Skip to main content
NCBI home page
Hepatology Forum logoLink to Hepatology Forum
. 2026 May 11;7(2):125–132. doi: 10.14744/hf.2025.84568

Plasma PDGFRβ and PSD4 methylation levels for non-invasive staging of liver fibrosis

Burge Ulukan 1,†, Hong Yang 2, Hande Uludag 1, Cheng Zhang 2, Ali Mutlu 3, Murat Dayangac 4, Murat Akyildiz 1, Buket Yigit 1, Shaghayegh Soleimani 1, Hale Kirimlioglu 5, Burcu Saka 6, Cigdem Ataizi Celikel 7, Fatih Eren 8,10, Adil Mardinoglu 2,11, Yusuf Yilmaz 9, Mujdat Zeybel 1,✉
PMCID: PMC13442723  PMID: 42564411

Abstract

Background and Aim

While liver biopsy remains the reference standard for assessing hepatic fibrosis, the major prognostic factor in metabolic dysfunction-associated steatotic liver disease (MASLD), its inherent limitations have driven the search for innovative non-invasive diagnostic tests. In this study, we sought to evaluate the diagnostic accuracy of plasma PDGFRβ and PSD4 methylation levels, integrated within machine learning algorithms, for non-invasive staging of hepatic fibrosis in patients with MASLD, and to compare its performance against established non-invasive biomarkers.

Materials and Methods

Patients with biopsy-proven MASLD and healthy controls were recruited from the three institutions. Quantitative methylation of circulating cell-free DNA was assessed using bisulfite modification and pyrosequencing. The resulting data were used to develop linear discriminant analysis, random forest, and support vector machine algorithms to identify patients at different stages of fibrosis.

Results

The study included 234 patients with histologically confirmed MASLD and 43 healthy controls. Each dataset was validated using an independent cohort. Advanced fibrosis was associated with elevated plasma PDGFRβ and PSD4 methylation levels in both the discovery and validation cohorts. The integration of plasma DNA methylation markers with clinical parameters demonstrated performance comparable to that of existing noninvasive biomarkers for detecting both significant and advanced fibrosis, achieving area under the curve scores exceeding 0.75 across all models.

Conclusion

Supervised machine learning algorithms incorporating plasma PDGFRβ and PSD4 DNA methylation levels show promise as a non-invasive approach for assessing hepatic fibrosis in MASLD.

Keywords: Cell-free DNA, DNA methylation, liver fibrosis, machine learning algorithms, MASLD, predictive models

Highlights & Insights

  • Scientific Gap: Liver biopsy remains an imperfect gold standard for fibrosis assessment due to sampling errors and inter-observer variability. Although non-invasive methods have improved evaluation, they cannot fully replace biopsy; accurate, timely fibrosis detection in MASLD remains needed.

  • Key Finding: Machine learning using cell-free epigenetic markers and clinical data identified significant differences in plasma PDGFRβ and PSD4 methylation by MASLD status and fibrosis stage

  • Clinical Impact: These methylation signatures may offer a promising non-invasive approach for assessing fibrosis in MASLD.

Introduction

Metabolic dysfunction-associated steatotic liver disease (MASLD), is a complex and prevalent clinical condition, affecting over 30% of the global population.[1] It is closely associated with obesity, type 2 diabetes, dyslipidemia and metabolic syndrome, which collectively contribute to hepatic injury and increase the risk of severe hepatic complications, including advanced fibrosis, cirrhosis and hepatocellular carcinoma (HCC).[2,3] Notably, MASLD has emerged as the leading cause of chronic liver disease worldwide and is a significant contributor to liver-related morbidity and mortality.

Until recently, no pharmacological treatment had received regulatory approval for MASLD. However, the U.S. Food and Drug Administration (FDA) has recently granted accelerated approval to resmetirom for the treatment of Metabolic dysfunction-associated steatohepatitis (MASH) in patients with moderate-to-advanced liver fibrosis (stages F2 and F3). Notably, resmetirom, an oral thyroid hormone receptor-β -agonist, has demonstrated efficacy in reducing hepatic fibrosis and improving key histopathological features of MASH, including hepatic fat content and inflammation.[4]

Fibrosis represents a critical prognostic determinant in chronic liver diseases of various etiologies. The early detection and accurate staging of hepatic fibrosis are essential to prevent the progression to more advanced stages, such as decompensated cirrhosis.[5] While liver biopsy remains the reference standard for fibrosis assessment, its invasive nature, susceptibility to sampling errors, and inter-observer variability highlight the need for alternative diagnostic approaches.[3,6–8] Non-invasive techniques (NIT), including serum biomarkers, routine laboratory tests, scoring systems (e.g., FIB-4 index, the NAFLD fibrosis score), and imaging-based methods such as vibration-controlled Transient Elastography (VCTE; FibroScan®, Echosens, Paris, France), have emerged as valuable and effective tools for excluding advanced fibrosis. Among these, VCTE is particularly effective for assessing liver stiffness and staging fibrosis in patients with MASLD. Nevertheless, limitations such as higher failure rates in obese individuals and practical challenges related to its application and stability restrict its broader feasibility.[9] Despite their utility, NITs cannot fully replace liver biopsy in clinical practice.[9,10]

The mechanisms driving the development and progression of MASLD are intricate and remain incompletely understood. Multiple factors, including dietary habits, lifestyle, genetic predisposition, and epigenetic modifications, have been implicated in the pathophysiology of MASLD.[11,12] Among these, DNA methylation, a key epigenetic modification, predominantly occurs at cytosine-guanine dinucleotides, known as CpG sites, and is closely linked to alterations in gene expression. DNA methylation is mediated by DNA methyltransferases (DNMTs), which catalyze the addition of a methyl group to the fifth carbon of the cytosine ring, forming 5-methylcytosine. It plays a significant role in the pathogenesis of metabolic disorders including obesity, diabetes, and MASLD.[13] Extensive research has demonstrated the association between DNA methylation patterns and liver diseases.[3,14–17] For example, Gerhard et al.[14] identified the correlation between DNA methylation levels in patients with MASLD and key signaling pathways. Similarly, Xu et al.[16] identified ten methylation markers for diagnosing HCC. Furthermore, our previous studies revealed significantly higher DNA methylation levels within the PPARγ and PPARα promoters, two genes with anti-fibrogenic effects, in patients with severe MASLD.[17,18]

Predictive models are widely employed in the diagnosis and prognosis of liver diseases. For instance, Batista et al.[19] applied Linear Discriminant Analysis (LDA) to develop a model for predicting fibrosis stages. Similarly, recent research has highlighted the effectiveness of the support vector machine (SVM) model in predicting fibrosis risk based on the patients’ radiomics data.[20] Building on the significance of predictive modeling, we evaluated whether methylation levels of circulating cell-free plasma DNA at specific CpG sites could serve as biomarkers to stratify fibrosis stages in MASLD. Two genes, PDGFRβ and Pleckstrin and Sec7 Domain Containing 4 (PSD4), were selected due to their potential association with liver disease or fibrosis biology.[16,21] Subsequently, three predictive models were developed and compared to identify the most effective approach for fibrosis stage stratification. These findings are anticipated to contribute to the development of novel NITs for liver fibrosis, offering valuable insights for future diagnostic advancements.

Material and Methods

Study Cohorts

The study was conducted on MASLD patients (Discovery Cohort n=115, Validation Cohort n=119) and healthy individuals (Discovery Cohort n=21, Validation Cohort n=22). Blood samples and clinical data were collected after obtaining informed consent from participants in compliance with ethical standards and approved by Koc University Research Ethics Committees [approval number: 2015.053.IRB1.014 (30.03.2015), 2016.024.IRB2.005 (15.02.2016), 2017.139.IRB2.048 (10.09.2017), 2022.246.IRB2.040 (30.07.2022), 09.2018.086 (24.09.2018)]. The study adhered to the ethical principles outlined in the Declaration of Helsinki. The diagnosis of MASLD was established based on the guidelines of American Association for the Study of Liver Diseases.[22] The fibrosis severity in the study participants was assessed using histopathological evaluation.

DNA Methylation and Pyrosequencing Analysis

200 μL of plasma was used for DNA extraction with the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany). The DNA was then treated with bisulfite using the EZ DNA Methylation Gold kit (Zymo Research, Irvine, CA, US) according to the manufacturer’s instructions. The bisulfite-modified DNA was amplified using custom-designed primers. The primers were designed using PyroMark Assay Design Software 2.0, and their sequences are provided in Supplementary Table 1. Amplification was performed using the PyroMark-PCR kit (Qiagen) according to conditions in Supplementary Table 2. After thermal cycling, all samples were loaded onto a 1.5% agarose gel and run at 100 volts for 30 minutes. Samples were processed for the pyrosequencing experiment only after their band intensities shown in Supplementary Figure 1 had been checked. Sample preparation and pyrosequencing were subsequently carried out on the PyroMark Q96 ID instrument (Qiagen). Unmethylated and methylated EpiTect DNA controls (Qiagen) and negative PCR templates were used for each experiment. The average methylation percentages were analyzed using PyroMark software (version 2.5.8; Supplementary Fig. 2–6).

Predictive Models for Staging Fibrosis and Statistical Analysis

The discovery dataset was used to develop three binary classification models aimed at distinguishing between different stages of fibrosis. The first model was designed to differentiate patients with significant fibrosis (F≥2) from healthy subjects and those with mild or no fibrosis (F≤1). The second model focused on identifying patients with advanced fibrosis (F≥3), while the third model aimed to differentiate patients with cirrhosis (F=4). To construct these models, three machine-learning algorithms were employed. Linear Discriminant Analysis was implemented using the LDA () function from the MASS package (version 7.3-60.0.1). The Random Forest algorithm was implemented using the random Forest () function in the randomForest R package (version 4.7-1.1). Finally, the SVM algorithm was implemented using the SVM () function from the e1071 package (version 1.7-14). These algorithms were selected to assess their accuracy in classifying fibrosis stages using the provided data. Initially, the discovery dataset, consisting of 85 patients with MASLD and 21 healthy controls, was divided into a training set (70%) and a testing set (30%) using the create Data Partition () function from the caret R package (v 6.0-94). Hyperparameter optimization was performed through 5-fold cross-validation on the training set to maximize the area under the receiver operating characteristic (AUC ROC) curve. Once the models were trained and stabilized, the predict () function in caret was employed to estimate the class probabilities for the samples in the test set. The trained models were then independently validated on an external dataset comprising 103 patients with MASLD and 22 healthy controls. Samples with missing feature data were excluded from the model training and prediction procedures. The pROC R package (v 1.18.5) was used to visualize the ROC curve. Performance metrics, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), precision, recall, and F1 scores, were calculated and presented in Supplementary Tables 3 and 4.

All statistical analyses were performed using GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA, USA). For comparisons between cohorts, descriptive statistics were used. The Shapiro-Wilk test was used to check the normality of continuous variables. Normally distributed variables were analyzed using the Student’s t-test and presented as “mean±standard deviation” whereas non-normally distributed variables were analyzed using the Mann-Whitney U test and presented as median (interquartile range). Categorical variables were analyzed with Fisher’s exact test and presented as frequencies (percentages). Two-tailed p-values <0.05 were considered statistically significant.

Results

The study cohort included 43 healthy subjects and 234 patients diagnosed with MASLD. The histological assessments and Fibroscan evaluations revealed that 115 patients exhibited mild fibrosis (stages F0-F2), whereas 119 had advanced fibrosis (stages F3-F4). The general characteristics of study participants are summarized in Table 1.

Table 1.

Demographic and clinical characteristics of patients

Discovery cohort (n=136) Validation cohort (n=141)
Healthy (n=21) MASLD (n=115) Healthy (n=22) MASLD (n=119)
Male, n (%) 11 (52.3) 69 (60.0) 7 (31.8) 57 (47.9)
Female, n (%) 10 (47.6) 46 (40.0) 15 (68.2) 62 (52.1)
Age (years)   37 (24, 42) 55 (47, 61) 35 (25, 40) 53 (46, 59)
Diabetes, n (%) 0 69 (60.0) 0 85 (71.4)
BMI (kg/m2) 23.9±2.1 29.5 (26.7, 34.2)**** 23.86±2.4 33.3 (29.7, 36.6)
Albumin (g/dL) 4.82±0.1** 4.5 (4, 4.8) 4.65±0.1 4.4 (4.1, 4.7)
ALT (IU/L), 22.0±10.0 40 (25.5, 69.5)* 19.93±7.3 49 (31.0, 98.0)
AST (IU/L) 21.68±7.6 32.7 (25.0, 47.0)** 19.10±5.3 41 (29.0, 62.0)
Platelet (x109), 258.5±56.4 213.5 (133.0, 285.0) 240.0±46.4 208.0 (148.0, 248.0)
FIB-4 Index 0.6±0.2 1.2 (0.7, 3.2)* 0.6±0.2 1.4 (1.0, 2.3)
Fibrosis stage, n (%)
   0 – 16 (13.91) – 0 (0.00)
   1 – 23 (20.00) – 25 (21.00)
   2 – 16 (13.91) – 35 (29.41)
   3 – 27 (23.48) – 30 (25.21)
   4 – 33 (28.70) – 29 (24.37)
NAS score – 3.0 (2.0, 4.0)**** – 6.0 (5.0, 7.0)
Open in a new tab

Categorical variables are shown as frequencies (percentages). Normally distributed data are presented as “mean±standard deviation” and non-normally distributed data are presented as “median (interquartile range)”. Comparisons between cohorts of each subgroup were made using the Student’s t-test for continuous and Fisher’s exact test for categorical variables. Statistically significant variables are annotated on the discovery cohort and p-values are given as p<0.05 (*), 0.01 (**), 0.0001 (****).

ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; BMI: Body mass index; MASLD: Metabolic dysfunction-associated steatotic liver disease; NAS: NAFLD Activity score.

Higher plasma PDGFRβ and PSD4 DNA methylation are associated with advanced hepatic fibrosis.

We investigated DNA methylation profiles at specific CpG sites within PSD4, a locus recently implicated in HCC, and PDGFRβ, a receptor for PDGF strongly expressed in hepatic stellate cells and activated myofibroblasts.[16,21] Cell-free DNA was extracted from blood samples, and pyrosequencing assays were optimized for targeted CpG sites (Supplementary Table 1). Plasma DNA methylation levels were quantified at five CpG loci within the PDGFRβ promoter and three CpG loci within the PSD4 promoter (Fig. 1, 2). The discovery cohort consisted of patients recruited from two centers. In this study, all CpG sites within PDGFRβ exhibited higher DNA methylation levels in patients with MASLD than in healthy controls (Fig. 1a-e). Notably, methylation levels at PDGFRβ CpGs 5 and 6 were significantly higher in patients with MASLD and advanced fibrosis (F3-F4) compared with those with mild fibrosis (F0-F2) and healthy controls (Fig. 1a, b). Other CpG sites (CpG 7–9) showed higher methylation levels in MASLD groups than in healthy controls (Fig. 1c–e). These findings were validated in an independent patient cohort recruited from the third center (Fig. 2). The elevated methylation levels at PDGFRβ CpG sites 5 and 6 in F3-4 MASLD remained significant but were less pronounced in the validation cohort (Fig. 2a, b).

Figure 1.

Figure 1

Open in a new tab

Plasma DNA methylation level of MASLD patients in the discovery cohort. Plots were shown as the correlation between the CpG methylation and fibrosis stage of the MASLD. The analysis examines the plasma DNA methylation levels across fibrosis stages F0-2 and F3-4, alongside those of healthy subjects. Bisulfite modification and pyrosequencing were used to assess the plasma CpG methylation at five loci in PDGFRβ and three loci in the PSD4 gene. Changes in PDGFRβ (a–e) and PSD4 (f–h) CpG methylation levels are associated with the fibrosis stage in the discovery cohort. DNA methylation is quantified and represented as a percentage.

*p<0.05, **p<0.01 and ***p<0.001.

Figure 2.

Figure 2

Open in a new tab

Plasma DNA methylation level of MASLD patients in the validation cohort. Plots were shown as the correlation between the CpG methylation and fibrosis stage of the MASLD. The analysis examines the plasma DNA methylation levels across fibrosis stages F0-2 and F3-4, alongside those of healthy subjects in the validation cohort. Bisulfite modification and pyrosequencing were used to assess the plasma CpG methylation at five loci in PDGFRβ and three loci in the PSD4 gene. Changes in PDGFRβ (a–e) and PSD4 (f–h) CpG methylation levels are associated with the fibrosis stage in the validation cohort. DNA methylation is quantified and represented as a percentage.

*p<0.05, **p<0.01 and ***p<0.001.

For the PSD4 promoter, plasma DNA methylation levels at all targeted CpG loci were considerably higher in patients with MASLD than in healthy controls in both the discovery and validation cohorts (p<0.001) (Fig. 1f-h, Fig. 2f-h). Remarkably, patients with advanced fibrosis exhibited higher methylation levels than those with mild fibrosis in both cohorts.

Machine Learning Algorithms for Accurate Staging of Liver Fibrosis

To identify fibrosis stages, we developed predictive models integrating plasma DNA methylation data and clinical parameters. Four diagnostic approaches were assessed across four distinct datasets: the FIB-4 index, clinical data alone, DNA methylation levels alone, and a combination of clinical data and DNA methylation levels. Patients were stratified into four fibrosis categories for comparison: mild/no fibrosis (F0-1), significant fibrosis (≥F2), advanced fibrosis (≥F3), and cirrhosis (F4). To benchmark the models’ performance, we conducted AUC analyses using three supervised machine learning approaches: (i) lda(), (ii) randomForest(), (iii) svm().

Among the three supervised machine learning algorithms tested, SVM achieved the best diagnostic performance for detecting significant fibrosis, advanced fibrosis, and cirrhosis (Fig. 3). For identifying or excluding significant fibrosis, algorithms based on DNA methylation with or without clinical parameters performed comparably to the FIB-4 index, achieving AUC values of 0.84, 0.87, and 0.81 in the discovery cohort, and 0.63, 0.81, and 0.76 in the validation cohort, respectively (Fig. 3a, b). In detecting advanced fibrosis, SVM-based models outperformed the FIB-4 index, with higher positive predictive values of 0.87 and 0.74 compared with 0.66. For exclusion of advanced fibrosis, all three methods demonstrated excellent negative predictive values (Supplementary Tables 3 and 4). Additionally, models based on plasma DNA methylation levels of PSD4 and PDGFRβ showed excellent diagnostic accuracy for detecting and ruling out stage 4 fibrosis (cirrhosis) (Supplementary Tables 3 and 4).

Figure 3.

Figure 3

Open in a new tab

SVM analysis model in MASLD patients. This figure shows the SVM analysis in patients with different stages of fibrosis. (a–b) ROC curves of four markers (FIB-4, Clinical parameters, DNA Methylation, and Clinical Parameters + DNA methylation combination) in the discovery and validation cohort of patients with significant fibrosis. (c–d) ROC curves of four markers (FIB-4, Clinical parameters, DNA Methylation, and Clinical Parameters + DNA methylation combination) in the discovery and validation cohort of patients with advanced fibrosis. (e–f) ROC curves of four markers (FIB-4, Clinical parameters, DNA Methylation, and Clinical Parameters + DNA methylation combination) in cirrhotic patients’ discovery and validation cohort.

Overall, combining clinical data with DNA methylation levels yielded promising results for diagnosing or ruling out various stages of fibrosis, with AUC values consistently exceeding 0.75. Detailed comparisons of cohort performance across predictive models are presented in Figure 3 and Supplementary Figures 7 and 8; sensitivity and specificity metrics are provided in Supplementary Tables 3 and 4.

Discussion

Early and accurate identification of hepatic fibrosis is crucial for determining the clinical prognosis of chronic liver diseases and plays a vital role in preventing and treating advanced liver diseases.[5] Significant efforts have been directed towards developing novel NITs, including blood-based biomarkers, for fibrosis assessment.[9] In this study, we explored the potential of cell-free DNA methylation signatures as biomarkers for liver fibrosis in MASLD. Specifically, we analyzed DNA methylation levels in two candidate genes (PDGFRβ and PSD4) and identified significant differences across distinct fibrosis stages (significant fibrosis, advanced fibrosis, and cirrhosis).

PDGFRβ is the receptor for PDGF-B, a well-established pro-fibrotic growth factor predominantly produced by Kupffer cells, pericytes, and hepatic stellate cells. PDGF has a pivotal role in the extracellular matrix remodeling, particularly collagen synthesis and is implicated in various fibroproliferative disorders.[23] In this study, a significant increase in PDGFRβ methylation levels at 5 CpG sites was observed in patients with mild or advanced fibrosis compared with healthy controls in the discovery cohort. Notably, CpG 5-6 methylation levels of PDGFRβ showed marked differences between MASLD F0-2 and MASLD F3-4 patients in the discovery data; in the validation cohort, the difference was modest but statistically significant. Previous research by Lambrecht et al.[24] indicated that extracellular vesicles derived from activated murine and human hepatic stellate cells are enriched for PDGFRβ expression. Furthermore, circulating PDGFRβ protein levels, combined with a PDGFR β-based (PRTA) score, were identified as biomarkers of hepatic fibrosis, achieving AUROC values of 0.78 for significant fibrosis, 0.74 for advanced fibrosis, and 0.79 for cirrhosis across three different etiologies. In our study, apoptotic or necrotic myofibroblasts might have contributed to the circulating cell-free DNA methylation pool, thereby leading to an increase in PDGFRβ methylation in association with hepatic fibrosis.

PSD4 was previously identified as a prominent circulating cell-free methylation marker for HCC in a large cohort study.[16] Additionally, prior research has demonstrated that TNF-α-induced phospho-p65 activation leads to the recruitment of DNMT1, resulting in hypermethylation of the PSD4 promoter. This epigenetic modification is associated with carcinogenesis, particularly in individuals with excessive alcohol consumption.[25] Based on these findings, we investigated whether plasma methylation levels of the PSD4 gene are associated with fibrosis stages in MASLD. Our analysis revealed that changes in cell-free DNA methylation within the PSD4 promoter region were remarkably consistent across cohorts and notably higher in MASLD patients than in healthy controls (Fig. 1f-h, Fig. 2f-h). These results suggest that PSD4 methylation levels could serve as reliable predictors not only for hepatic fibrosis but also for HCC.

We next evaluated whether DNA methylation markers outperformed other biomarkers, including the FIB-4 index. To this end, we applied three widely used supervised machine learning algorithms: LDA, Random Forest, and SVM.[20,26,27] While differences were observed between cohorts for each marker, the AUROC scores consistently favored the combination of DNA methylation and clinical data in SVM analyses. The observed discrepancies between cohorts may be attributed to differences in sample size and fibrosis severity among patients. Notably, SVM results demonstrated that combining DNA methylation markers with clinical data yielded strong diagnostic performance, as reflected by AUROC scores and NPVs.[28–30] Importantly, even DNA methylation-based algorithms without incorporating clinical parameters showed robust performance in ruling out fibrosis. These algorithms achieved an NPV of 0.83 for significant fibrosis and an NPV of 1 for advanced fibrosis, highlighting their potential utility in primary care settings for stratifying MASLD patients and identifying those who require referral to specialized centers.

There are several limitations to this study. First, the relatively modest sample size may constrain the generalizability of our findings. Second, although multicenter, the data collection was limited to three medical centers within a single country, potentially introducing geographical and ethnic biases that may affect the generalizability of our results. Another significant limitation concerns participant selection: study included only a subset of metabolic profiles, which may not fully represent the diverse spectrum of metabolic manifestations observed in clinical practice. From an analytical perspective, although pyrosequencing provides highly accurate methylation analysis, its labor-intensive nature may present challenges for widespread clinical implementation. This methodological constraint could affect the immediate translation of our findings into routine clinical practice, particularly in resource-limited settings.

Conclusion

In conclusion, we explored the potential of machine learning algorithms as a non-invasive diagnostic tools for assessing fibrosis in patients with MASLD. Our findings demonstrate that methylation levels in circulating cell-free DNA are significantly altered in this patient group, aligning with prior research.[3] Collectively, our findings underscore the potential of DNA methylation as a biomarker for fibrosis staging.

Acknowledgement

The authors acknowledge the use of the facilities at the Koc University Research Center for Translational Medicine (KUTTAM), funded by the Presidency of Turkiye, Head of Strategy and Budget.

Footnotes

How to cite this article: Ulukan B, Yang H, Uludag H, Zhang C, Mutlu A, Dayangac M, et al. Plasma PDGFRβ and PSD4 methylation levels for non-invasive staging of liver fibrosis. Hepatology Forum 2026; 7(2):125–132.

Ethics Committee Approval

Blood samples and clinical data were collected after obtaining informed consent from participants in compliance with ethical standards and approved by Koc University Research Ethics Committees [approval number: 2015.053.IRB1.014 (30.03.2015), 2016.024.IRB2.005 (15.02.2016), 2017.139.IRB2.048 (10.09.2017), 2022.246.IRB2.040 (30.07.2022), 09.2018.086 (24.09.2018)].

Informed Consent

Blood samples and clinical data were collected after obtaining informed consent from participants in compliance with ethical standards.

Conflict of Interest

Yusuf Yilmaz has received honoraria for lectures from Echosens and consulting fees from Zydus, Akero, and Novo Nordisk. Adil Mardinoglu is the founder of SZA Longevity, Trustlife Therapeutics, ScandiBio Therapeutics, and ScandiEdge Therapeutics. All other authors have no conflicts of interest to declare.

Financial Disclosure

This work was supported by grants from the European Commission, Horizon 2020 Marie Curie Sklodowska Individual Fellowship (to MZ) and Newton-Katip Celebi Fund 2551 (to MZ).

Use of AI for Writing Assistance

The authors declare that no artificial intelligence was used in this study.

Author Contributions

Concept: MZ, YY; Design: MZ, YY; Supervision: YY; Data Collection and/or Processing: MZ, MD, MA, FE, YY; Analysis and/or Interpretation: BU, BY, SS, AM, HY, CZ, BS, HK, CAC, AMa; Literature Review: BU, HU; Writing: BU, HU, MZ; Critical Review: BU, HY, HU, CZ, AM, MD, MA, BY, SS, HK, BS, CAC, FE, AMa, YY, MZ.

Peer-review

Externally peer-reviewed.

References

  • 1.Teng ML, Ng CH, Huang DQ, Chan KE, Tan DJ, Lim WHY, et al. Global incidence and prevalence of nonalcoholic fatty liver disease. Clin Mol Hepatol. 2023;29(Suppl):S32–S42. doi: 10.3350/cmh.2022.0365. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Grander C, Grabherr F, Tilg H. Non-alcoholic fatty liver disease: pathophysiological concepts and treatment options. Cardiovasc Res. 2023;119(9):1787–1798. doi: 10.1093/cvr/cvad095. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Hardy T, Zeybel M, Day CP, Dipper C, Masson S, McPherson S, et al. Plasma DNA methylation: a potential biomarker for stratification of liver fibrosis in non-alcoholic fatty liver disease. Gut. 2017;66(7):1321–1328. doi: 10.1136/gutjnl-2016-311526. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Kingwell K. NASH field celebrates ‘hurrah moment’ with a first FDA drug approval for the liver disease. Nat Rev Drug Discov. 2024;23(4):235–237. doi: 10.1038/d41573-024-00051-1. [DOI] [PubMed] [Google Scholar]
  • 5.Ravndal L, Lindvig KP, Jensen EL, Sunde A, Nassehi D, Thiele M, et al. Algorithms for early detection of silent liver fibrosis in the primary care setting-a scoping review. Expert Rev Gastroenterol Hepatol. 2023;17(10):985–997. doi: 10.1080/17474124.2023.2255522. [DOI] [PubMed] [Google Scholar]
  • 6.Anstee QM, Castera L, Loomba R. Impact of non-invasive biomarkers on hepatology practice: Past, present and future. J Hepatol. 2022;76(6):1362–1378. doi: 10.1016/j.jhep.2022.03.026. [DOI] [PubMed] [Google Scholar]
  • 7.European Association for the Study of the Liver EASL Clinical Practice Guidelines on non-invasive tests for evaluation of liver disease severity and prognosis-2021 update. J Hepatol. 2021;75(3):659–689. doi: 10.1016/j.jhep.2021.05.025. [DOI] [PubMed] [Google Scholar]
  • 8.Tincopa MA, Loomba R. Non-invasive diagnosis and monitoring of non-alcoholic fatty liver disease and non-alcoholic steatohepatitis. Lancet Gastroenterol Hepatol. 2023;8(7):660–670. doi: 10.1016/S2468-1253(23)00066-3. [DOI] [PubMed] [Google Scholar]
  • 9.Wang J, Qin T, Sun J, Li S, Cao L, Lu X. Non-invasive methods to evaluate liver fibrosis in patients with non-alcoholic fatty liver disease. Front Physiol. 2022;13:1046497. doi: 10.3389/fphys.2022.1046497. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.van Kleef LA, Sonneveld MJ, de Man RA, de Knegt RJ. Poor performance of FIB-4 in elderly individuals at risk for chronic liver disease-implications for the clinical utility of the EASL NIT guideline. J Hepatol. 2022;76(1):245–246. doi: 10.1016/j.jhep.2021.08.017. [DOI] [PubMed] [Google Scholar]
  • 11.Ali H, Shahzil M, Moond V, Shahzad M, Thandavaram A, Sehar A, et al. Non-pharmacological approach to diet and exercise in metabolic-associated fatty liver disease: Bridging the gap between research and clinical practice. J Pers Med. 2024;14(1):61. doi: 10.3390/jpm14010061. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Yang H, Atak D, Yuan M, Li M, Altay O, Demirtas E, et al. Integrative proteo-transcriptomic characterization of advanced fibrosis in chronic liver disease across etiologies. Cell Rep Med. 2025;6(2):101935. doi: 10.1016/j.xcrm.2025.101935. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Wu YL, Lin ZJ, Li CC, Lin X, Shan SK, Guo B, et al. Epigenetic regulation in metabolic diseases: mechanisms and advances in clinical study. Signal Transduct Target Ther. 2023;8(1):98. doi: 10.1038/s41392-023-01333-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Gerhard GS, Malenica I, Llaci L, Chu X, Petrick AT, Still CD, DiStefano JK. Differentially methylated loci in NAFLD cirrhosis are associated with key signaling pathways. Clin Epigenetics. 2018;10(1):93. doi: 10.1186/s13148-018-0525-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Ulukan B, Sila Ozkaya Y, Zeybel M. Advances in the epigenetics of fibroblast biology and fibrotic diseases. Curr Opin Pharmacol. 2019;49:102–109. doi: 10.1016/j.coph.2019.10.001. [DOI] [PubMed] [Google Scholar]
  • 16.Xu RH, Wei W, Krawczyk M, Wang W, Luo H, Flagg K, et al. Circulating tumour DNA methylation markers for diagnosis and prognosis of hepatocellular carcinoma. Nat Mater. 2017;16(11):1155–1161. doi: 10.1038/nmat4997. [DOI] [PubMed] [Google Scholar]
  • 17.Zeybel M, Hardy T, Robinson SM, Fox C, Anstee QM, Ness T, et al. Differential DNA methylation of genes involved in fibrosis progression in non-alcoholic fatty liver disease and alcoholic liver disease. Clin Epigenetics. 2015;7(1):25. doi: 10.1186/s13148-015-0056-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Zeybel M, Hardy T, Wong YK, Mathers JC, Fox CR, Gackowska A, et al. Multigenerational epigenetic adaptation of the hepatic wound-healing response. Nat Med. 2012;18(9):1369–1377. doi: 10.1038/nm.2893. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Batista AD, Barros CJP, Costa TBBC, de Godoy MMG, Silva RD, Santos JC, et al. Proton nuclear magnetic resonance-based metabonomic models for non-invasive diagnosis of liver fibrosis in chronic hepatitis C: Optimizing the classification of intermediate fibrosis. World J Hepatol. 2018;10(1):105–115. doi: 10.4254/wjh.v10.i1.105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Meng F, Wu Q, Zhang W, Hou S. Application of interpretable machine learning models based on ultrasonic radiomics for predicting the risk of fibrosis progression in diabetic patients with nonalcoholic fatty liver disease. Diabetes Metab Syndr Obes. 2023;16:3901–3913. doi: 10.2147/DMSO.S439127. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Bonner JC. Regulation of PDGF and its receptors in fibrotic diseases. Cytokine Growth Factor Rev. 2004;15(4):255–273. doi: 10.1016/j.cytogfr.2004.03.006. [DOI] [PubMed] [Google Scholar]
  • 22.Kanwal F, Neuschwander-Tetri BA, Loomba R, Rinella ME. Metabolic dysfunction-associated steatotic liver disease: Update and impact of new nomenclature on the American Association for the Study of Liver Diseases practice guidance on nonalcoholic fatty liver disease. Hepatology. 2024;79(5):1212–1219. doi: 10.1097/HEP.0000000000000670. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Kocabayoglu P, Lade A, Lee YA, Dragomir AC, Sun X, Fiel MI, et al. β-PDGF receptor expressed by hepatic stellate cells regulates fibrosis in murine liver injury, but not carcinogenesis. J Hepatol. 2015;63(1):141–147. doi: 10.1016/j.jhep.2015.01.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Lambrecht J, Verhulst S, Mannaerts I, Sowa JP, Best J, Canbay A, et al. A PDGFRβbased score predicts significant liver fibrosis in patients with chronic alcohol abuse, NAFLD and viral liver disease. EBioMedicine. 2019;43:501–512. doi: 10.1016/j.ebiom.2019.04.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Shi J, Song S, Li S, Zhang K, Lan Y, Li Y. TNF-α/NF-κB signaling epigenetically represses PSD4 transcription to promote alcohol-related hepatocellular carcinoma progression. Cancer Med. 2021;10(10):3346–3357. doi: 10.1002/cam4.3832. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Gorden DL, Myers DS, Ivanova PT, Fahy E, Maurya MR, Gupta S, et al. Biomarkers of NAFLD progression: a lipidomics approach to an epidemic. J Lipid Res. 2015;56(3):722–736. doi: 10.1194/jlr.P056002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Wu CC, Yeh WC, Hsu WD, Islam MM, Nguyen PAA, Poly TN, et al. Prediction of fatty liver disease using machine learning algorithms. Comput Methods Programs Biomed. 2019;170:23–29. doi: 10.1016/j.cmpb.2018.12.032. [DOI] [PubMed] [Google Scholar]
  • 28.European Association for the Study of the Liver EASL Clinical Practice Guidelines on non-invasive tests for evaluation of liver disease severity and prognosis-2021 update. J Hepatol. 2021;75(3):659–689. doi: 10.1016/j.jhep.2021.05.025. [DOI] [PubMed] [Google Scholar]
  • 29.Loomba R, Adams LA. Advances in non-invasive assessment of hepatic fibrosis. Gut. 2020;69(7):1343–1352. doi: 10.1136/gutjnl-2018-317593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Polo TCF, Miot HA. Use of ROC curves in clinical and experimental studies. J Vasc Bras. 2020;19:e20200186. doi: 10.1590/1677-5449.200186. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Hepatology Forum are provided here courtesy of Turkish Association for the Study of the Liver

RESOURCES