecology and liver injury. Therefore, systematic analysis of intestinal flora-related metabolites in plasma provides a valuable opportunity for discovering new diagnostic markers for DILI. However, existing researches have focused on single type of omics data or isolated risk factors, failing to effectively integrate host clinical characteristics with deep molecular metabolic information. Therefore, it is urgent to build a multi-dimensional identification model that can synergistically integrate baseline clinical information and plasma metabolomics characteristics, to break through the limitations of existing diagnostic models, achieve early, non-invasive and precise identification and risk stratification of DILI, and provide scientific basis for subsequent early intervention and individualized treatment strategies.

[0003] I. Epidemiology of DILI

[0004] Globally, the incidence of drug-induced liver injury (DILI) is on the rise. Common hepatotoxic drugs include anti-tuberculosis drugs, traditional Chinese medicines / herbal medicines, and analgesic and antipyretic drugs, among which traditional Chinese medicine-related DILI accounts for more than 25%, which has become a focus and difficulty in the field of DILI research in China.

[0005] II. Technical bottlenecks of existing diagnostic tools in the diagnosis of DILI

[0006] Current diagnosis of DILI mainly relies on comprehensive clinical evaluation, including: ① clinical manifestations, such as jaundice, fatigue, and elevated transaminases; ② laboratory tests, such as changes in serum ALT, AST, ALP, and TBIL; ③ etiological exclusion, excluding other causes of liver damage such as viral hepatitis, autoimmune liver disease, and alcohol-induced liver disease; ④ causality assessment tools, the most commonly used is RUCAM (Roussel Uclaf Causality Assessment Method) scoring system to assess the correlation between drugs and liver damage; ⑤ imaging and liver biopsy, used to exclude structural liver disease or obtain histological evidence when necessary. However, these diagnostic methods have significant limitations. First, the clinical manifestations of DILI lack specificity and are easily confused with other liver diseases. Second, routine biochemical indicators can only reflect the extent of liver function damage, but cannot determine the specific cause or mechanism of liver damage. Although RUCAM is an internationally recognized tool, it is highly subjective and information-dependent, and its application is easily influenced by clinical experience, medical record completeness, and other factors, leading to inconsistent causality determination. In addition, imaging methods have low sensitivity for early functional liver damage, and liver biopsy can provide pathological evidence, but it is invasive and has no specific features in most DILI. In particular, in regions where Chinese herbal medicine is widely used, such as China and Southeast Asia, the diagnosis of DILI faces greater challenges: the composition of Chinese herbal medicine is complex, the quality standards are not uniform, and patients often cannot provide accurate medication history, further increasing the difficulty of exclusion. Therefore, the current diagnosis of DILI relies on multidisciplinary collaboration and clinical experience, and lacks a rapid, accurate, and standardized tool, which greatly limits the early identification, risk stratification, and individualized intervention of DILI. In recent years, with the rapid development of multi-omics technology, non-invasive diagnostic strategies represented by metabolomics have shown significant potential in DILI research. Studies have shown that DILI patients have significant differences in the expression profiles of various gut flora-related plasma metabolites, such as secondary bile acids, short-chain fatty acids, tryptophan metabolites, and microbial-derived oxidized lipids. These metabolic characteristics are expected to become new biomarkers for early identification and risk stratification of DILI. With the increasing understanding of the role of the 'gut-liver axis' in the pathogenesis of DILI, targeting gut flora and their metabolic microenvironment has become a potential treatment direction.

[0007] III. The value of gut flora-related metabolites in the progression of DILI

[0008] Multiple studies have revealed the core role of gut microbiota and its metabolites in the occurrence and development of DILI from the perspective of the 'gut-liver axis' theory. Gut microbiota significantly affects the susceptibility and course of DILI through various mechanisms, such as metabolizing drug precursors, regulating the enterohepatic circulation of bile acids, and producing bioactive metabolites. Specifically, gut microbiota metabolites can directly or indirectly participate in the occurrence and development of DILI by regulating key pathophysiological processes such as liver inflammatory response, oxidative stress level, mitochondrial function stability, and bilirubin metabolism. During the progression of DILI, the disturbance of gut microbiota structure leads to changes in a series of characteristic metabolite profiles. These metabolites include various conjugated bile acids and their derivatives involved in bile acid metabolism pathways, antioxidants reflecting oxidative stress levels, and gut-host co-metabolites involved in inflammation regulation. Changes in their blood concentrations not only reflect the degree of hepatocyte injury and metabolic dysfunction but also show significant correlation with the severity and prognosis of DILI. For example, the proportion of hydrophobic bile acids (such as deoxycholic acid DCA) significantly increases due to gut microbiota imbalance during DILI, and such bile acids have strong hepatotoxicity, which can directly exacerbate hepatocyte injury and apoptosis through pathways such as oxidative stress, mitochondrial dysfunction, and endoplasmic reticulum stress. In addition, the inflammatory signal molecules lipopolysaccharide (LPS) and trimethylamine (TMA) produced by gut microbiota metabolism of dietary components are oxidized to trimethylamine-N-oxide (TMAO) in the liver, which can activate Kupffer cells in the liver and mediate strong inflammatory responses through Toll-like receptor 4 (TLR4) and other signaling pathways, thereby amplifying the hepatotoxicity of drugs. Clinical studies have shown that the increase in plasma TMAO and LPS binding protein (LBP) levels is significantly associated with the severity of inflammation and adverse prognosis in DILI patients. Furthermore, the reduction of protective metabolites from gut microbiota is also an important mechanism for the deterioration of DILI. Short-chain fatty acids (SCFAs) (such as butyric acid, propionic acid, and acetic acid) are beneficial metabolites produced by gut microbiota fermentation of dietary fiber, which not only provide energy for the intestinal epithelium, maintain barrier integrity, and prevent bacterial translocation but also exert systemic anti-inflammatory effects by regulating the differentiation and function of regulatory T cells (Tregs).

[0009] Therefore, the disturbance of gut microbiota-related metabolites is not only an important pathological feature of DILI but also a key factor driving its occurrence and progression. By systematically analyzing the specific gut microbiota metabolic phenotypes of DILI, these quantifiable and easily monitored molecular indicators provide important evidence for early diagnosis and risk stratification of DILI. Moreover, intervention strategies targeting such metabolic abnormalities may also become a new direction for DILI treatment research, aiming to improve patient prognosis and effectively delay disease progression. SUMMARY

[0010] The present application provides a combined biomarker for drug-induced liver injury diagnosis, which provides a basis for early diagnosis of drug-induced liver injury, so as to improve the prognosis of patients and effectively delay the progression of the disease.

[0011] Therefore, the present application provides the following solutions: The first aspect of the present application is to provide a biomarker for diagnosing drug-induced liver injury, which is a plasma metabolite, including glycocholic acid, L-carnosine and valine-phenylalanine.

[0012] The second aspect of the present application is to provide the use of a detection reagent of the biomarker in the first aspect in the preparation of a drug-induced liver injury diagnosis product.

[0013] Further, the product is a kit, a detection system or an instrument.

[0014] Further, the detection reagent is used to detect the abundance of the biomarker in the plasma metabolite. And / or, the detection reagent comprises at least one of a mass spectrometry reagent or a chromatography reagent.

[0015] The third aspect of the present application is to provide a drug-induced liver injury diagnosis kit, which comprises a plasma metabolite abundance detection reagent and a prediction formula; the plasma metabolite abundance detection reagent is used to detect the abundance of the biomarker in claim 1; and the prediction formula is constructed based on the biomarker in the first aspect.

[0016] Further, the prediction formula is: p(positive) = 1 / (1 + exp(-(2.8348 -0.7049 Valine-phenylalanine abundance value + 0.3937 Glycocholic acid abundance value - 0.4499·L-carnosine abundance value))) when P(positive) > 0.412, it is determined as a drug-induced liver injury patient.

[0017] Further, the kit further comprises a serum glutamic-pyruvic transaminase content detection reagent, and the prediction formula is: p(positive) = 1 / (1 + exp(-(1.7976-2.4956 Valine-phenylalanine abundance value + 1.1172 Glycocholic acid abundance value - 1.1187 L-carnosine abundance value + 0.1886·serum glutamic-pyruvic transaminase concentration value))) when P(positive) > 0.936, it is determined as a drug-induced liver injury patient.

[0018] Further, the kit further comprises a plasma metabolite extraction reagent.

[0019] A fourth aspect of the present application provides a device for diagnosing drug-induced liver injury, comprising a detection unit and a data analysis unit, wherein: The detection unit is used to obtain a plasma sample, detect and determine the detection results of sample plasma metabolite abundance and serum glutathione transaminase concentration; the plasma metabolite is the biomarker of the first aspect; The data analysis unit is used to analyze the detection results of the detection unit combined with a prediction formula to make a diagnosis result; the prediction formula is: p(positive) = 1 / (1+exp(-(1.7976-2.4956 Valine abundance value + 1.1172 Glycocholic acid abundance value - 1.1187 L-carnosine abundance value + 0.1886 Serum glutathione transaminase concentration value))) when P(positive) > 0.936 is determined as a drug-induced liver injury patient.

[0020] Further, the detection unit comprises at least one of LC-MS, GC-MS, LC-MS / MS detection reagent, system and instrument.

[0021] Compared with the prior art, the present application has the following beneficial effects: The present application is based on plasma metabolite data related to intestinal flora, and obtains plasma metabolites with abundance difference. These metabolites show excellent differential diagnostic value for distinguishing drug-induced liver injury (DILI) from healthy (HC) people, and thus serve as a combination biomarker for diagnosing drug-induced liver, so as to realize accurate and non-invasive identification of DILI and healthy people, and provide a new basis for realizing clinical accurate diagnosis and treatment.

[0022] The present application constructs a prediction model with high specificity and high accuracy by combining plasma metabolites and baseline clinical information serum glutathione transaminase. The model can effectively realize the distinction between drug-induced liver injury patients and healthy people, and provides an auxiliary decision-making tool with conversion application value for early diagnosis and treatment of drug-induced liver injury in clinical practice. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 Figure 2 shows the influence of drug-induced liver injury severity on human plasma metabolites and intestinal flora in the examples. A figure: PLS-DA analysis of plasma metabolites in the mild (Mild) and moderate to severe (Moderate to Severe) groups; B figure: volcano plot of the differential plasma metabolites in the Mild and Moderate to Severe groups.

[0024] Figure 2 Variable importance on the projection (VIP) analysis results of the first principal component of the plasma metabolites for the difference between Mild and Moderate to Severe groups in the Examples.

[0025] Figure 3 Intestinal fungal alpha diversity results for DILI patients and HC in the Examples; A figure: shannon index Wilcoxon rank-sum test bar plot of fecal bacteria for Mild and Moderate to Severe groups; B figure: sobs index Wilcoxon rank-sum test bar plot of fecal bacteria for Mild and Moderate to Severe groups.

[0026] Figure 4 Beta diversity PCA analysis results of fecal bacteria for Mild and Moderate to Severe groups in the Examples.

[0027] Figure 5 Analysis results of the top 10 major difference bacteria in relative abundance at the species level for Mild and Moderate to Severe groups in the Examples.

[0028] Figure 6 Spearman correlation heat map of the top 50 VIP plasma difference metabolites and the top 10 major difference bacteria in relative abundance at the species level for Mild and Moderate to Severe groups in the Examples, to screen endogenous plasma metabolites with strong correlation with difference flora.

[0029] Figure 7 AUC curves of the top 3 individual metabolites in the training set in the Examples, and AUC curves of the logistic regression model established by joint metabolites in the samples.

[0030] Figure 8 AUC curves of the top 3 individual metabolites in the validation set in the Examples, and AUC curves of the logistic regression model established by joint metabolites in the samples.

[0031] Figure 9 AUC curves of part of the individual metabolites in the training set in the Examples, AUC curves of patient TB, and AUC curves of the logistic regression model established by joint metabolites and patient TB in the samples.

[0032] Figure 10 AUC curves of part of the individual metabolites in the validation set in the Examples, AUC curves of patient TB, and AUC curves of the logistic regression model established by joint metabolites and patient TB in the samples. DETAILED DESCRIPTION

[0033] The technical solutions of the present application will be described clearly and completely below in combination with preferred embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the protection scope of the present application.

[0034] In one embodiment, based on the plasma metabolite data related to intestinal flora of drug-induced liver injury (DILI) patients and healthy people, plasma metabolites with abundance difference are obtained. These metabolites show excellent diagnostic value in distinguishing drug-induced liver injury from healthy people through verification, and can be used as biomarkers to prepare a diagnostic kit for DILI. These biomarkers are a combination of N-Choloylglycine (common name: glycine cholate; CAS registry number: 475-31-0), Carnosine (common name: L-carnosine; CAS registry number: 305-84-0), and Val Phe (full name: Valylphenylalanine; common name: valine phenylalanine; CAS registry number: 3918-92-1).

[0035] In a preferred embodiment, the abundance value of the above biomarkers in plasma metabolites is detected, and an identification model is combined to facilitate the direct distinction between drug-induced liver injury (DILI) patients and healthy people (HC).

[0036] In the above embodiment, the detection of the abundance value of plasma metabolites is performed by detecting the proportion of the target metabolite in all metabolites, and the content determination of the target metabolite includes but is not limited to known gas chromatography (GC) and liquid chromatography (LC), combined with mass spectrometry (MS) technology to form GC-MS, LC-MS, and liquid chromatography-tandem mass spectrometry (LC-MS / MS) and other high-efficiency separation and detection methods, thus including necessary detection reagents and detection systems or instruments.

[0037] In a preferred embodiment, the identification model is p(positive) = 1 / (1 + exp(-(2.8348 + -0.7049 Val Phe + 0.3937 N-Choloylglycine + -0.4499 Carnosine), the DILI positive probability p(positive) was calculated by detecting the relative abundance (percentage, %) of the above plasma metabolites. When P(positive) > 0.412, the patient was determined as DILI patient. This model showed excellent diagnostic performance: AUC = 1.000, sensitivity 100%, specificity 100%, positive predictive value 100%, and negative predictive value 100%.

[0038] In a more preferred embodiment, the identification model is p(positive) = 1 / (1 + exp(-(1.7976 +-2.4956 Val Phe + 1.1172 N-Choloylglycine + -1.1187 Carnosine + 0.1886 ALT). The DILI positive probability p(positive) was calculated by detecting the relative abundance (percentage, %) of the above plasma metabolites and the serum glutamic-pyruvic transaminase concentration (U / L). This model showed perfect diagnostic performance (AUC = 1.000) in the samples, with a sensitivity of 100% and a specificity of 100%. That is, when the intestinal flora-related metabolites and patient ALT were combined, P(DILI) was greater than 0.936, the patient was determined as DILI patient.

[0039] Embodiment

[0040] I. Plasma metabolites and intestinal bacteria sequencing results of healthy controls and DILI patients

[0041] 1. Purpose of the experiment: By collecting basic information, blood samples and fecal samples of healthy controls and DILI patients, the differences in plasma metabolites and fecal bacteria between DILI patients and healthy controls were analyzed by sequencing to determine whether the plasma metabolites and intestinal bacteria of DILI patients have changed.

[0042] 2. Experimental methods: The basic information and clinical indicators such as hematology of hospitalized patients (including healthy controls, DILI) were collected prospectively. The inclusion criteria and exclusion criteria of DILI patients were as follows: (1) Patients who met all the following inclusion criteria were included: ① Age > 18 years old ② Meet the diagnostic criteria of DILI in the "Guidelines for the diagnosis and treatment of DILI" ③ Have a history of drug use ④ Clinical data is relatively complete, and the compliance is good. (2) Patients were excluded according to the following exclusion criteria: ① Hepatocellular carcinoma or liver metastasis ② Combined with infectious liver diseases, such as hepatitis A virus, hepatitis B virus, hepatitis C virus, hepatitis D virus, hepatitis E virus, human immunodeficiency virus ③ Combined with non-infectious liver diseases, such as non-alcoholic fatty liver disease, alcoholic liver disease, autoimmune liver disease, immunoglobulin G4-related liver disease, Wilson's disease, alpha-1-antitrypsin deficiency, Biliary syndrome and other congenital liver diseases ④ Combined with serious organic lesions ⑤ Pregnant and lactating women ⑥ Used antibiotics or probiotics affecting intestinal flora in the past 1 month. The healthy control (HC) group met the following criteria (1) Age > 18 years old (2) No history of liver disease or other organic disease. Blood and fecal samples were collected from subjects at admission, plasma samples were placed in blood collection tubes, fecal samples were placed in sterile plastic cups, and were stored at -80°C until extraction.

[0043] Metabolomics analysis: Liquid chromatography-MS / MS technique (UPLC-MS / MS) analysis, 100 μL of plasma was added to a 1.5 mL centrifuge tube, 800 μL of solution (acetonitrile:methanol = 1:1 (v:v)) containing four internal standards (0.02 mg / mL L-2-chlorophenylalanine, etc.) was added to extract metabolites. After mixing the sample in a vortex shaker for 30 seconds, it was ultrasonically treated at low temperature for 30 minutes (5°C, 40 KHz), followed by being placed at -20°C for 30 minutes to precipitate proteins. Then, the sample was centrifuged for 15 minutes (4°C, 13000 g). The supernatant was removed and blown dry under nitrogen. The sample was then redissolved with 100 μL of solution (acetonitrile:water = 1:1) and extracted by low-temperature ultrasonic treatment for 5 minutes (5°C, 40 KHz), followed by centrifugation at 13000 g and 4°C for 10 minutes. The supernatant was transferred to a sample bottle for LC-MS / MS analysis. A mixed quality control sample (QC) was prepared by mixing all samples in equal volumes. According to the operating instructions, data were collected using a Thermo UHPLC-Q Exactive HF-X mass spectrometer. The raw data were peak detected using Progenesis QI 2.3 (Water Corporation, Milford, USA) software. The internal standard peaks were removed from the data matrix. At least 80% of the metabolites detected in any one group of samples were retained. The maximum mass error allowed was ±10 ppm, and metabolites with MS / MS fragment scores higher than 30 were identified.

[0044] 3. Experimental results

[0045] 3.1 Drug-induced liver injury affects the distribution of plasma metabolites

[0046] Figures 1-2 Effects of drug-induced liver injury on the distribution of plasma metabolites in humans. In this study, plasma samples were analyzed for metabolites by liquid chromatography-tandem mass spectrometry (LC-MS / MS) technique, and data matrices were obtained by metabolomics processing software (Progenesis QI (Waters Corporation, Milford, USA)). MS and MSMS mass spectrum information were matched with public databases HMDB and Metlin and the Meiji self-built library to obtain metabolite information, and finally the effects of drug-induced liver injury on plasma metabolites were analyzed. The results showed that in PLS-DA analysis, the separation degree of samples of DILI patients and HC was large, and the classification effect was very significant (R2X = 0.99, R2Y = 0.99, Q2 = 0.99) Figure 1A). Volcano plot of differential metabolites with VIP > 1 and FDR < 0.05 by Wilcox'T test. There were 77 up-regulated metabolites and 119 down-regulated metabolites in DILI group Figure 1 B). The distribution of plasma metabolites in the two groups changed significantly. The differential metabolites were used as a supervised model by OPLS-DA / PLS-DA, and 7-fold cross validation was used to test and predict the different changes of the paired samples. The first principal component variable projection importance VIP analysis determined the top 50 differential metabolites that contributed to classification Figure 2 ), which were used for correlation analysis with differential bacteria.

[0047] 3.2 Drug-induced liver injury affects the distribution of intestinal bacteria

[0048] Figures 2-5 The impact of drug-induced liver injury on human intestinal bacterial distribution. This study systematically analyzed the impact of drug-induced liver injury on human intestinal bacterial community by 16S rRNA sequencing technology. The results showed that the intestinal fungal alpha diversity (including Shannon index and sobs index) of DILI patients and HC showed significant differences (P < 0.05) Figure 2 、 Figure 3 Beta diversity analysis showed that the bacterial community structure was significantly separated between groups (P < 0.05) Figure 4 , indicating that drug-induced liver injury significantly changed the composition characteristics of intestinal bacteria. At the species level, Faecalibacterium prausnitzii 、 Gemmiger formicilis 、 Ruminococcoides bili , etc. Beneficial bacteria were significantly reduced in DILI, Veillonella parvula 、 Streptococcus macedonicus , etc. Opportunistic pathogens were significantly increased in DILI Figure 5 . These findings suggest that drug-induced liver injury can lead to intestinal dysbiosis, characterized by reduced microbial diversity, increased potential pathogenic bacteria, and reduced beneficial bacteria, playing an important role in the development of DILI.

[0049] 4. Experimental conclusion

[0050] The above experiments found that there were significant differences in metabolite distribution between DILI patients and healthy control (HC) groups, which provided potential biomarkers for early non-invasive diagnosis of DILI and important evidence for exploring the disease mechanism. In terms of intestinal microecology, DILI can cause significant changes in intestinal flora structure, and this change is closely related to the disease state. Compared with HC, DILI patients have significant dysbiosis, manifested as decreased bacterial diversity, increased potential pathogenic bacteria, and decreased beneficial bacteria. By analyzing the correlation between the main difference flora and the main difference plasma metabolites, it was found that there was a wide and significant correlation between the two. These main difference metabolites and the expression levels of the above key difference bacterial genera showed strong relevance, further confirming the "gut-liver axis" in the occurrence and development of DILI, and providing important evidence for in-depth understanding of the pathogenesis of DILI and the development of targeted microecological intervention strategies.

[0051] II. Data processing

[0052] A total of 56 study subjects were included in this study, and the Majorbio Cloud (Majorbio Bio-Pharm Technology Co. Ltd., Shanghai, China, https: / / cloud.majorbio.com) online platform was used for metabolomics and intestinal flora analysis. At the same time, the data to be analyzed were entered into SPSS 26.0 for statistical analysis. The normality test was performed on the measurement data, and the results were expressed as ±s or M(QR). According to the normality, t-test or non-parametric test was used for comparison between groups. The count data was expressed as the number of cases (percentage), and the comparison between the two groups was performed according to the data distribution characteristics using chi-square test, corrected chi-square test or Fisher's exact test. Sensitivity, specificity, positive predictive value, and negative predictive value were used to evaluate the effectiveness of intestinal flora-related metabolites as diagnostic tools. Model testing used the receiver operating characteristic curve (ROC) and calculated the area under the curve (AUC). Chi-square test was used to evaluate the difference between different components, and p<0.05 was considered statistically significant. Visualization of experimental data was performed using GraphPad Prism 9 software, R language (v4.5.1), and Python (v3.12.9). The above software can meet all statistical analysis and plotting needs of this study.

[0053] III. Discriminatory diagnosis model of DILI and HC intestinal flora-related plasma metabolites

[0054] (1) In DILI and HC, first, for plasma metabolites, OPLS-DA / PLS-DA was used as a supervised model to test and predict the different changes of paired samples by 7-fold cross validation. Variable projection importance analysis of the first principal component determined the important metabolites that contributed to classification, and based on VIP value analysis, 50 key metabolites with the best ROC analysis effect in DILI and HC classification were screened out Figure 2 ); then for intestinal bacteria, the difference in bacterial flora at the species level between DILI and HC was analyzed by rank sum test, and the top 10 main difference bacteria were screened out Figure 5 based on relative abundance value, which provided a new perspective for understanding the metabolite characteristics and bacterial characteristics of DILI and HC, and also provided a theoretical basis for developing diagnostic markers and intervention targets based on specific intestinal flora-related metabolites.

[0055] (2) Spearman correlation analysis was performed on the above 50 VIP value difference metabolites and the top 10 main difference bacteria based on relative abundance value based on Spearman correlation coefficient. According to the results of spearman correlation analysis Figure 6 ), we screened out the top 3 key endogenous metabolites N-Choloylglycine (common name: Glycine N-Cholate; CAS Registry Number: 475-31-0), Carnosine (common name: L-Carnosine; CAS Registry Number: 305-84-0), Val Phe (full name: Valylphenylalanine; common name: Valine-phenylalanine; CAS Registry Number: 3918-92-1) with the strongest correlation, and evaluated the diagnostic performance of the above three intestinal flora-related metabolites. Stratified random sampling was realized by the train_test_split function in python, and 56 research objects (including 23 HC and 33 DILI) were divided into training set (n=45) and validation set (n=11) at a ratio of 8:2. The K-fold cross-validation method (K=5) was used to predict the model in the validation set. Receiver operating characteristic (ROC) curve analysis showed that these metabolites showed excellent diagnostic value for DILI: N-Choloylglycine (AUC=0.874), Carnosine (AUC=0.940), Val Phe (AUC=1). It is worth noting that among single metabolites, Val Phe showed the most excellent diagnostic performance (AUC=1), with a sensitivity and specificity of 1.0 Figure 7 ). These results suggest that specific intestinal flora-related plasma metabolites not only have important pathophysiological significance, but also can be used as non-invasive biomarkers for the auxiliary diagnosis of drug-induced liver injury.

[0056] (3) Based on the ROC curve analysis results, we optimized the multi-metabolite joint diagnosis model for metabolites with excellent diagnostic performance. The study found that the joint diagnosis model composed of the above 3 intestinal flora-related metabolites showed the best disease prediction ability (AUC=1.000, sensitivity 100%, specificity 100%, positive predictive value 100%, negative predictive value 100%) (Table 1). Figure 8 ), and the DILI prevalence probability was calculated by the logistic regression equation p(positive) = 1 / (1 + exp(-(2.8348 + -0.7049·Val Phe + 0.3937·N-Choloylglycine + -0.4499·Carnosine))). That is, when P(positive) > 0.412, it can be determined as DILI patients. This model shows excellent diagnostic performance (AUC=1.000, sensitivity 100%, specificity 100%, positive predictive value 100%, negative predictive value 100%) (Table 1).

[0057] Table 1: DILI and HC intestinal flora-related plasma metabolite differential diagnosis model characteristics

[0058] Four, DILI and HC intestinal flora-related plasma metabolite joint patient ALT differential diagnosis model

[0059] (1) In the analysis of patient basic information, we found that DILI patients had significantly higher ALT than HC patients (231.00(97.00-445.00) vs 22.00(17.00-27.00), P<0.0001) (Table 2).

[0060] Table 2: DILI and HC basic information and clinical data

[0061] Table 2 includes the basic information of age, gender, BMI, clinical manifestations of jaundice, ascites, hepatic encephalopathy, liver cirrhosis, clinical indicators of total bilirubin (TB), direct bilirubin (DB), alanine aminotransferase (ALT), aspartate aminotransferase (AST), alanine aminotransferase / aspartate aminotransferase (ALT / AST), alkaline phosphatase (ALP), gamma glutamyltransferase (GGT), albumin (ALB), international normalized ratio (INR) of healthy controls (n=23), DILI patients (n=33). Continuous variables were expressed as mean ± SD when they were normally distributed, and were tested by independent sample T test. Non-normally distributed data were expressed as median and were tested by Kruskal-Wallis test. p<0.05 was considered statistically significant.

[0062] Therefore, on the basis of the previously established metabolic marker model, the important clinical variable ALT was further integrated, and an innovative "gut microbiota-related metabolites-ALT" multi-parameter diagnostic model was constructed. The logistic regression prediction formula was established: p(positive) = 1 / (1 + exp(-(1.7976 + -2.4956·Val Phe + 1.1172·N-Choloylglycine + -1.1187·Carnosine + 0.1886·ALT))). The model showed perfect diagnostic performance (AUC=1.000) in the sample, with a sensitivity of 100% and a specificity of 100%. That is, when the gut microbiota-related metabolites are combined, the patient's ALT is calculated P(DILI) greater than 0.936, the patient is determined as a DILI patient. Figure 9

[0063] Table 3: DILI and HC gut microbiota-related plasma metabolites combined with ALT differential diagnosis model characteristics

[0064] (2) The model constructed in this study showed almost perfect diagnostic performance in distinguishing DILI patients from HC group. In particular, the validation set (n=15) showed a perfect diagnostic performance (AUC=1.000), with a sensitivity of 100% and a specificity of 100%. Figure 10 ​The AUC value of the model reached 1.000, which means its discriminative ability is superior to any model with an AUC less than 1. According to the optimal threshold value of 0.530 calculated, the model achieved full marks on all key performance indicators: sensitivity (100%), specificity (100%), positive predictive value (100%), negative predictive value (100%), accuracy (100%), and precision (100%). The Youden index was 1.000, further confirming that the model achieved the ideal state of zero misjudgment at the optimal threshold value. This result indicates that the model can completely and accurately distinguish DILI individuals from healthy populations in the study cohort, demonstrating great potential for clinical application, especially for early screening scenarios that require high precision and zero tolerance.

[0065] While embodiments of the present application have been shown and described, it is to be understood that the embodiments described are merely divergences of the principles and spirit of the present application, and that many changes, modifications, substitutions and variations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A biomarker for diagnosing drug-induced liver injury, characterized in that, The biomarkers are plasma metabolites, including glycocholic acid, L-carnosine, and pentylelanine.

2. The use of the detection reagent for the biomarker described in claim 1 in the preparation of diagnostic products for drug-induced liver injury.

3. The application according to claim 2, characterized in that, The products mentioned are reagent kits, detection systems, or instruments.

4. The application according to claim 2, characterized in that, The detection reagent is used to detect the abundance of biomarkers in plasma metabolites; And / or, the detection reagent includes at least one of the reagents for mass spectrometry analysis or the reagents for chromatographic analysis.

5. A kit for diagnosing drug-induced liver injury, characterized in that, The kit includes a plasma metabolite abundance detection reagent and a prediction formula; the plasma metabolite abundance detection reagent is used to detect the abundance of the biomarker of claim 1; the prediction formula is constructed based on the biomarker of claim 1.

6. The reagent kit according to claim 5, characterized in that, The prediction formula is: p(positive) = 1 / (1+ exp(-(2.8348 -0.7049)). Pentylphenylalanine abundance value +0.3937 The abundance value of glycocholic acid is -0.4499·L-carnosine abundance value. When P(positive)>0.412, the patient is identified as having drug-induced liver injury.

7. The reagent kit according to claim 5, characterized in that, The kit also includes a serum alanine aminotransferase (ALT) level detection reagent, and the prediction formula is: p(positive) = 1 / (1+exp(-(1.7976-2.4956)). Pentylphenylalanine abundance value +1.1172 Glycinecholic acid abundance value: -1.1187 L-carnosine abundance value + 0.1886 Serum alanine aminotransferase (ALT) concentration value; when P(positive) > 0.936, the patient is identified as having drug-induced liver injury.

8. The reagent kit according to claim 5, characterized in that, The kit also includes plasma metabolite extraction reagents.

9. A device for diagnosing drug-induced liver injury, characterized in that, It includes a detection unit and a data analysis unit, wherein: The detection unit is used to acquire plasma samples, detect and determine the abundance of plasma metabolites and the concentration of serum alanine aminotransferase; the plasma metabolites are the biomarkers described in claim 1; The data analysis unit is used to analyze the detection results of the detection unit in conjunction with a prediction formula to make a diagnostic result; the prediction formula is: p(positive) = 1 / (1+exp(-(1.7976-2.4956)). Pentylphenylalanine abundance value +1.1172 Glycinecholic acid abundance value: -1.1187 L-carnosine abundance value + 0.1886 Serum alanine aminotransferase (ALT) concentration value; when P(positive) > 0.936, the patient is identified as having drug-induced liver injury.

10. The apparatus according to claim 9, characterized in that, The detection unit includes at least one of LC-MS, GC-MS, LC-MS / MS detection reagents, systems, and instruments.

Images