MASLD early screening marker based on plasma proteomics and application

Abstract

The invention provides an MASLD early screening marker based on plasma proteomics. The early screening marker comprises the following five proteins: CDHR2, FUOM, KRT18, ACY1 and GGT1. The early screening marker provided by the invention can provide ultra-early warning within 16 years before disease attack, is high in prediction accuracy, and is superior to the conventional model in short-term and long-term prediction aspects; meanwhile, a key proteome pathway is disclosed, and reference can be provided for clinical application and prevention strategies. The invention also provides application of the early screening marker based on the MASLD.

Description

Technical Field

[0001] This invention relates to a biomarker for early screening of MASLD based on plasma proteomics and its application. Background Technology

[0002] Metabolic dysfunction-related fatty liver disease (MASLD) is a type of metabolic fatty liver with a complex and unclear etiology, influenced by multiple factors. It often precedes serious chronic liver diseases such as liver fibrosis, cirrhosis, and hepatocellular carcinoma. Recently, the increase in diabetes, hyperlipidemia, and other chronic diseases has led to a dramatic rise in the global incidence of MASLD, projected to reach 25% by 2021. Therefore, MASLD and its related complications have become a major contributor to the global health burden. Due to the slow progression of MASLD, early prevention, accurate prediction, and timely intervention are crucial. Currently, there is no specific treatment for MASLD; treatment primarily focuses on lifestyle modifications such as weight control, increased physical activity, and dietary adjustments to reduce carbohydrate and saturated fat intake while increasing fiber and unsaturated fat intake. Therefore, identifying accurate clinical predictors of MASLD pathogenesis is essential.

[0003] However, existing studies on predictive biomarkers for MASLD incidence typically rely on traditional clinical indicators, including physical markers and common serological markers such as body mass index (BMI), waist circumference, and triglycerides (TG). These indicators are insufficient to achieve higher specificity and sensitivity, especially in long-term MASLD prediction. In contrast, high-throughput proteomics holds great promise for elucidating disease pathogenesis, identifying therapeutic targets, and discovering novel biomarkers. To date, few studies have extensively screened plasma proteins as biomarkers for early MASLD prediction, particularly in large cohort studies.

[0004] Olink proteomics employs Proximity Extended Assay (PEA) technology, which is extremely sensitive and can measure large amounts of protein from small samples, overcoming the limitations of previous technologies. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the main objective of this invention is to provide a MASLD early screening biomarker based on plasma proteomics and its application.

[0006] To achieve the above-mentioned main objectives, on the one hand, the present invention provides a MASLD early screening biomarker based on plasma proteomics, which includes the following five proteins: CDHR2, FUOM, KRT18, ACY1, and GGT1.

[0007] According to another specific embodiment of the present invention, the method for selecting early screening biomarkers includes the following steps:

[0008] A. Utilize prospective cohort study data from biological databases (e.g., the UK Biobank) to analyze a large number (e.g., tens of thousands) of adult individuals without MASLD at baseline and measure several (e.g., thousands) of plasma proteins.

[0009] B. Identify MASLD-related proteins using Cox regression analysis;

[0010] C. Use 5-fold cross-validation to perform machine learning to quantify the importance of the protein determined in step S2.

[0011] On the other hand, the present invention provides an application of the above-mentioned plasma proteomics-based MASLD early screening biomarker, which is used in the preparation of MASLD early screening kits.

[0012] The present invention has the following beneficial effects:

[0013] The early screening biomarkers of this invention can provide ultra-early warnings up to 16 years before the onset of disease, with high predictive accuracy, outperforming previous models in both short-term and long-term predictions; at the same time, they reveal key proteomic pathways, which can provide a reference for clinical applications and prevention strategies.

[0014] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0015] Figure 1A-Figure 1B This is a volcano plot showing the association between plasma proteins and new-onset MASLD, displaying the HR (x-axis) and -log10 (P-value) (y-axis) of 2737 proteins with new-onset Crohn's disease; Figure 1A and Figure 1B Results for Cox proportional hazards regression model 0 and model 1 are presented separately; model 0 adjusted for age, sex, and race; model 1 additionally adjusted for BMI, Townsend deprivation index, smoking status, frequency of alcohol consumption, physical activity, dietary habits, anxiety, depression, and history of antibiotic and NSAID use; p-values ​​are two-tailed tests, and proteins above the horizontal dashed line indicate a significant association with new-onset MASLD after Bonferroni correction;

[0016] Figure 1C This is a graph showing the enrichment of GO and KEGG pathways;

[0017] Figure 2 It is the cumulative AUC value of the top 30 proteins;

[0018] Figure 3A The accuracy of the selected protein in predicting MASLD events;

[0019] Figure 3BThe accuracy of the selected protein in predicting MASLD events over a five-year period;

[0020] Figure 3C The accuracy of the selected protein in predicting MASLD events over a decade;

[0021] Figure 3D The accuracy of the selected protein in predicting MASLD events over a decade;

[0022] Figure 4A It is a cumulative risk curve of protein FUOM;

[0023] Figure 4B This is a cumulative risk curve for protein ACY1;

[0024] Figure 4C This is a cumulative risk curve for the protein GGT1;

[0025] Figure 4D This is a cumulative risk curve for the protein CDHR2;

[0026] Figure 4E This is a cumulative risk curve for protein KRT18;

[0027] Figure 5A This is a graph showing the changes in protein FUOM concentration from 16 years ago until the diagnosis of MASLD;

[0028] Figure 5B This is a graph showing the changes in the concentration of protein ACY1 from 16 years ago until the diagnosis of MASLD;

[0029] Figure 5C This is a graph showing the changes in the concentration of protein GGT1 from 16 years ago until the diagnosis of MASLD;

[0030] Figure 5D This is a graph showing the changes in the concentration of the protein CDHR2 from 16 years ago until the diagnosis of MASLD;

[0031] Figure 5E This is a graph showing the changes in the concentration of protein KRT18 from 16 years ago until the diagnosis of MASLD;

[0032] Figure 6 This is a research overview diagram of Example 1. Detailed Implementation

[0033] Example 1

[0034] In this study, the inventors analyzed proteomics data from 52,952 UK Biobank participants who did not have baseline MASLD. Multivariate Cox regression was used to select proteins and build the model. The top-ranked proteins were combined with clinical data and a multigene risk score to develop and validate the predictive model. The cumulative incidence of MASLD at different baseline protein concentration quintiles was compared, and the time trend of protein levels was assessed. Normalized protein compression (NPX) values ​​of 2,737 plasma proteins were analyzed using the Olink 3072 Explore platform.

[0035] I. Methods

[0036] 1. Participants

[0037] The UK Biobank (UKB) dataset comprises over 500,000 participants (aged 37–73) recruited between 2006 and 2010 from England, Scotland, and Wales. Participants underwent a series of physical assessments and biostatistical examinations and completed questionnaires covering socioeconomic characteristics, lifestyle factors, and health-related conditions. Hospital admission records were regularly updated via links to the England Hospital Case Statistics, the Scotland Morbidity Records, and the Welsh Patient Case Database. Mortality data were obtained from the National Health Service (NHS) Digital System and the NHS Central Registry. The study received ethical approval from the North West Multi-Centre Research Ethics Committee, and informed consent was obtained from all participants. Individuals with a baseline diagnosis of metabolic dysfunction-associated fatty liver (MASLD) and those without proteomics data were excluded from the inventors' study (n = 449,317), resulting in a final cohort of 52,952 participants.

[0038] 2. Results

[0039] MASLD cases were reported by hospital specialist clinicians, general practitioners in the primary care system, or staff in the UK death registry. The inventors obtained different diagnostic codes recorded in primary or secondary locations in all participants' inpatient medical records according to the International Classification of Diseases, 10th Revision (ICD-10). In the primary analysis of this study, MASLD (including MASH) was determined using ICD-10 coding according to the latest expert panel consensus statement. Specifically, MASLD was classified under ICD-10 codes K76.0 (fatty [variable] liver, elsewhere unclassified) and K75.8 (other specific inflammatory liver diseases). The corresponding date for each diagnosis was first recorded in all inpatient medical records (field 41280). Results include MASLD events retrieved and compiled from hospital inpatient data summaries (field 41270). Follow-up begins on the date the subject goes to the UK Biobank Assessment Centre (field 53) (automatically obtained during the reception phase) and continues until the earliest recorded date of diagnosis, date of death (field 40000), or the last available date provided by the hospital or general practitioner (whichever comes first).

[0040] 3. Variables

[0041] At baseline, sociodemographic characteristics (age, sex, and race), socioeconomic status (household income, education level, and Thomson Poverty Index [TDI]), lifestyle factors (body mass index [BMI], frequency of alcohol consumption, smoking habits, physical activity, and dietary habits), and comorbidities (hypertension, diabetes) were collected.

[0042] 4. Definition of genetic risk

[0043] For detailed information on polygenic risk score (PRS) calculations, please refer to the supplementary materials.

[0044] 5. Missing values ​​and data preparation

[0045] In addressing the issue of missing data in predictors, the inventors quantified the number and percentage of missing values. Using chain equation techniques (with the MICE package in R), they performed multiple attribution and predictive mean matching. This approach integrates regression models and nearest neighbor matching to handle missing data. The inventors estimated values ​​on five datasets, iterating 50 times for each, and then merged the estimates according to Rubin's rule.

[0046] 6. Plasma proteomics

[0047] The serum proteomics data used by the inventors came from a large-scale UK Biobank Pharmaceutical Proteomics Project, a collaboration between Benjamin B. Sun of the University of Cambridge and 13 biopharmaceutical companies, conducted in 2023. This project employed antibody-based proximal extension assays (PEA) to measure, process, and analyze 2,941 plasma analytes from 54,219 UK Biobank participants. Details of plasma collection and testing are available in the supplementary materials.

[0048] 7. Statistical Analysis

[0049] Participants who developed MASLD during follow-up were assigned to the event MASLD group, while those who did not develop MASLD were assigned to the control group. To assess the difference between the MASLD event group and the control group, the inventors used a chi-square test for categorical variables and a Student's t-test for continuous variables, with statistical significance determined by the p-value. The inventors established a Cox proportional hazards regression model to reveal the association between the NPX values ​​(scales) of 2737 plasma proteins in UKB-PPP and MASLD events using HR values, 95% CIs, and bonferroni-corrected p-values. To identify the predictive proteins with stable performance, the inventors performed Cox regression analysis on various covariates. Based on the results of different covariate models, the inventors could obtain data from volcano plots (…). Figure 1A The stability of the best-performing protein was observed in the study. The inventors chose a more comprehensive model with covariates (age, sex, race, household income, education level, TDI, BMI, frequency of alcohol consumption, smoking habits, physical activity, dietary habits, hypertension, and diabetes) to ultimately select the predictive protein.

[0050] After Bonferroni correction, the inventors performed enrichment analysis on meaningful proteins (P < 0.05). Enrichment analysis helps to understand the extent to which plasma proteins are enriched in other functions, pathways, or specific biological processes. The inventors used the DAVID website, referencing the full suite of Olink proteins, to gather deeper biological insights. Statistical significance is expressed as a p-value from Fisher's exact test, followed by false discovery correction using the Benjamini-Hochberg method.

[0051] The proteome prediction model was built based on single protein concentration, multiple protein combinations, clinical factors (age, sex, race, education level, TD index, income, body mass index, hypertension, diabetes, healthy diet, smoking, alcohol consumption, activity), and the MASLD polygenic risk score (PRS). The inventors divided the UKB dataset into a training set (70%) and a test set (30%), using the 70% proportion for initial model development. Furthermore, replication validation analyses were performed using 75% and 80% of the training proportion to ensure the robustness of the results. The training set was used to build and train the Cox proportional hazards model, which was then evaluated on the test set. Finally, guided internal validation was conducted to further ensure the reliability of the results. The inventors evaluated the predictive performance of the model across different timeframes, spanning up to 16 years, including 5 years, 10 years, 16.6 years (hierarchical), and over 10 years. The results of internal validation were visually displayed using receiver operating characteristic (ROC) curves, and the area under the receiver operating characteristic (AUROC) was used to assess the predictive accuracy of proteins. The Delong test was performed using the R software package pROC to assess the statistical significance of the differences in AUC values ​​between different models.

[0052] Plasma proteins were classified according to the quintiles of the relevant protein NPX values, categorizing participants into five groups: Q1, Q2, Q3, Q4, and Q5. After adjusting for age, sex, race, household income, education level, TDI, BMI, frequency of alcohol consumption, smoking habits, physical activity, dietary habits, hypertension, and diabetes, Cox proportional hazards regression analysis was performed to understand the relationship between individual proteins and events (MASLD events). Cumulative incidence curves were then plotted to visually represent the clinical prognosis of MASLD events over time.

[0053] To compare the temporal trajectories of key protein concentrations before the onset of illness in the case and control groups, the inventors employed a nested case-control study to match individuals by age, sex, and ethnicity (control to case ratio of 5:1). Subsequently, the inventors modeled the mean protein concentration up to the date of onset using a locally weighted scatter plot smoothing (LOWESS) curve. The Mann-Kendall trend test was used to assess differences in slope, with p-values ​​derived from a two-tailed test.

[0054] II. Results

[0055] 1. Participant characteristics

[0056] The inventors' study included 52,952 patients without MASLD at baseline from the UK Biobank, with a median age of 58 years, of whom 46.1% were male. During a 16.6-year follow-up, 782 cases of MASLD were identified, with 406 occurring within 10 years, 114 within 5 years, and 376 after more than 10 years. The median age of MASLD patients (59 years) was not significantly different from that of healthy individuals (58 years). Men who participated less in physical activity appeared to be more likely to develop MASLD. In the MASLD group, 26.1% did not meet the World Health Organization (WHO) guidelines for physical activity, a higher proportion than in the healthy group (18.9%). Other covariates showing differences between groups included body mass index, education level, smoking status, alcohol consumption, annual household income, diabetes, and hypertension (P<0.001).

[0057] 2. Identify 30 proteins associated with MASLD.

[0058] In the inventors' research cohort, after adjusting for age, sex, race, education level, TD index, income, body mass index, hypertension, healthy diet, smoking, alcohol consumption, diabetes, and physical activity, 323 out of the 2737 proteins tested were significantly associated with the occurrence of MASLD, with Bonferroni-corrected p-values ​​less than 0.05. See details in [link to results]. Figure 1A Proteins were ranked according to their importance and added to the model in descending order. The DeLonghi test was used to compare each new model with the previous one to assess whether there were significant differences in their predictive power. Figure 2 The top 5 and top 30 proteins were selected for further analysis and used to construct a 5-protein / 30-protein prediction panel. Among all the proteins tested, aminoacylase-1 (… ACY1 The correlation between ) and MASLD events was strongest (HR = 2.28, P = 1.34 × 10⁻⁶). -81 The second highest concentration was fucoxonogenase (FUOM, HR = 2.27, P = 1.54 × 10⁻⁶). -70 ) and glutathione hydrolase 1 proenzyme (GGT1, HR = 2.34, P = 1.22 × 10⁻⁶) -69 Keratin type I cytoskeleton 18 (KRT18, HR = 1.68, P = 3.11 × 10⁻⁶) -68 ) and Cadherin-related family member 2 (CDHR2, HR = 2.21, P = 3.98 × 10) -62 Higher levels of these substances also increase the risk of developing MASLD (Supplementary Table 2).

[0059] 3. Biological pathway analysis

[0060] Subsequent enrichment analysis ( Figure 1B This study elucidated important biological pathways related to plasma proteins, particularly in the immune system, including inflammatory responses, immune responses, and IL-10 signaling.

[0061] 4. Predictive performance of plasma proteins

[0062] The inventors internally validated the predictive accuracy of the selected proteins for MASLD events using a guided method (Figure 3). They found that the top five proteins (ACY1, GGT1, FUOM, KRT18, CDHR2) exhibited good predictive performance for MASLD events (AUC > 0.7) in all time constraints, including those within 5 years, 10 years, and longer. Notably, among other time constraints, single proteins showed the best predictive performance for MASLD occurring within 5 years: plasma CDHR2 only (AUC = 0.825), plasma FUOM only (AUC = 0.815), plasma KRT18 only (AUC = 0.810), plasma GGT1 only (AUC = 0.803), and plasma ACY1 only (AUC = 0.797). Over a 16-year time span, combinations of 30 proteins demonstrated significant predictive accuracy for MASLD events across different timeframes: within 5 years (AUC = 0.887), within 10 years (AUC = 0.818), over 10 years (AUC = 0.764), and all timeframes (AUC = 0.785). When the number of proteins was reduced to a more clinically feasible 5-protein panel, the model still exhibited high predictive accuracy: within 5 years (AUC = 0.857), within 10 years (AUC = 0.775), over 10 years (AUC = 0.739), and all timeframes (AUC = 0.758). Combined with clinical predictive factors, predictive performance was further improved: 30 proteins within 5 years (AUC = 0.910), 30 proteins within 10 years (AUC = 0.851), 30 proteins over 10 years (AUC = 0.817), and 30 proteins across all time periods (AUC = 0.828); 5-protein panel within 5 years (AUC = 0.894), 5-protein panel within 10 years (AUC = 0.840), 5-protein panel over 10 years (AUC = 0.807), and 5-protein panel across all time frames (AUC = 0.822).

[0063] 5. Plasma proteins and the cumulative incidence of MASLD

[0064] Cumulative hazard curves were used to assess the risk of developing MASLD when baseline plasma protein levels were too high or too low (Figure 4). Baseline protein levels were divided into five equal parts, and then cumulative hazard curves were plotted. The hazard ratio assessed the risk of developing MASLD when protein levels were between Q1 and Q5 throughout the follow-up period. CDHR2 (HR = 9.81 [7.11, 13.54], P = 8.6 × 10⁻⁶) -44 ),FUOM(HR=9.45[6.73,13.27],P=2.2×10 -38 ), KRT18 (HR=9.02[6.62,12.29], P=3.25×10 -44 ), ACY1 (HR=8.69[6.48,11.65], P=2.73×10 -47 ) or GGT1 (HR=7.05[5.28,9.41],P=4.5×10 -40 Compared to the Q1 level, the risk of MASLD is increased.

[0065] 6. Replicate and verify the analysis results

[0066] The results of the replication analysis were consistent with those of the primary analysis. To verify robustness, the inventors randomly assigned participants to the training and test sets at a ratio of 75% and 80%, respectively. The predictive performance of single proteins for MASLD events remained strong: over 5 years, CDHR2 (75%: AUC = 0.812, 80%: 0.796), FUOM (75%: 0.802, 80%: 0.787), KRT18 (75%: 0.794, 80%: 0.781); over 10 years, ACY1 (75%: 0.742, 80%: 0.736), CDHR2 (75%: 0.741, 80%: 0.740), KRT18 (75%: 0.7... 38, 80%: 0.734); Over 10 years, ACY1 (75%: 0.732, 80%: 0.732), FUOM (75%: 0.705, 80%: 0.710), KRT18 (75%: 0.697, 80%: 0.699); All along, ACY1 (75%: 0.734, 80%: 0.734), FUOM (75%: 0.728, 80%: 0.729), KRT18 (75%: 0.724, 80%: 0.724). The combined model of the five proteins also showed good predictive accuracy: 5 years (75%: AUC = 0.842, 80%: 0.826); 10 years (75%: 0.771, 80%: 0.768); over 10 years (75%: 0.737, 80%: 0.738); and all time (75%: 0.755, 80%: 0.756). The 30-protein panel further demonstrated strong performance (AUC = 0.761–0.879).

[0067] 7. Preclinical trajectory of plasma proteins

[0068] The inventors created a graph depicting changes in protein concentrations from 16 years prior to MASLD diagnosis and compared them with those of individuals who did not have MASLD during the same period (Figure 5). The locally weighted scatter plot smoothed curves showed that the levels of these five proteins (ACY1 / FUOM / GGT1 / KRT18 / CDHR2) remained consistently elevated in patients diagnosed with MASLD. In the MASLD group, GGT1 levels (P = 1.25 × 10⁻⁶) were significantly higher. -4 (Mann-Kendall trend test) and ACY1 (P = 2.30 × 10⁻⁶) -3 The level of FUOM in the plasma increased faster over time than in individuals without MASLD (P = 2.14 × 10⁻⁶). -2 CDHR2 (P = 6.00 × 10) -3 KRT18 (P = 2.60 × 10) -3 This also shows a significant gap.

[0069] III. Discussion

[0070] In this study involving 52,952 participants, the inventors conducted a large-scale proteomic analysis to identify biomarkers predicting MASLD up to 16 years before onset. Of the 2,737 proteins analyzed, five emerged as outstanding predictors of future MASLD: CDHR2, FUOM, KRT18, ACY1, and GGT1. Using these five key proteins, the inventors developed simplified models that demonstrated superior performance across different timeframes, particularly in long-term prediction. These findings highlight the potential of these biomarkers and models to inform new strategies for the early detection and prevention of MASLD.

[0071] The top five proteins—CDHR2, FUOM, KRT18, ACY1, and GGT1—showed significant biases in plasma concentrations between patients with and without disease progression. Participants with higher baseline levels had a much higher risk of developing MASLD compared to those with lower concentrations. Furthermore, the study found these key proteins enriched in immune-related pathways, highlighting their potential importance in the disease process.

[0072] Of the five key proteins discovered by the inventors, two are newly identified. Previous studies have shown that ACY1, GGT1, and KRT18 are closely related to the pathogenesis of MASLD, possibly through mechanisms associated with amino acid metabolism, oxidative stress, and hepatocyte apoptosis. However, the predictive role of these three proteins in MASLD has not been explored before. The inventors' findings pioneer the use of these three proteins as future predictive biomarkers for MASLD. Notably, the inventors' research is the first to highlight the importance of these two novel proteins in the early development of MASLD and demonstrate their strong predictive potential. Specifically, CDHR2, known for its involvement in cell adhesion and tissue integrity, may promote the pathogenesis of MASLD by influencing cell-cell interactions. FUOM, an enzyme involved in glycoprotein synthesis, has the potential to disrupt normal lipid metabolism, a key factor in MASLD progression. Increased FUOM expression may lead to lipid accumulation and inflammation in the liver, further exacerbating disease development. These results suggest that both proteins may play important roles in metabolic and inflammatory processes, which are driving forces in the pathogenesis of MASLD.

[0073] Previous studies on predictive models for MASLD incidence have had some limitations, particularly in terms of predictive accuracy and time span. Most studies rely on traditional clinical indicators such as body mass index (BMI), waist circumference, triglycerides, and high-density lipoprotein (HDL), typically based on cross-sectional data. For example, Deng et al. used 33 longitudinal clinical variables and a Long Short-Term Memory (LSTM) algorithm to achieve a predictive performance of 0.818 approximately one year in advance. Similarly, Huang et al. used 11 physical and serological predictors with 6 machine learning methods to construct a 5-year MASLD predictive model, where logistic regression achieved AUROCs of 0.778 and 0.806 in internal and external validation, respectively. However, these models are relatively inaccurate and lack the ability to provide long-term predictions. Furthermore, the reliance on a relatively large number of indicators also limits their clinical applicability.

[0074] By selecting and integrating key proteins identified through association analysis, the inventors developed a proteomics-based model to predict the future incidence of MASLD. Individual proteins demonstrated significantly strong predictive power for MASLD, particularly in short-term predictions over the next five years. Plasma proteins such as CDHR2 (AUC = 0.825), FUOM (AUC = 0.815), KRT18 (AUC = 0.810), GGT1 (AUC = 0.803), and ACY1 (AUC = 0.797) showed extremely high accuracy when used alone, making this method both simple and clinically applicable. When combined into a more clinically suitable 5-protein panel, the predictive accuracy further improved, with an AUC of 0.857 for MASLD within 5 years, 0.775 within 10 years, 0.739 for more than 10 years, and 0.758 for all time scenarios. This 5-protein panel exhibited excellent long-term predictive ability, capable of predicting MASLD incidence up to 16 years into the future. Furthermore, incorporating clinical predictive factors into the model further improved the performance of the 5-protein panel, achieving an AUC of 0.894 for predicting MASLD within 5 years, 0.840 for predicting MASLD within 10 years, 0.807 for predicting MASLD over 10 years, and 0.822 for predicting MASLD across all time scenarios. These results highlight the inventors' model's superior performance in 5-year, 10-year, and long-term predictions, making it highly valuable for early clinical intervention and patient management. Its simplicity and powerful performance distinguish it from existing models, particularly in its ability to predict the risk of MASLD in the long term.

[0075] The main strength of the inventors' research lies in the long-term tracking of a large community sample and the use of high-throughput proteomics analysis to discover novel plasma biomarkers and develop innovative proteomics-based predictive models. The researchers discovered several important protein biomarkers with high accuracy in predicting MASLD events 16 years prior to onset, which will be significant for screening and early intervention in high-risk populations. However, some limitations should be considered when interpreting these results. First, while Olink proteomics provides a comprehensive assessment of circulating proteins, it does not cover the entire human proteome and may be subject to selection bias in measuring secreted proteins. Second, the UK Biobank is predominantly composed of Caucasian Europeans, which limits the generalization of the findings to other populations. While validation on independent external datasets is ideal, there are currently no large-scale prospective cohorts using Olink detection methods for long-term tracking. To address this challenge, the inventors conducted internal validation and replication analyses through 1000 iterations, randomly assigning participants to training and test sets with varying proportions. The consistent results obtained from these replication methods further strengthen the robustness of the inventors' findings.

[0076] Using a data-driven proteomics approach, the inventors identified key plasma biomarkers from the largest long-term prospective community cohort to date. Their research revealed a strong correlation between plasma levels of FUOM, ACY1, GGT1, CDHR2, and KRT18 and the onset of MASLD, indicating that these levels begin to change more than 16 years before diagnosis and show significant differences among individuals at different risk levels. These biomarkers, whether used alone or in combination with clinical factors, have demonstrated the strongest predictive performance to date in both short-term and long-term risk assessment. Integrating these biomarkers can predict MASLD up to 16 years before diagnosis, thereby significantly enhancing early screening and intervention strategies for high-risk populations.

[0077] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the scope of the invention. Any person skilled in the art can make modifications without departing from the scope of the invention; all equivalent modifications made in accordance with the invention should be covered by the scope of the invention.

Claims

1. A MASLD early screening biomarker based on plasma proteomics, characterized in that, The early screening biomarkers include the following five proteins: CDHR2, FUOM, KRT18, ACY1, and GGT1.

2. The MASLD early screening biomarker as described in claim 1, characterized in that, The method for selecting early screening biomarkers includes the following steps: A. Using prospective cohort study data from the biobank, we analyzed a large number of adult individuals without MASLD at baseline and measured several plasma proteins. B. Identify MASLD-related proteins using Cox regression analysis; C. Use 5-fold cross-validation to perform machine learning to quantify the importance of the protein determined in step S2.

3. An application of the MASLD early screening biomarker based on plasma proteomics as described in claim 1, characterized in that, The early screening biomarkers are used in the preparation of the MASLD early screening kit.

Images