A protein marker combination for masld staging screening and diagnosis and application thereof

Abstract

This invention provides a combination of protein biomarkers for staging screening and diagnosis of MASLD and its application. The combination of protein biomarkers includes AFM, CFHR5, APOA2, and ORM1. Based on this protein composition, a predictive model for MASLD staging screening and diagnosis is constructed. The model primarily uses serum samples, which, compared to other bodily fluid samples, offers advantages such as convenient extraction and less patient discomfort, providing a new direction for MASLD staging screening and diagnosis. This invention also develops accompanying predictive software based on the model formula. This software automatically outputs scores and grading results through a graphical user interface, making it easy to operate and suitable for clinical promotion and large-scale population screening. In practical clinical applications, this model can also be combined with the patient's specific medical history and other existing detection methods to improve diagnostic accuracy, thereby increasing the disease detection rate and ultimately helping to improve patient prognosis.

Description

Technical Field

[0001] This invention relates to the field of screening and diagnosis of metabolic-associated fatty liver disease, specifically to a combination of protein biomarkers for screening and diagnosis of MASLD staging and its application. Background Technology

[0002] The global disease spectrum is undergoing a significant transformation, with metabolic-associated fatty liver disease (MASLD) emerging as a new public health challenge. Epidemiological data shows that the disease prevalence in the global adult population has reached one-quarter, and its systemic health hazards are significant—all-cause mortality data indicate approximately 12.6 deaths per 1,000 person-years, with cardiovascular events accounting for the highest proportion (4.2 cases), followed by malignant tumors (2.83 cases) and primary liver diseases (0.92 cases). Notably, this disease exhibits a significant comorbidity with various components of metabolic syndrome. When the disease progresses to metabolic-associated steatohepatitis (MASH), the risk of hepatocellular carcinoma and liver-related mortality increases tenfold, highlighting the importance of early identification.

[0003] In terms of clinical diagnostic methods, although traditional liver biopsy is considered the benchmark for pathological diagnosis, its application has significant limitations, including invasiveness (potentially leading to complications such as bleeding and organ damage), high technical requirements, and low cost-effectiveness. Non-invasive imaging techniques each have their advantages and disadvantages: conventional ultrasound and CT can observe morphological changes in the liver, but lack the ability to differentiate early lesions; MRI, while free of ionizing radiation, is limited by examination time and economic costs; transient elastography (such as Fibroscan) can assess the degree of liver fibrosis, but is easily affected by other liver disease factors.

[0004] Regarding serological testing, although MASH patients are known to frequently exhibit abnormal liver enzyme profiles (ALT / AST) and elevated inflammatory markers, these indicators lack disease specificity. Existing non-invasive diagnostic models based on traditional biomarkers generally suffer from insufficient accuracy or lack of cross-population validation. While composite scoring systems such as FAST and NIS4, developed in recent years, integrate multimodal data, their predictive efficacy for long-term prognosis still requires further validation. Against this backdrop, developing novel non-invasive diagnostic tools with high sensitivity and specificity has significant clinical value.

[0005] Proteomics, as an important research tool in the post-genomic era, can systematically analyze the expression, modification, and interactions of proteins in disease states, providing new insights for discovering highly specific biomarkers. Compared to metabolomics, proteomics directly reflects the end products of gene expression and is closer to the functional phenotype of diseases. In recent years, the rapid development of mass spectrometry (MS) and high-throughput protein chip detection technologies has made large-scale, high-precision protein quantification possible. Of particular note is serum / plasma proteomics, which offers advantages such as convenient sampling, high stability, and dynamic monitoring, making it highly suitable for the development of non-invasive diagnostic models and providing new insights for assessing and diagnosing metabolic-related fatty liver diseases.

[0006] Advances in artificial intelligence have further propelled the application of proteomics in the medical field. Machine learning can unearth potential biomarker combinations from massive amounts of protein data and construct high-precision predictive models. Recent studies show that proteomics and machine learning-based methods have made breakthroughs in areas such as cancer and cardiovascular diseases, but their application in the diagnosis of MASL, MASH, and liver fibrosis is still in the exploratory stage. Therefore, integrating proteomics and machine learning technologies to develop non-invasive predictive models for MASL, MASH, and liver fibrosis applicable to multiple population groups holds promise for overcoming the shortcomings of existing diagnostic methods and providing a more reliable basis for clinical decision-making.

[0007] To address the aforementioned issues, this study explores novel non-invasive scoring models based on proteomics and machine learning technologies. These models are of great significance for early screening and diagnosis of simple fatty liver (MASL), early-stage MASH, and fibrotic MASH, thereby improving patient prognosis. Summary of the Invention

[0008] This invention provides a combination of protein biomarkers for staging screening and diagnosis of MASLD and its application. The combination of protein biomarkers includes AFM, CFHR5, APOA2, and ORM1. The levels of the four selected serum proteins can be obtained by various methods, including mass spectrometry and ELISA methods on different platforms. After Z-score normalization, a predictive model for staging screening and diagnosis of MASLD is calculated, enabling non-invasive, rapid, and accurate stratified diagnosis of simple fatty liver (MASL), early metabolic-associated steatohepatitis (early MASH), and metabolic-associated steatohepatitis with fibrosis (fibrotic MASH).

[0009] The technical solution adopted by the present invention to achieve the above objectives is as follows: In the first aspect of the present invention, a combination of protein biomarkers for staging screening and diagnosis of MASLD is disclosed, wherein the combination of protein biomarkers includes AFM, CFHR5, APOA2 and ORM1.

[0010] A second aspect of the present invention discloses the application of the above-described protein biomarker combination in the preparation of a kit for MASLD staging screening and diagnosis.

[0011] Preferably, the test sample for the kit is serum.

[0012] A third aspect of the present invention discloses a method for constructing a staging screening and diagnostic model for MASLD, comprising the following steps: (1) Data collection and grouping: The liver biopsy cohort was selected as the modeling sample. All subjects were excluded from excessive drinking, viral hepatitis and other chronic liver diseases, and history of hepatotoxic drug use. Demographic information, physical indicators, biochemical indicators, liver biopsy pathology results and fasting serum samples were collected from the subjects. According to the SAF scoring system, the subjects were divided into non-MASLD, simple fatty liver MASL, early MASH and fibrotic MASH groups. (2) Screening of protein biomarker combinations: Differentially expressed proteins among different MASLD stage groups were screened. Combined with liver tissue transcriptome and proteome data, common differentially expressed proteins with consistent expression trends and enrichment in the same metabolic pathway were screened. Based on the protein enrichment characteristics, secretion properties and correlation with liver pathology scores, 13 candidate proteins were initially screened. The protein biomarker combination as described in claim 1 was determined using the optimal subset regression algorithm. (3) Determine the model variable set: Combine the protein biomarkers and multiple indicators together as model input variables; (4) Data standardization and variable assignment: Z-score standardization was performed on the expression data of protein markers; binary classification 0-1 quantization was performed on the two categories of categorical variables, gender and type 2 diabetes status; (5) Construct a multi-factor logistic regression scoring model: Simple fatty liver MASL, early MASH, and fibrotic MASH are used as binary outcome variables respectively. The multi-factor stepwise logistic regression algorithm is used to fit the model. Based on the regression constant term and the regression coefficients of each variable, the scoring formulas for simple fatty liver MASL, early MASH, and fibrotic MASH are constructed respectively. (6) Setting diagnostic thresholds: Based on ROC curve analysis of the modeling cohort, high thresholds were set for simple fatty liver MASL and early MASH, and the diagnostic cutoff value for fibrotic MASH was determined; wherein the high thresholds were used for positive diagnosis of the disease. (7) Model validation: Conduct internal and external validation, evaluate model performance, and complete the construction of the model.

[0013] Preferably, the 13 candidate proteins in step (2) include: AFM, C3, CP, ITIH1, RBP4, APOC3, ORM1, HRG, KNG1, SERPINC1, VTN, CFHR5 and APOA2.

[0014] Preferably, the multiple indicators in step (3) include BMI, gender, type 2 diabetes status and FIB4 index.

[0015] Preferably, the MASL scoring formula for simple fatty liver, the MASH scoring formula for early stage, and the MASH scoring formula for fibrosis in step (5) are as follows: Simple fatty liver MASL score = -6.297 + 1.283 × AFM – 0.645 × APOA2 + 0.345 × CFHR5 + 0.643 × ORM1 + 0.294 × BMI + 1.667 × gender + 1.271 × type 2 diabetes Early MASH score = -4.702 + 1.045 × AFM – 0.182 × APOA2 – 0.055 × CFHR5 – 0.203 × ORM1 + 0.183 × BMI – 0.179 × gender + 0.666 × type 2 diabetes Fibrosis MASH score = -5.951 + 0.540 × AFM – 0.422 × APOA2 + 0.284 × CFHR5 – 0.206 × ORM1 + 0.109 × BMI – 0.004 × gender + 1.216 × type 2 diabetes + 1.058 × FIB4.

[0016] Preferably, the binary classification 0-1 quantization of gender is assigned as male=1 and female=0; the binary classification 0-1 quantization of type 2 diabetes is assigned as yes=1 and no=0.

[0017] Preferably, in step (6), the high threshold for simple fatty liver MASL is ≥2.80, the high threshold for early MASH is ≥2.19, and the diagnostic cutoff value for fibrotic MASH is ≥0.48.

[0018] A fourth aspect of the present invention discloses a system for staging screening and diagnosis of MASLD, comprising: The data acquisition module is used to acquire the following data of the sample to be tested through the data input area of ​​the graphical user interface: expression level data of the above protein biomarker combination after Z-score normalization, and clinical variable data, including BMI value, gender, type 2 diabetes status and FIB4 index value. The model calculation module is used to input the expression level data of the protein biomarker combination and the clinical variable data into the MASLD staging model, and automatically calculate the scores of each dimension according to the above-mentioned MASL scoring formula for simple fatty liver, MASH scoring formula for early stage and MASH scoring formula for fibrosis. The grading output module is used to determine the MASLD histological grading based on the scores of each dimension and the predetermined critical threshold, and to display the MASLD grading results through the graphical user interface. The grading results include one of non-MASLD, simple fatty liver, early MASH, or fibrotic MASH.

[0019] In some embodiments of this application, the target population may include different individuals, such as healthy individuals, individuals with hepatic steatosis but not diagnosed with MASLD, and individuals suspected of having MASLD. Serum samples are collected from the subjects to be tested, and the protein concentration in the body fluid samples is quantitatively detected using an instrument. The graded diagnosis of MASLD (simple fatty liver MASL, early metabolic-associated steatohepatitis, metabolic-associated steatohepatitis with fibrosis) is calculated using a formula, and the long-term all-cause mortality risk is predicted.

[0020] Compared with the prior art, the beneficial effects that at least one technical solution adopted in the embodiments of this specification can achieve include at least: (1) The present invention screens a combination of protein biomarkers. Based on this, serum proteomics and machine learning technology are used to construct a screening and diagnostic model that can non-invasively and accurately assess each stage of MASLD (simple fatty liver MASL, early MASH, fibrotic MASH). This effectively avoids the trauma, risk and high cost of liver biopsy. By using this model in conjunction with traditional diagnostic criteria, the risk of MASH can be effectively predicted, making the screening results more reliable. (2) This invention uses multiple clinical parameters and proteomics, and uses machine learning to optimize the model, which can more comprehensively capture the various lesions and functional states of the liver during the course of MASLD. It can be used as a new non-invasive assessment method for screening MASLD risk and clinical prognosis. (3) The model of the present invention has been validated in multiple independent cohorts, demonstrating excellent cross-population applicability and stability; (4) This invention uses serum as the main test sample, which is convenient to collect, highly standardized, well accepted by patients, low in testing cost, and low in economic burden on patients. It is easy to promote in clinical practice and large-scale screening. (5) Based on the model formula, the present invention has developed a supporting system for MASLD staging screening and diagnosis. Through a graphical interactive interface, the system realizes the automatic output of variable input, score calculation and grading results. It is easy to operate and does not require professional programming background. This further reduces the threshold for using the model of the present invention in clinical practice, makes it easier for front-line medical staff to quickly obtain prediction results, and improves the accessibility of the model and the efficiency of practical application transformation. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0022] Figure 1 The image shows the proteomic features of MASLD histological progress, where (A) is a heatmap of the dynamic changes in protein abundance of representative pathways and the enrichment of functional pathways in the histological progress features; and (B) is a protein interaction network diagram related to the histological progress features. Figure 2 The following is a graph showing the results of the analysis of MASLD-related biomarkers screened based on liver biopsy proteomics. (A) is a heatmap showing the normalized expression intensity of 13 proteins that are significantly associated with the progression stage of MASLD; (B) is a graph of MASH candidate feature screening based on optimal subset regression, which reduces 13 core variables to 4 MASH candidate variables to maximize the adjusted coefficient of determination; (C) is a grouped quantitative expression graph showing the quantitative expression intensity of proteins screened from the tissue proteome in subjects at different pathological progression stages of MASLD. In the graph, eMASH represents early MASH and fMASH represents fibrotic MASH. Figure 3 The diagnostic efficacy of four protein combination biomarkers in Chinese and Danish populations is shown in the graphs. (A) is the receiver operating characteristic (ROC) curve of the four protein combinations for predicting MASL, MASH and advanced liver fibrosis, corresponding to the LIVERMONITOR Phase I cohort (liver tissue proteome), Phase II cohort (serum proteome), and the Shanghai and Beijing validation cohorts in China, respectively; (B) is the ROC curve of the four protein combinations for predicting MASL and MASH in the Danish liver biopsy cohort based on liver tissue and serum samples. Figure 4The graph shows the accuracy of the predictive model of this invention in predicting liver-related mortality in a UK biobank cohort. (A) is a flowchart of the two-step non-invasive diagnostic method for MASLD histological staging, with a diagnostic cutoff value set to ensure a specificity of 95%; (B) is a Venn diagram showing the proportion of patients with simple metabolic-associated fatty liver disease (MASL) and metabolic-associated steatohepatitis (MASH); (C) is a Venn diagram showing the proportion of fibrotic MASH among all MASH patients; and (D) is a survival curve showing that the risk of liver-related mortality is significantly higher in individuals identified as high-risk by the model than in those without MASLD. p <0.001; Cox proportional hazards model analysis was used); (E) is a survival curve showing that the risk of liver-related death in patients with fibrotic MASH is approximately 8 times that of patients with early-stage MASH ( p <0.001; Cox proportional hazards model analysis was used); In the figure, eMASH represents early MASH, and fMASH represents fibrotic MASH; Figure 5 This invention provides a system interface for MASLD staging screening and diagnosis. In the data input area of ​​the graphical user interface, the user sequentially enters the expression values ​​of four protein markers (AFM, APOA2, CFHR5, ORM1) after Z-score standardization, as well as BMI (kg / m²), gender (male / female), type 2 diabetes status (yes / no), and FIB4 index value. After inputting the data, the user clicks the "Predict" button, and the software automatically calculates the results in the background. Finally, the software simultaneously outputs the MASH score, fibrotic MASH (fMASH) score, and the comprehensive MASLD grading result in the "Prediction Result" display area. Figure 6To further simplify the operation process, a system interface for MASLD staging screening and diagnosis was established based on the AFM, APOA2, CFHR5, and ORM1 values ​​measured by the ELISA method. In the data input area of ​​the software's graphical user interface, users sequentially enter the concentrations of the four proteins directly measured using ELISA kits manufactured by RayBiotech, including AFM (RayBiotech, Cat no. ELH-AFM-1), APOA2 (RayBiotech, Cat no. ELH-ApoA2-1), CFHR5 (RayBiotech, Cat no. ELH-CFHR5-1), and ORM1 (RayBiotech, Cat no. ELH-ORM1-1), as well as BMI (kg / m²), gender (male / female), type 2 diabetes status (yes / no), and FIB4 index value. After inputting the data, the user clicks the "Predict" button, and the software automatically calculates the results in the background. Finally, the software simultaneously outputs the MASH score, the fibrous MASH (fMASH) score, and the comprehensive MASLD classification result in the "PredictionResult" display area. Detailed Implementation

[0023] The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. Without departing from the concept of the present invention, those skilled in the art can make various modifications or substitutions to the present invention, and all such modifications or substitutions should fall within the scope of protection of the present invention.

[0024] Example 1: Clinical Sample and Multi-omics Data Acquisition Methods for Screening Serum Protein Biomarkers at Different Stages of MASLD 1. Study population and ethical statement Liver biopsy samples from participants of the Liver Monitoring Program (LIVERMONITOR: Liver Multi-omics Dynamic Risk Non-invasive Detection Program) were used to determine the liver multi-omics atlas of different stages of MASLD. The LIVERMONITOR program prospectively enrolled participants who underwent liver biopsy at Zhongshan Hospital affiliated with Fudan University (Shanghai, China) for suspected metabolic-associated steatohepatitis (MASH), aiming to screen for novel multi-omics risk factors driving MASLD pathological progression and related metabolic outcomes. The program consisted of two phases: Phase I focused on liver tissue, and Phase II focused on serum multi-omics biomarkers. A 1:1:1:1 propensity score matching strategy was used to select 100 participants (26 non-MASLD, 24 simple fatty liver MASL, 26 early MASH, and 24 fibrotic MASH), matched for age and sex; after excluding 4 patients with insufficient liver tissue protein, 96 patients were ultimately included in the Phase I analysis. The second phase (II) of the LIVERMONITOR project included 196 participants (46 non-MASLD, 47 MASL with simple fatty liver, 47 early MASH, and 56 fibrotic MASH), with comparable age and sex distribution, used to screen for serum multi-group biomarkers associated with MASLD progression; after excluding 34 patients who were included in the first phase analysis, a total of 162 participants were included in this study analysis.

[0025] Liver transcriptomic and proteomic atlases comprehensively revealed the molecular regulatory network driving the pathological progression of MASLD. Combined multi-omics analysis of liver tissue and serum proteomics identified common candidate pathways and biomarkers associated with MASLD progression. The LIVERMONITOR project employed a multi-step machine learning strategy to establish a serum predictive model for non-invasive identification of patients with simple fatty liver (MASL), early-stage MASH, and fibrotic MASH. This model was validated in an internal validation cohort of 245 patients in Shanghai, China, and two external validation cohorts (including 90 patients in Beijing, China, and 41 patients in Denmark). All three liver biopsy cohorts excluded patients with excessive alcohol consumption, chronic liver diseases other than MASLD / MASH (including hepatitis B or C), or those using hepatoprotective / hepatotoxic drugs. Furthermore, in the UK Biobank, a study was conducted among 39,920 participants with proteomic data to investigate whether this predictive model could predict long-term mortality from liver disease and related metabolic disorders. This study was conducted in accordance with the Declaration of Helsinki, and the study protocol was approved by the Ethics Review Committee of Zhongshan Hospital Affiliated to Fudan University (No. B2022-152) and the Ethics Review Committee of Beijing You'an Hospital Affiliated to Capital Medical University (No. 2022053). All patients signed written informed consent forms.

[0026] 2. Pathological staging of MASLD The liver pathology features were assessed by a hepatologist who was unaware of all other clinical data. The severity of MASLD was assessed using the SAF scoring system (steatodegeneration, activity, and fibrosis score). The SAF score consists of steatosis (grades 0-3), activity (grades 0-4), and fibrosis (grades 0-4), with activity comprised of ballooning degeneration and lobular inflammation. Patients were further categorized according to the SAF scoring system: non-MASLD (steatodegeneration <1 point), simple fatty liver MASL (steatodegeneration ≥1 point, hepatocellular ballooning degeneration or lobular inflammation = grade 0), early MASH (eMASH, steatosis ≥1 point, hepatocellular ballooning degeneration ≥1 point, lobular inflammation ≥1 point, fibrosis ≤1 point), and fibrotic MASH (fMASH, steatosis ≥1 point, hepatocellular ballooning degeneration ≥1 point, lobular inflammation ≥1 point, fibrosis ≥2 points).

[0027] 3. Physical and biochemical tests Participants were barefoot and naked before their weight and height were measured, and their body mass index (BMI) was calculated. Systolic and diastolic blood pressure were measured using an electronic blood pressure monitor. After an overnight fast, venous blood samples were collected from participants, and biochemical parameters, including peripheral blood platelet count, fasting plasma glucose (FPG), 2-hour postprandial blood glucose, alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), alkaline phosphatase (ALP), total cholesterol, triglycerides, and high-density lipoprotein cholesterol (HDL-C), were measured using the Friedewald formula. Low-density lipoprotein cholesterol (LDL-C) was estimated using the Friedewald formula. The liver fibrosis-4 index (FIB4) was calculated by analyzing patients' age, AST, ALT, and platelet count. Glycated hemoglobin (HbA1c) was measured using high-performance liquid chromatography (HPLC) and certified by the National Glycated Hemoglobin Standardization Program. Serum insulin concentration was measured using electrochemiluminescence immunoassay. According to the American Diabetes Association criteria, diabetes is defined as HbA1c ≥ 6.5%, FPG ≥ 7.0 mmol / L, 2-hour postprandial or random blood glucose ≥ 11.1 mmol / L, or a previous diagnosis of diabetes, or self-reported use of hypoglycemic medications.

[0028] 4. Preparation of RNA and protein samples DNA, RNA, and protein were extracted from liver tissue using the AllPrep DNA / RNA / Protein Mini Kit (Qiagen, catalog number 80004). Liver tissue samples approximately 1 cm in length and 1 mm in cross-sectional area were homogenized using a tissue homogenizer. Genomic DNA was first bound to an AllPrep DNA column, washed, and eluted. Ethanol was then added to the eluent, and RNA was purified using an RNeasy column and stored at -80°C for later use. Tissue proteins were precipitated from the remaining eluent using Buffer APP (a novel aqueous protein precipitant), and the precipitate was collected by centrifugation. The precipitate was resuspended and denatured by adding 8M urea buffer (8M urea, 100mM ammonium bicarbonate, 1× protease inhibitor [Roche, catalog number 04693159001], 2× PhosSTOP phosphatase inhibitor [Roche, catalog number 04906845001]), and centrifuged at 21000g. The protein-containing supernatant was collected. Serum samples were prepared according to the previously described procedure. High-abundance proteins were removed using a highly selective human serum albumin / immunoglobulin removal centrifuge column (Thermo Fisher Scientific, catalog number A36366). Protein concentrations in tissues and serum were determined using the BCA method. Subsequently, protein samples were reduced by adding 5 mM DTT and incubating at 56°C for 30 minutes; alkylated by adding 15 mM IAA and incubating at room temperature in the dark for 30 minutes; and quenched by adding 20 mM DTT and incubating at room temperature for 30 minutes. The protein solution was diluted with 100 mM ammonium bicarbonate to a final urea concentration below 2 M, and trypsin (Beijing Shengxia Protein Technology Co., Ltd., catalog number HLSTRY001C; enzyme:protein = 1:50, weight ratio) was added and incubated at 37°C for 16 hours. A second round of enzymatic digestion was performed by adding trypsin (enzyme:protein = 1:100, weight ratio) and incubating for an additional 3 hours to ensure complete protein digestion. The enzymatically hydrolyzed peptides were desalted using a Sep-Pak tC18 column (Waters, WAT054960) and dried using a SpeedVac vacuum dryer. The desalted peptides were stored at -80℃ for later use.

[0029] 5. RNA sequencing RNA sequencing services were provided by CloudSeq (Shanghai, China). Ribosomal RNA (rRNA) was removed from the samples using the GenSeq® rRNA Removal Kit (GenSeq). Sequencing libraries were constructed from the rRNA-removed samples using the GenSeq® Low Input RNALibrary Prep Kit (GenSeq). The quality and quantity of the prepared libraries were assessed using a BioAnalyzer 2100 system (Agilent Technologies, USA), followed by 150bp paired-end sequencing. Raw sequencing data were obtained, filtered using fastP software (v0.22.0), and then quality controlled using fastqc software (0.12.1) and multiqc software (1.14). High-quality reads were aligned to the human reference genome GRCh38.p13 using STAR software (2.7.11b), and the alignment results were converted to BAM format using samtools software (1.22). Transcriptome quantification was performed using rsem software (v1.3.3) in conjunction with the GeneCode v43 annotation file. The FPKM values ​​generated by rsem were used for subsequent differential analysis.

[0030] 6. Tissue proteomics analysis based on tandem quality tag (TMT) markers The enzymatically digested peptides from tissue samples were dissolved in 100 mM triethylamine bicarbonate (TEAB) buffer. Following the manufacturer Thermo Fisher's instructions, 11-tandem quality tagging (TMT) reagents (Thermo Fisher, catalog numbers 90406 and A37725) were used for labeling. Simultaneously, to facilitate batch-to-batch comparisons, equal volumes of peptides from all samples were mixed to create an internal reference sample. The 11-tag TMT labeling procedure was as follows: First, all samples were grouped into 11 groups, with 10 samples from tissue samples originating from different clinical cases and 1 sample serving as the internal reference. The 10 peptide samples from different tissue sources within each group were labeled using TMT channels -127N, -127C, -128N, -128C, -129N, -129C, -130N, -130C, -131N, and -131C, respectively. The internal reference sample was labeled using TMT channel -126. After labeling, the 11 TMT-labeled samples in each group were mixed in equal amounts according to peptide content in a ratio of 1:1:1:…:1.

[0031] The mixed TMT-labeled peptide samples were further separated using high-pH reversed-phase liquid chromatography. The instruments and consumables used were as follows: an Agilent Technologies HPLC system (model 1100); and a Waters XBridge BEH C18 column (catalog number: 186003116). An 85-minute linear gradient was used to fractionate the TMT-labeled peptide samples. Mobile phase A was a 2% acetonitrile aqueous solution (pH adjusted to 10.0 with ammonia), and mobile phase B was a 98% acetonitrile aqueous solution (pH adjusted to 10.0 with ammonia). Specific separation conditions were as follows: flow rate maintained at 1 mL / min; 0-10 minutes, mobile phase B concentration gradually increased from 2% to 10%; 10-60 minutes, mobile phase B concentration gradually increased from 10% to 40%; 60-80 minutes, mobile phase B concentration gradually increased from 40% to 85%; 80-85 minutes, mobile phase B concentration was maintained at 85%. Liquid chromatography was performed using a UV detector at a wavelength of 210 nm to detect the separated peptides. The eluent was collected every 1 minute, for a total of 85 fractions. Finally, all fractions were combined into 30 fractions in the order of "pre-middle-post" of receipt, and the combined fractions were dried using a SpeedVac vacuum dryer.

[0032] Each fractionated component was dissolved in buffer A (0.1% formic acid, 2% acetonitrile (ACN) aqueous solution) and separated using an Easy-nLC 1200 system (Thermo Fisher Scientific) with a self-made C18 capillary column (ID 75 μm × 220 mm, particle size 1.9 μm, pore size 120 Å, Dr. Maisch). Separation conditions were: flow rate 300 nL / min, 70-minute linear gradient: 0–57 min, 6–30% buffer B (0.1% formic acid, 80% acetonitrile aqueous solution); 57–61 min, 30–45% buffer B; 61–65 min, 45–80% buffer B; 65–70 min, 80% buffer B. Samples were analyzed by tandem mass spectrometry (MS / MS) using a Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific).¹ 4After elution using a C18 capillary column, the peptides were ionized at a nanofluidic electrospray ionization source with a voltage of 2.1 kV. All ionized peptide ions were then detected by mass spectrometry. MS1 parameters: a high-voltage electrostatic orbital trap (Orbitrap) was used as the detector; resolution was set to 120,000 (m / z 200); mass-to-charge ratio (m / z) was 350-1600; automatic gain control (AGC) target value was 3E6; and the maximum ion implantation time was 45 ms. MS2 (secondary mass spectrometry) parameters: A high-voltage electrostatic orbital trap (Orbitrap) was used as the detector, with a resolution of 45000 m / z 200. The top 12 ions with the highest MS1 signal intensity were selected for secondary fragmentation, with a high-energy collision dissociation (HCD) collision energy of 32%. Precursor ions with charges of +1 and >+6 were excluded. The scan range had a mass-to-charge ratio of 200-2000; the automatic gain control (AGC) target value was 1E5; the maximum ion implantation time was 30 ms; the isolation window had a mass-to-charge ratio of 0.8; and the dynamic exclusion time was 30 ms. The mass spectrometry data were then analyzed using MaxQuant software (v2.0.1.0) and compared with the UniProt human database (downloaded May 10, 2020). For tissue proteomics, MS2 reporter ion quantification based on 11-fold TMT was used, with a precursor intensity fraction (PIF) of 0.75. Fixed modifications were set to cysteine ​​carbamylation, and variable modifications were set to methionine oxidation and protein N-terminal acetylation. A maximum of two missed cleavage sites were allowed. The false discovery rate (FDR) for both peptides and proteins was controlled below a threshold of 0.01.

[0033] 7. Proteomics Data Processing Tissue protein quantification was performed using the proteinGroups.txt file from the MaxQuant analysis results. For serum proteomics, label-free quantification was used. Proteins labeled as reverse proteins, potential contaminants, or with more than 50% missing values ​​in all four pathological stages were removed. All remaining missing values ​​were imputed using the low-abundance imputation method in Perseus software. Tissue proteins were normalized to median, with the log2 transformation ratio of the sample to the internal reference sample used as the standardized protein intensity. Serum proteins were normalized using the Total Scale of Occurrence (FOT) method, multiplied by 1E5, and then Z-score converted for subsequent analysis. Batch effects were eliminated using the `combat` function in the `sva` package of R software.

[0034] 8. Serum proteomics analysis Serum proteomic analysis was performed using liquid chromatography-mass spectrometry (LC-MS) with an EASY-nLC 1200 ultra-high pressure system (Thermo Fisher Scientific), coupled with a nano-electrospray ionization source (Thermo Fisher Scientific) and a Fusion Lumos Orbitrap mass spectrometer (Thermo Fisher Scientific). The 14 most abundant proteins in the serum samples were removed by immunoassay. Trypsin was added at a 1:25 enzyme-to-protein ratio, and the samples were digested overnight at 37°C. The digests were then extracted and dried. LC-MS / MS detection was performed using the EASY-nLC 1200 ultra-high pressure liquid chromatography system in series with the Fusion Lumos Orbitrap mass spectrometer, with nano-electrospray ionization source. Peptides were dissolved in loading buffer and loaded onto a C18 capture column, followed by linear gradient elution on a C18 analytical column. Data acquisition was performed in data-independent acquisition (DIA) mode. FragPipe software combined with the MSFragger engine was used to compare DIA data with the UniProt human protein database for peptide and protein identification. Fixed and variable modifications were set, and reverse database screening was used to control the false positive rate (FDR) of protein identification to within 1%. SpectraST was used to construct a spectral library. Subsequently, DIA-NN software was used for DIA data analysis, and protein quantification was performed using the label-free intensity absolute quantification method (iBAQ). Protein abundance was normalized using total fraction (FOT), and missing values ​​were set to 10. -5 Imputation was performed to complete the qualitative and quantitative analysis of serum proteomics.

[0035] 9. ELISA method for detecting serum protein biomarkers Serum samples were retrieved at -80℃ and biomarker detection was performed simultaneously. The levels of AFM, APOA2, CFHR5, and ORM1 proteins in serum were measured using the following kits: Human AFM ELISA Kit (RayBiotech, catalog number ELH-AFM-1), Human Apolipoprotein A-II (APOA2) ELISA Kit (RayBiotech, catalog number ELH-ApoA2-1), Human Complement Factor H-Related Protein 5 (CFHR5) ELISA Kit (RayBiotech, catalog number ELH-CFHR5-1), and Human α1-Acid Glycoprotein (ORM1) ELISA Kit (RayBiotech, catalog number ELH-ORM1-1).

[0036] 10. Statistical Analysis All statistical analyses were performed using R software (v4.1.0). Continuous parameters are expressed as mean ± standard deviation (SD), and categorical parameters are expressed as number of cases (percentage). The Wilcoxon rank-sum test was used to compare mRNA and protein expression levels among different pathological stages of MASLD. The Wilcoxon rank-sum test was used to calculate the fold change (FC) and p-value between groups. Venn diagram analysis was performed on significantly upregulated or downregulated RNA (p-value < 0.05 and FC ≥ 1.5) and protein (p-value < 0.05 and FC ≥ 1.2). Pathway enrichment analysis was performed using the WebGestaltR package (v1.0.0) in conjunction with the Gene Ontology (GO) database to comprehensively visualize the molecular regulatory pathways of MASLD pathological progression. Biomarkers showing consistent changes in expression in the liver transcriptome and proteome at all stages of MASLD were screened as candidate biomarkers for predictive models applicable to all different stages of MASLD. Spearman correlation analysis of tissue proteome / transcriptome with NAFLD activity score (NAS) was performed to further screen proteins with p-values ​​<0.05 and absolute correlation coefficients >0.2. Proteins belonging to differentially expressed pathways in both the liver tissue proteome and transcriptome were input into the String database to obtain protein-protein interaction networks supported by experimental evidence or database annotations and with confidence scores >0.7. The networks were visualized using Cytoscape software (v3.10.2).

[0037] Biomarkers of highly expressed, secreted proteins in the liver were screened using the Human Proteome Database. These proteins were required to rank in the top 25% of the proteome and exhibit consistent expression trends in liver tissue and serum proteomes. Optimal subset regression (using R package leafs 3.1) was then used to further refine the candidate biomarkers. Based on these selected biomarkers, multivariate logistic regression analysis was used to establish predictive models for simple fatty liver disease (MASLD), early-stage fatty liver disease (MASH), and fibrotic MASH. These models were used for staging and diagnosis of MASLD and validated in an internal validation cohort in Shanghai, China, and two external validation cohorts in Beijing, China, and Denmark. The area under the receiver operating characteristic (AUROC) curve was used to assess diagnostic efficacy. All statistical tests were two-tailed, and a p-value <0.05 was considered statistically significant.

[0038] Example 2: Joint analysis of serum and liver tissue proteomes to identify candidate predictors of MASLD progression Serum proteomic analysis was performed on 162 patients in the Phase II study of the Liverpool Monitor project, and 1139 proteins showed significant differences among different MASLD stages (Kruskal). Wallis rank-sum test, P<0.05. Differentially correlated serum proteins were enriched in acute inflammatory response, platelet α-granules, protein folding, and protein... Pathways such as lipid complexation, vitamin binding, and fatty acid metabolism show trends consistent with changes in the liver tissue proteome and transcriptome. Figure 1 In addition, a total of 207 proteins showed differential expression in both liver tissue and serum proteomes. Further screening identified 13 liver-enriched and secreted proteins as candidate biomarkers for MASLD progression, including: AFM, complement component 3 (C3), ceruloplasmin (CP), interstitial α-trypsin inhibitor heavy chain H1 (ITIH1), retinol-binding protein 4 (RBP4), and apolipoprotein C. III (APOC3), ORM1, histidine-rich glycoprotein (HRG), kininogen 1 (KNG1), serine protease inhibitor C1 (SERPINC1), vitrin (VTN), CFHR5, and APOA2.

[0039] Example 3: Screening of protein biomarker combinations for predicting MASLD histological staging based on machine learning Based on liver tissue transcriptional expression (p-value < 0.05 and FC ≥ 1.5), protein variation amplitude (p-value < 0.05 and FC ≥ 1.2), liver enrichment pathways, and liver secretory potential, 13 candidate proteins were optimized and identified from the transcriptomic and proteomic data of the phase I cohort. Figure 2 A). Through the optimal subset regression method (see section 10. Statistical Analysis in Example 1 for details), the four-protein combination consisting of AFM, CFHR5, APOA2, and ORM1 was finally determined to have strong predictive power for disease progression, and the differences were significant in each stage of MASLD progression. Figure 2 B, C).

[0040] Example 4: Validation of the predictive power of protein biomarker combinations for MASLD In the Phase I cohort of the LIVERMONITO project, the combined detection of four proteins—AFM, CFHR5, APOA2, and ORM1—in liver tissue showed the following results: Figure 3A) demonstrated excellent predictive power for MASL, MASH, and advanced liver fibrosis, with corresponding area under the receiver operating characteristic (AUROC) curves of 0.96 (95% confidence interval: 0.92–1.00), 0.95 (95% confidence interval: 0.91–0.99), and 0.87 (95% confidence interval: 0.81–0.93), respectively. Consistent with these results, the serum expression levels of these four proteins also showed stable predictive ability. In Phase II of the project, the AUROCs for diagnosing MASL, MASH, and advanced liver fibrosis were 0.87 (95% confidence interval: 0.80–0.93), 0.78 (95% confidence interval: 0.71–0.85), and 0.82 (95% confidence interval: 0.71–0.93), respectively. Figure 3 A).

[0041] External validation results showed that the combination of four protein biomarkers maintained good discriminative ability in two independent liver biopsy cohorts in Shanghai and Beijing (see Figure 3A): the AUROC for diagnosing MASL, MASH, and advanced liver fibrosis was as follows: Shanghai cohort 0.84 (95% confidence interval: 0.77~0.90), Beijing cohort 0.95 (95% confidence interval: 0.87~1.00); Shanghai cohort 0.75 (95% confidence interval: 0.66~0.84), Beijing cohort 0.92 (95% confidence interval: 0.85~1.00); Shanghai cohort 0.78 (95% confidence interval: 0.70~0.85), Beijing cohort 0.87 (95% confidence interval: 0.75~0.99). Furthermore, this study validated the results using a published Danish dataset. In this Danish liver biopsy cohort, the four proteins mentioned above were detected in the proteome of liver tissue and serum. The predicted AUROCs for MASL were 0.91 (95% confidence interval: 0.82–1.00) and 0.82 (95% confidence interval: 0.68–0.96), respectively; the predicted AUROCs for MASH were both 0.86 (95% confidence intervals: 0.75–0.98 and 0.73–0.98, respectively) (see [reference needed]). Figure 3 B).

[0042] Example 5: Model Construction and Validation for MASLD Staging Screening and Diagnosis Based on the four core candidate proteins (AFM, CFHR5, APOA2, and ORM1) and basic clinical indicators (BMI, gender, and presence of type 2 diabetes), a multivariate logistic regression model was used to construct a staging screening and diagnostic model for MASLD in the LIVERMONITOR phase II cohort. External validation was then performed in independent liver biopsy cohorts from Shanghai and Beijing. The specific methods are as follows: (1) Data collection and grouping: The liver biopsy cohort of the LIVERMONITOR project was selected as the modeling sample. All subjects were excluded from excessive drinking, viral hepatitis and other chronic liver diseases, and history of hepatotoxic drug use. Demographic information, physical indicators, biochemical indicators, liver biopsy pathology results and fasting serum samples of the subjects were collected. According to the SAF scoring system, the subjects were divided into non-MASLD, simple fatty liver MASL, early MASH, and fibrotic MASH groups. (2) Screening of protein biomarker combinations: Differentially expressed proteins among different MASLD stage groups were screened. Combined with liver tissue transcriptome and proteome data, common differentially expressed proteins with consistent expression trends and enrichment in the same metabolic pathways were screened. Based on the protein enrichment characteristics, secretion properties and correlation with liver pathology scores, 13 candidate proteins were initially screened. Using the optimal subset regression algorithm, AFM, CFHR5, APOA2 and ORM1 were determined as protein biomarker combinations. (3) Determine the model variable set: Combine the protein biomarkers with BMI, gender, type 2 diabetes status, and FIB4 index as model input variables; (4) Data standardization and variable assignment: Z-score standardization was performed on the expression data of the four protein biomarkers; binary classification 0-1 quantization was performed on the two categories of categorical variables, gender and type 2 diabetes status; (5) Construct a multi-factor logistic regression scoring model: Simple fatty liver MASL, early MASH, and fibrotic MASH are used as binary outcome variables respectively. The multi-factor stepwise logistic regression algorithm is used to fit the model. Based on the regression constant term and the regression coefficients of each variable, the scoring formulas for simple fatty liver MASL, early MASH, and fibrotic MASH are constructed respectively. (6) Setting diagnostic thresholds: Based on ROC curve analysis of the modeling cohort, high thresholds were set for simple fatty liver MASL and early MASH, and the diagnostic cutoff value for fibrotic MASH was determined; the high threshold was used for positive diagnosis of the disease.

[0043] (7) Model validation: The LIVERMONITOR project cohort was used for internal validation, and the Shanghai liver biopsy cohort, Beijing liver biopsy cohort, and Danish public liver biopsy dataset were used for external validation. The model performance was evaluated by AUROC, sensitivity, specificity, and positive predictive value, and the model construction was completed.

[0044] Before analysis, the expression data of the four protein biomarkers were standardized using Z-scores. The MASL and early MASH prediction scores for simple fatty liver can be calculated using the following formula: Simple fatty liver MASL score = -6.297 + 1.283 × AFM – 0.645 × APOA2 + 0.345 × CFHR5 + 0.643 × ORM1 + 0.294 × BMI + 1.667 × gender (male = 1, female = 0) + 1.271 × type 2 diabetes (yes = 1, no = 0) Early MASH score = -4.702 + 1.045 × AFM – 0.182 × APOA2 – 0.055 × CFHR5 – 0.203 × ORM1 + 0.183 × BMI – 0.179 × Gender (Male = 1, Female = 0) + 0.666 × Type 2 Diabetes (Yes = 1, No = 0) In the modeling cohort, the area under the curve (AUROC) of the predictive score for distinguishing histological MASL was 0.95 (95% confidence interval: 0.91–1.00), and the AUROC for distinguishing MASH was 0.83 (95% confidence interval: 0.74–0.86). External validation showed that the score demonstrated good to excellent discriminative power in two independent liver biopsy cohorts (AUROC: 0.82–0.92).

[0045] The study established cutoff values: high cutoff values ​​achieving 95% specificity for diagnosing MASL (cutoff value ≥ 2.80) and early MASH (cutoff value ≥ 2.19). In the validation cohort, the high cutoff values ​​showed specificity of 95%–100% for MASL and 100% for early MASH. Across all cohorts, the positive predictive value (PPV) for MASL corresponding to the high cutoff values ​​was 99%–100%, and the positive predictive value for early MASH was 89%–100%.

[0046] Significant liver fibrosis (≥F2 stage) is also an important criterion for patient enrollment in clinical trials. Therefore, this study simultaneously constructed a fibrosis MASH score as follows: Fibrosis MASH score = -5.951 + 0.540 × AFM – 0.422 × APOA2 + 0.284 × CFHR5 – 0.206 × ORM1 + 0.109 × BMI – 0.004 × Gender (Male = 1, Female = 0) + 1.216 × Type 2 Diabetes (Yes = 1, No = 0) + 1.058 × FIB4 The MASH score for fibrosis was superior to the FIB4 index in predicting ≥F2 stage liver fibrosis (AUROC: 0.81 vs 0.64, P < 0.001), and this was validated in the validation cohort, with AUROCs of 0.70 (95% confidence interval: 0.63–0.76) and 0.73 (95% confidence interval: 0.60–0.85), respectively. In all cohorts, a high cutoff value ≥0.48 identified 78%–83% of patients with significant liver fibrosis.

[0047] Example 6: Application of the model of the present invention in predicting long-term all-cause mortality. This study employed a two-step approach, combining proteomics data from the UK Biobank, to non-invasively predict the histological stage of metabolic-associated fatty liver disease (MASLD) in participants. Figure 4 A). Based on this four-protein prediction model, in the UK biobank cohort, a total of 13,652 MASL subjects (34.2%), 2,846 early MASH subjects (7.1%), and 1,944 fibrotic MASH subjects (4.9%) were identified. Figure 4 (B, 4C). The median follow-up of the 39,920 participants in this cohort was 13.9 years. The all-cause mortality rate was 787 per 100,000 people-years; among them, the mortality rate from cardiovascular disease was 160 per 100,000 people-years, the mortality rate from liver disease was 13 per 100,000 people-years, and the mortality rate from extrahepatic cancer was 323 per 100,000 people-years.

[0048] Compared with subjects without MASLD, those identified as having MASL and MASH using the four-protein model had a significantly increased risk of all-cause mortality: the adjusted hazard ratio (aHR) was 1.16 (95% confidence interval CI: 1.08–1.24). P <0.001), 1.52 (95%CI: 1.39~1.66, P <0.001); the risk of cardiovascular death was also higher in both groups, with aHR of 1.37 (95% CI: 1.17–1.61), respectively. P <0.001), 2.22 (95%CI: 1.84~2.68, P <0.001); the risk of death from extrahepatic cancer was also significantly increased, with an aHR of 1.20 (95% CI: 1.08–1.34). P =0.001), 1.27 (95%CI: 1.10~1.47, P =0.001). Notably, the risk of liver-related death in MASH patients was significantly higher than in those without MASLD (aHR=3.89, 95% CI: 2.25–6.73). Figure 4 D); among them, the risk of liver-related death in patients with fibrotic MASH was approximately 8 times that of patients with early-stage MASH (aHR=7.74, 95% CI: 2.64–22.69). Figure 4 E). Compared with early-stage MASH patients, fibrotic MASH patients also had a significantly increased risk of all-cause mortality and cardiovascular mortality, with corresponding aHRs of 1.61 (95% CI: 1.37–1.87). P<0.001), 2.13 (95%CI: 1.55~2.92, P <0.001).

[0049] Example 7: Development of Prediction Software Based on the Model Scoring Formula of the Invention Based on the model for MASLD staging screening and diagnosis constructed and validated in Example 5, this example further encapsulates the model's scoring formula into a predictive software tool that is easy for end users to operate. This software uses the MASL scoring formula for simple fatty liver, the MASH scoring formula for early stage, and the MASH scoring formula for fibrosis determined in Example 5 as its core algorithms. It receives model variables input by the user through a graphical user interface, and after automatic calculation in the background, outputs the predicted score and the corresponding MASLD histological grading results, thereby enabling convenient deployment and application of the model in clinical and research scenarios. Figure 5 ).

[0050] The software's operation process is as follows: In the data input area of ​​the software's graphical user interface, the user sequentially enters the expression values ​​of four protein markers (AFM, APOA2, CFHR5, and ORM1) after Z-score normalization. Figure 5 Alternatively, to further simplify the operation, a system interface for MASLD staging screening and diagnosis can be used, based on the AFM, APOA2, CFHR5, and ORM1 values ​​measured by the ELISA method. In the data input area of ​​the software's graphical user interface, users can sequentially enter the concentrations of the four proteins directly measured using an ELISA kit manufactured by RayBiotech. Figure 6 The user inputs BMI (kg / m²), gender (male / female), type 2 diabetes status (yes / no), and FIB4 index value. After inputting the data, the user clicks the "Predict" button. The software automatically calculates the scores for each dimension based on the three scoring formulas mentioned above, and performs MASLD histological grading based on the diagnostic cutoff values ​​determined in Example 5 (MASL high cutoff value ≥2.80; early MASH high cutoff value ≥2.19; fibrotic MASH diagnostic cutoff value ≥0.48). Finally, the software simultaneously outputs the MASH score, fibrotic MASH (fMASH) score, and the comprehensive MASLD grading result (non-MASLD, simple fatty liver MASL, early MASH, or fibrotic MASH) in the "Prediction Result" display area.

[0051] The software is a standalone executable program or a web application, supporting local deployment and remote access; the software interface supports switching between Chinese and English, and can quickly assist in the diagnosis and judgment of liver pathology grading in clinical scenarios.

[0052] The above embodiments are merely illustrative of the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made based on the essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A combination of protein biomarkers for staging screening and diagnosis of MASLD, characterized in that, The combination of protein biomarkers includes AFM, CFHR5, APOA2, and ORM1.

2. The use of the protein biomarker combination as described in claim 1 in the preparation of a kit for MASLD staging screening and diagnosis.

3. The application as described in claim 2, characterized in that, The test sample for this kit is serum.

4. A method for constructing a staging screening and diagnostic model for MASLD, comprising the following steps: (1) Data collection and grouping: The liver biopsy cohort was selected as the modeling sample. All subjects were excluded from excessive drinking, viral hepatitis and other chronic liver diseases, and history of hepatotoxic drug use. Demographic information, physical indicators, biochemical indicators, liver biopsy pathology results and fasting serum samples were collected from the subjects. According to the SAF scoring system, the subjects were divided into non-MASLD, simple fatty liver MASL, early MASH and fibrotic MASH groups. (2) Screening of protein biomarker combinations: Differentially expressed proteins among different MASLD stage groups were screened. Combined with liver tissue transcriptome and proteome data, common differentially expressed proteins with consistent expression trends and enrichment in the same metabolic pathway were screened. Based on the protein enrichment characteristics, secretion properties and correlation with liver pathology scores, 13 candidate proteins were initially screened. The protein biomarker combination as described in claim 1 was determined using the optimal subset regression algorithm. (3) Determine the model variable set: Combine the protein biomarkers and multiple indicators together as model input variables; (4) Data standardization and variable assignment: Z-score standardization was performed on the expression data of protein markers; binary classification 0-1 quantization was performed on the two categories of categorical variables, gender and type 2 diabetes status; (5) Construct a multi-factor logistic regression scoring model: Simple fatty liver MASL, early MASH, and fibrotic MASH are used as binary outcome variables respectively. The multi-factor stepwise logistic regression algorithm is used to fit the model. Based on the regression constant term and the regression coefficients of each variable, the scoring formulas for simple fatty liver MASL, early MASH, and fibrotic MASH are constructed respectively. (6) Setting diagnostic thresholds: Based on ROC curve analysis of the modeling cohort, high thresholds were set for simple fatty liver MASL and early MASH, and the diagnostic cutoff value for fibrotic MASH was determined; wherein the high thresholds were used for positive diagnosis of the disease. (7) Model validation: Conduct internal and external validation, evaluate model performance, and complete the construction of the model.

5. The construction method as described in claim 4, characterized in that, The 13 candidate proteins in step (2) include: AFM, C3, CP, ITIH1, RBP4, APOC3, ORM1, HRG, KNG1, SERPINC1, VTN, CFHR5, and APOA2.

6. The construction method as described in claim 4, characterized in that, The indicators in step (3) include BMI, gender, type 2 diabetes status and FIB4 index.

7. The construction method as described in claim 4, characterized in that, The MASL scoring formulas for simple fatty liver, early MASH scoring formula, and fibrotic MASH scoring formula in step (5) are as follows: Simple fatty liver MASL score = -6.297 + 1.283 × AFM – 0.645 × APOA2 + 0.345 × CFHR5 + 0.643 × ORM1 + 0.294 × BMI + 1.667 × gender + 1.271 × type 2 diabetes Early MASH score = -4.702 + 1.045 × AFM – 0.182 × APOA2 – 0.055 × CFHR5 – 0.203 × ORM1 + 0.183 × BMI – 0.179 × sex + 0.666 × type 2 diabetes Fibrosis MASH score = -5.951 + 0.540 × AFM – 0.422 × APOA2 + 0.284 × CFHR5 – 0.206 × ORM1 + 0.109 × BMI – 0.004 × gender + 1.216 × type 2 diabetes + 1.058 × FIB4.

8. The construction method as described in claim 7, characterized in that, The binary classification 0-1 quantization assignment for gender is male = 1, female = 0; the binary classification 0-1 quantization assignment for type 2 diabetes is yes = 1, no = 0.

9. The construction method as described in claim 4, characterized in that, In step (6), the high threshold for simple fatty liver MASL is ≥2.80, the high threshold for early MASH is ≥2.19, and the diagnostic cutoff value for fibrotic MASH is ≥0.

48.

10. A system for staging screening and diagnosis of MASLD, characterized in that, include: The data acquisition module is used to acquire the following data of the sample to be tested through the data input area of ​​the graphical user interface: expression level data of the protein biomarker combination of claim 1 after Z-score normalization, and clinical variable data, including BMI value, gender, type 2 diabetes status and FIB4 index value. The model calculation module is used to input the expression level data of the protein biomarker combination and the clinical variable data into the MASLD staging model, and automatically calculate the scores of each dimension according to the MASL scoring formula for simple fatty liver, the early MASH scoring formula and the fibrotic MASH scoring formula as described in claim 7. The grading output module is used to determine the MASLD histological grading based on the scores of each dimension and the predetermined critical threshold, and to display the MASLD grading results through the graphical user interface. The grading results include one of non-MASLD, simple fatty liver, early MASH, or fibrotic MASH.

Images