Human Serum Albumin Profiling by Top-down Analysis Enables Multi-Class Liver Fibrosis Staging: A Cross-Platform Validation Study

Human Serum Albumin Profiling by Top-down Analysis Enables Multi-Class Liver Fibrosis Staging: A Cross-Platform Validation Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Human Serum Albumin Profiling by Top-down Analysis Enables Multi-Class Liver Fibrosis Staging: A Cross-Platform Validation Study Souleiman El Balkhi, Racym Berrah, François Ludovic Sauvage, Léa Le Du, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9371068/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 10 You are reading this latest preprint version Abstract Chronic liver disease (CLD) affects millions worldwide, yet accurately staging its progression without liver biopsy remains a major clinical challenge. Human serum albumin (HSA), the most abundant blood protein synthesized exclusively by the liver, undergoes measurable structural modifications as liver disease advances, making it a potential molecular marker of disease severity. Using high-resolution liquid chromatography–mass spectrometry (LC-HR-MS), we quantified native HSA and nine modified isoforms in plasma from 172 CLD patients spanning all fibrosis stages and 82 healthy controls. Native HSA declined markedly with disease severity, reaching 4.1–4.2 g/L in decompensated cirrhosis versus 12.2 g/L in controls. Modified isoforms showed stage-specific patterns, and their ratios to native HSA amplified the diagnostic signal for advanced disease. A machine learning classifier trained on the full albumin spectral profile achieved substantial agreement with biopsy-based staging and outperformed the widely-used FIB-4 index in clinical triage (81.5% vs. 59.3% accuracy). Critically, these results were fully reproduced on two independent instruments from different manufacturers (McNemar p = 0.149), confirming the platform-independence of the albumin signature. These findings establish HSA spectral profiling as a promising non-invasive staging tool for CLD, with cross-platform reproducibility supporting its translation to multicenter clinical practice. Health sciences/Biomarkers Health sciences/Diseases Health sciences/Gastroenterology Health sciences/Medical research Chronic liver disease biomarker Post-transcriptional modification albumin Machine learning Cross-platform validation Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Chronic liver diseases (CLDs) are a global health burden, contributing significantly to morbidity and mortality worldwide ( 1 ). The progression of CLDs through various stages, from initial injury to fibrosis, cirrhosis, and eventual decompensation or the development of hepatocellular carcinoma, underscores the critical need for accurate diagnostic and staging tools ( 2 – 7 ). Non-invasive tools for assessing liver fibrosis and significative portal hypertension have been widely developed in recent years, such as liver and more recently splenic elastometry and non-invasive biological fibrosis tests such as Fibrotest R and Fibrometer R . However, while they have made a definite contribution in practice, they do not meet the objective of EASL with an accuracy > 80%, except for advanced fibrosis diagnosis. Their limitation lies primarily in the difficulty of predicting intermediate fibrosis scores and the time to onset of liver-related events ( 8 ). While prognostic scores like the Model for End-Stage Liver Disease (MELD) and Child-Pugh scores are widely used ( 4 , 9 ), they also have limitations, and accurately predicting the disease outcomes, especially the transition between compensated and decompensated states or the risk of Acute on Chronic Liver Failure (ACLF) development, remains challenging ( 4 , 10 ). Improved biomarkers are needed to better stratify patients, monitor disease progression, help therapeutic decisions and interpret treatment response. Human serum albumin (HSA), the most abundant plasma protein ( 1 , 11 , 12 ), is synthesized exclusively by the liver ( 1 , 12 ). Traditionally recognized for its primary role in maintaining plasma oncotic pressure and modulating fluid distribution ( 9 , 11 ), HSA is now understood to possess a wide array of critical non-oncotic biological functions ( 1 , 4 , 7 , 9 – 13 ). These include binding, transport, and detoxification of numerous endogenous and exogenous substances (such as bilirubin, bile acids, fatty acids, metals, and drugs) ( 14 , 15 ), potent antioxidant and free-radical scavenging activities, immunomodulatory effects, and contributions to endothelial stabilization and homeostasis. The structural integrity of the HSA molecule, a 66.5 kDa globular protein organized into three homologous domains, is crucial for these non-oncotic functions ( 16 ). In healthy individuals, the majority (70–80%) of circulating HSA exists in the reduced form, known as human mercaptalbumin (HMA) ( 9 , 17 ), a smaller fraction (20–30%) as reversibly oxidized human nonmercaptalbumin-1 (HNA1), and a very minor fraction (< 5%) as irreversibly oxidized human nonmercaptalbumin-2 (HNA2)( 18 ). However, HSA is susceptible to a variety of other post-transcriptional modifications (PTMs) ( 13 , 17 ), especially under conditions of systemic inflammation and oxidative stress characteristic of advanced CLDs ( 3 , 5 , 10 – 12 ). The accumulation of these modified HSA molecules, or proteoforms, constitutes the microheterogeneity of circulating albumin ( 2 , 9 , 13 , 19 , 20 ). In patients with CLDs, particularly those with decompensated cirrhosis and ACLF, significant alterations in circulating HSA are observed, encompassing both quantitative reduction (hypoalbuminemia) and profound qualitative changes in its structure and function ( 1 , 4 , 7 , 10 , 11 ). Numerous studies utilizing high-performance liquid chromatography coupled mass spectrometry have documented extensive PTMs in HSA from cirrhotic patients ( 2 – 4 , 9 , 13 , 17 – 19 , 21 ) with a significant decrease in the proportion of native, structurally intact HMA and a corresponding increase in oxidized forms, HNA1 and particularly the irreversibly oxidized HNA2. Specific isoforms, such as cysteinylated, glycated, N-terminal truncated, C-terminal truncated, sulfinylated forms, as well as HSA homodimers, have been identified and found to be significantly more abundant in patients with decompensated cirrhosis compared to healthy controls or patients with compensated disease. Crucially, these qualitative changes impair HSA's non-oncotic functions and reduce binding capacity for various ligands ( 14 , 16 ), diminish antioxidant potential ( 3 ), and alter transport efficiency ( 7 ) of HSA in advanced liver disease ( 21 ). This functional impairment has led to the concept of "effective albumin concentration," suggesting that the overall biological function of albumin relates not just to its total concentration but also significantly to the preservation of its structural and functional integrity. Indeed, several studies have highlighted the prognostic significance of specific HSA isoforms: the level of native HMA, the irreversibly oxidized HNA2 fraction, the ischemia-modified albumin ratio (IMAR), certain cysteinylated/truncated isoforms, and specific homodimers might be independent predictors of CLD complications and short-term mortality in hospitalized patients with cirrhosis, suggesting their superiority compared to total serum albumin concentration ( 2 , 4 , 6 , 7 , 9 , 10 ). However, the utility of specific HSA isoforms as diagnostic markers for differentiating stages of CLD has not been systematically investigated. Most studies have focused on patients with established decompensation or ACLF, often comparing them to healthy controls or less well-defined cirrhotic groups ( 2 , 4 – 6 , 9 , 10 ). Furthermore, while various analytical methods have been employed ( 13 , 17 ), a standardized approach to profile and quantify a comprehensive panel of relevant HSA isoforms for diagnostic staging purposes is lacking. There is therefore a clear need: (i) to systematically determine whether the distinctive patterns of HSA isoform distribution can reliably discriminate CLD stages using a comprehensive quantitative approach; and (ii) to establish whether such a diagnostic signature is sufficiently robust to be reproducible across different analytical platforms, a prerequisite for multi-center clinical implementation. Therefore, this study pursued two co-primary objectives. First, to systematically characterize the complete profile of circulating HSA isoforms using LC-HR-MS across well-defined cohorts representing the full spectrum of chronic liver disease, including pre-cirrhotic fibrosis, compensated cirrhosis, and decompensated cirrhosis, compared to healthy controls. The primary aim was to evaluate the diagnostic relevance of individual and combined HSA isoforms in differentiating CLD stages, and to determine whether their multivariate signature provides superior performance to available non-invasive tools. Second, to validate the analytical reproducibility and instrument-independence of the ALBOM signature by applying the same classification model to samples analysed on two independent LC-HR-MS platforms from distinct manufacturers, and to quantify the degree of cross-platform agreement using formal method comparison statistics. Results Between January 2021 and January 2023, 172 patients diagnosed with CLD at various stages and 82 control subjects met the initial eligibility criteria. Detailed demographic and clinical characteristics of the patient cohort, stratified by liver fibrosis stage and cirrhosis severity, are presented in Table 1 . The patient group consisted predominantly of males (116/172, 67.4%), with a median age at inclusion of 61 years. Table 1 Population’s characteristics . Number of patients F0/F1 F2 F3 Child Pugh A Child Pugh B Child Pugh C Total 36(21%) 23(13%) 30(17%) 37(22%) 26(15%) 20(12%) 172 Gender Female 17 6 12 12 1 8 56 Male 15 17 18 25 25 12 116 Age (year) Average age 54 61 64 62 64 59 60 Median age 53 62 62 62 63 57 61 Etiology MASH 13 13 22 12 3 0 62 (36%) Alcohol 1 0 0 9 16 14 40(23,2%) HBV 10 1 2 2 0 0 15(8,7%) HCV 2 3 1 5 0 0 11(6,3%) Hemochromatosis 1 0 0 2 0 0 3(1,7%) Cardiac 0 0 0 0 1 0 1(0,6%) AIH 3 2 4 1 0 1 11(6,3%) Autoimmune cholangitis 3 0 0 0 0 0 3(1,7%) PBC 3 0 0 0 0 0 3(1,7%) PSC 0 2 0 0 0 0 2(1,1%) Crytogenic 0 0 1 3 0 0 4(2,3%) Mixed 0 2 0 3 7 5 17(9,9%) Mixed etiology Alcohol + MASH 0 0 0 2 5 4 11(6,3%) PBC + MASH 0 2 0 0 0 0 2(1,1%) Alcohol + HBV 0 0 0 0 0 1 1(0,6%) Alcohol + HCV 0 0 0 1 1 0 2(1,1%) Average MELD NS NS NS 8 15 21 14 Hepatic encephalopathy All stages NS NS NS 1 7 10 18 Moderate NS NS NS 1 4 8 13 Disabling NS NS NS 0 3 2 5 Ascite All stages NS NS NS 0 14 16 30 Average abundance NS NS NS 0 7 11 18 Tense or refractory NS NS NS 0 7 5 12 NS: Not suitable, MASH: Metabolic associated steatohepatitis, AIH: Autoimmune hepatitis, HCV: Hepatitis C virus, HBV: Hepatitis B virus, PBC: primary biliary cholangitis, PSC: primary sclerosing cholangitis, MELD: Model for end stage liver disease. The distribution across fibrosis stages was as follows: 36 (20.9%) F0/F1, 23 (13%) F2, 30 (17%) F3, and 83 (48%) F4 (cirrhosis). Fibrosis staging was mainly based on transient elastography. Among the 83 patients with cirrhosis (F4), 37 (44.6% of F4; 21.5% of total cohort) were classified as Child-Pugh Class A, 26 (31.3% of F4; 15.1% of total) as Class B, and 20 (24.1% of F4; 11.6% of total) as Class C. The Meld score averaged 15 for Child Pugh B and 21 for Child Pugh C patients respectively (Table 1 ). The primary etiologies of CLD were non-alcoholic steatohepatitis (MASH; 62 patients, 36.0%) and alcohol-related liver disease (ALD; 40 patients, 23.3%). A combination of ALD and MASH was identified in 11 patients (6.4%). Other etiologies included viral hepatitis B (HBV; 15 patients, 8.7%), viral hepatitis C (HCV; 11 patients, 6.4%), autoimmune hepatitis (AIH; 11 patients, 6.4%), and less common causes such as hemochromatosis, primary biliary cholangitis (PBC), primary sclerosing cholangitis (PSC), cardiac hepatopathy, and cryptogenic cirrhosis (Table 1 ). Ascites was present in 30 of the 83 (36.1%) individuals with cirrhosis. It was graded as moderate in 18 patients and was tense or refractory to diuretics in 12 patients. Hepatic encephalopathy (West Haven grade II or III) was documented in 18 (21.7%) cirrhotic patients. Further details on the distribution of clinical complications according to Child-Pugh score are provided in Table 2 . Table 2 – Diagnosis methods and Child-Pugh classification for F4 patients. Fibrosis stage Fibroscan Liver biopsy Fibroscan + liver biopsy F0/F1 35 (0.97) 4 (0.11) 3 (0.08) F2 23 ( 1 ) 5 (0.22) 5 (0.22) F3 28 (0.93) 12 (0.4) 10 (0.33) F4 32 (0.45) 19 (0.23) 4 (0.05) F4_Child Pugh A 25 (0.68) 5 (0.13) 2 (0.05) F4_Child Pugh B 3 (0.11) 6 (0.23) 1 (0.04) F4_Child Pugh C 4 (0.2) 8 (0.4) 1 (0.05) HSA isoforms profiling across liver disease stages The absolute concentrations of native HSA and its major isoforms, stratified by fibrosis stage and Child-Pugh class for cirrhotic patients are presented in Fig. 1 . The average value of the native albumin isoform (native HSA) in healthy controls was 12.2 g/L. Native HSA was slightly lower but not significantly in patients with F0/F1 fibrosis (10.6 g/L), F2 (9.9 g/L), and F3 (10.6 g/L), while it was significantly reduced in patients diagnosed with Child-Pugh A cirrhosis (F4_A: 10.2 g/L), Child-Pugh B cirrhosis (F4_B: 4.1 g/L), and Child-Pugh C cirrhosis (F4_C: 4.2 g/L) compared to healthy controls (Fig. 1 A). However, by combining F2 and F3 (F2/F3), a discrimination between F2/F3 and the control group was observed (Fig. 1 B). A significant difference was found between the F4_B/F4_C groups compared to the other groups. It is important to note that F4_A was distinguished from F4_B. However, no significant difference was detected between F4_B and F4_C. Concentrations of other modified HSA isoforms In total, up 10 non-native albumin isoforms were quantified in most patients, including N-terminally truncated (HSA-DA) and C-terminally truncated (HSA-L) albumin, cysteinylated (HSA + CYS), oxidized (HSA+SO 3 H), and glycated (HSA+Glyc) forms, as well as combination isoforms such as glycated cysteinylated (HSA + CYS+GLYC) and truncated cysteinylated (HSA-DA + CYS). We compared the mean concentration (g/L) of these albumin isoforms according to the stage of cirrhosis (Fig. 2 ). The isoforms revealed three distinct patterns of evolution corresponding to the progression of liver disease. First, several isoforms involved in moderate oxidative stress and glycation exhibited a biphasic trajectory. Specifically, the concentrations of cysteinylated (HSA + CYS), singly glycated (HSA+GLYC), and doubly glycated (HSA+2GLYC) isoforms generally increased from the control state through the stages of compensated cirrhosis. The mean concentration of HSA + Cys, for example, rose from 8.7 g/L in controls to a peak of 11.1 g/L in patients with Child-Pugh A cirrhosis. However, as the disease progressed into severe decompensation, the concentrations of these isoforms paradoxically declined, with HSA + CYS falling to 7.9 g/L and 6.8 g/L in Child-Pugh B and C patients, respectively (Figs. 2 .E, 2.F, 2.H). In contrast, the more complex, multiply-modified cysteinylated and doubly glycosylated isoform (HSA + CYS+2GLYC) showed a progressive and significant increase with advancing cirrhosis. Its concentration was markedly higher in Child-Pugh A (0.12 g/L), B (0.18 g/L), and C (0.2 g/L) stages compared to almost undetectable levels in healthy controls, suggesting it serves as a marker of cumulative, end-stage protein damage (Fig. 2 .I). Finally, a third group of isoforms, primarily those involving truncation or irreversible oxidation, showed a general trend of reduction, particularly in the most advanced disease stages. The N-terminally truncated isoform (HSA-DA) was significantly lower across all disease stages compared to its level in controls (0.2 g/L). Similarly, the irreversibly oxidized (HSA+SO 3 H) and the combined truncated-cysteinylated (HSA-DA + CYS) isoforms were notably decreased in patients with Child-Pugh B and C cirrhosis (Figs. 2 .A, 2.C, 2.D). The C-terminally truncated isoform (HSA-L) showed no significant variation across the disease spectrum (Fig. 2 .B). Discriminatory behavior of HSA isoforms To better assess the discriminant capacities of HSA isoforms, we normalized the average concentration of some clinically significant isoforms. The ratios of cysteinylated, glycated, and cysteinylated-glycated isoforms to native albumin (HSA + CYS/Native, HSA+GLYC/Native, and HSA + CYS+GLYC/Native) showed a marked and statistically significant increase with the evolution of the CLD stage. Notably, a significant increase in the concentration ratio of these isoforms was found in patients with decompensated cirrhosis B and C, whereas the concentration remained comparable in the earlier fibrotic stages (Fig. 3 A). The glycated isoform was the only one capable of discriminating between the control group and F2 (Fig. 3 A.2). Moreover, the GLYC and CYS+GLYC isoforms were able to distinguish between F2 and F4_B (Fig. 3 A.2 and 3A.3), which was not the case with the cysteinylated isoform (Fig. 3 A.1). We analysed the discriminant potential of these specific isoforms between the control group and the different fibrotic stages using ROC curves. For the cysteinylated isoform, a discrimination with a sensitivity of 65% and a specificity of 99% between the F4_C group of 20 patients and the control group was observed. For the glycated and cysteinylated-glycated isoforms, better discrimination of the F4_B and the F4_C group was observed. For the GLYC isoform, a diagnostic sensitivity of 85% and a specificity of 100% were observed for F4_B versus a sensitivity of 70% and a specificity of 99% for F4_C. For the CYS+GLYC isoform, a diagnostic sensitivity of 77% and a specificity of 99% was observed for F4_B versus a sensitivity of 70% and a specificity of 99% for F4_C (Fig. 3 B). In comparison, analysis of classical laboratory parameters confirmed the progressive deterioration of liver function across the patient cohorts. Markers of hepatic synthesis and excretion, such as serum albumin and bilirubin, remained stable through early fibrosis and compensated cirrhosis (Child-Pugh A). However, they showed significant deterioration in advanced decompensation, with albumin levels progressively decreasing while bilirubin markedly increased in Child-Pugh B and C stages. In parallel, the FIB-4 index and AST levels demonstrated a clear, stepwise increase corresponding to advancing disease severity, while ALT lacked discriminatory power (Figure S1 ). However, no single isoform was sufficient to discriminate all six CLD stages simultaneously, motivating the development of a multivariate classification approach incorporating the full spectral profile. Cross-Platform Validation To evaluate whether the diagnostic signature encoded in the albumin spectral profile is platform-independent, classification models were independently trained and evaluated on samples acquired on two distinct LC-HR-MS instruments. Despite differences in instrument architecture, software-based baseline correction algorithms, and spectral resolution between Platform 1 and Platform 2, the QWK of both classifiers fell within the range defined as 'substantial to near-perfect agreement', and their 95% bootstrap confidence intervals showed substantial overlap (Platform 1: [0.735–0.923]; Platform 2: [0.822–0.964]; Figure S4).Cross-platform equivalence was formally confirmed by McNemar's test applied to paired patient predictions (p = 0.149), indicating no statistically significant difference in classification decisions between platforms. The Jaccard Similarity Index of prediction errors between platforms was 0.696, meaning that approximately 70% of misclassified patients were identically misclassified by both instruments (Figure S5). This convergence of errors is a key finding: it indicates that the sources of misclassification reside in patient-level biological ambiguity at transitional fibrosis stages, particularly at the F2/F4A and F4A/F4B boundaries, rather than in instrument-specific noise or systematic bias. The slightly lower performance of Platform 1 compared to Platform 2 is consistent with a technical attenuation of high-mass albumin peaks (> 67,500 Da, corresponding to poly-glycated adducts) by the Platform 1 baseline correction algorithm, which geometrically misidentifies these broad, disease-specific peaks as baseline drift, partially suppressing the diagnostic signal in the F4B and F4C classes. This observation does not affect the validity of the Platform 1 classifier but provides an actionable optimization target for future implementations. Discussion This study provides the first comprehensive, quantitative analysis of a wide panel of human serum albumin (HSA) isoforms across the full spectrum of CLD, from early-stage fibrosis to end-stage decompensated cirrhosis. While previous research has established that structural and functional alterations of HSA occur in advanced liver disease, these investigations have predominantly focused on patients with advanced liver disease and have primarily investigated prognostic, rather than diagnostic, value. By employing a robust and sensitive top-down liquid chromatography high-resolution mass spectrometry (LC-HR-MS) method applied across two independent instrument platforms ( 18 ), our work moves beyond relative measurements to map the dynamic, stage-specific changes in the HSA isoforms landscape. This approach offers a novel "molecular staging" system that reflects the underlying pathophysiology of CLD progression, providing a more granular view than is achievable with conventional biomarkers or elastography alone. Our findings both confirm and substantially extend the existing literature. The observed progressive decline in native HSA concentration with increasing disease severity, particularly the precipitous drop in patients with Child-Pugh class B and C cirrhosis, is in strong agreement with the concept of a diminishing "effective albumin concentration" in advanced liver disease ( 9 , 10 , 12 ). This loss of the structurally and functionally intact HSA pool is a key pathophysiological feature, as native albumin is critical for mitigating the systemic inflammation and oxidative stress that drive disease progression ( 4 – 6 , 11 , 12 ). However, our data clarify that while native HSA is a reliable biomarker for predicting and identifying decompensated cirrhosis (Child-Pugh B and C), its utility for diagnosing the initial stages of fibrosis is limited, showing low sensitivity and specificity in these earlier phases. The analysis of modified isoforms reveals a highly dynamic and nuanced picture. For instance, the concentrations of cysteinylated (HSA + Cys) and glycated (HSA+Glyc) isoforms exhibit a biphasic pattern: they increase during the transition from a healthy state to compensated cirrhosis (F4_A), likely reflecting the escalating systemic oxidative and glycative stress, but subsequently decline in decompensated cirrhosis (F4_B and F4_C). This is a novel and critical observation that suggests a complex interplay of factors in end-stage disease. The decline in the absolute concentration of these modified forms may be due to the profound depletion of the native HSA as the substrate for modification, an accelerated clearance of these moderately modified isoforms, or their conversion into more complex, multiply-modified species, such as the doubly- and triply-modified isoforms (e.g., HSA + CYS+2GLYC) that we observed to increase in the most advanced stages. This apparent paradox, where the absolute concentration of a modified isoform decreases while the disease worsens, is resolved when considering the isoform ratios. The use of ratios, such as (HSA + Cys)/Native, normalizes for the overall drop in albumin synthesis and directly reflects the escalating proportion of damaged albumin, thus powerfully amplifying the diagnostic signal for advanced disease. This underscores that the dynamic flux of specific isoforms offers a far richer dataset for staging than a simple monotonic change in a single biomarker. The distinct behavior of individual isoforms suggests specific diagnostic utilities. The finding that the N-terminally truncated isoform (HSA-DA) was significantly reduced across all disease stages compared to controls is an unexpected result, as truncation is often considered a damage product. This may suggest that either the control population has a higher baseline for this specific proteoform, or that HSA-DA is subject to accelerated clearance or further modification even in early stages of liver injury, a hypothesis that warrants further investigation. Furthermore, the ROC curve analysis revealed that while cysteinylated isoforms were highly effective at distinguishing the most severe stage (F4_C) from controls, glycated and cysteinylated-glycated isoforms appeared superior in discriminating the transition to Child-Pugh B cirrhosis. However, these observations also make it clear that no single isoform is sufficient to accurately classify all disease stages. This highlights that HSA must be considered as an interdependent network of isoforms, where the concentration of one isoform is influenced by the availability and flux of others. Consequently, the true strength of this approach lies in the integrative analysis of the entire HSA isoform profile. While individual isoforms or classical biomarkers like FIB-4 (Figure S1 and S2 in supplemental data) can distinguish between distant stages (e.g., F0 vs. F4_C), they often fail to provide clear separation between contiguous intermediate stages. Our principal component analysis (PCA) compellingly demonstrates that the global HSA isoforms signature creates a distinct and progressively shifting pattern that maps directly to disease progression. The OrdinalForest classification model, trained on 75 spectral features within the albumin m/z 66,000–68,000 Da region and four routine clinical variables, achieved QWK values of 0.862 and 0.916 on two independent platforms, a performance levels that substantially exceed what is achievable by FIB-4 alone in multiclass staging (QWK = 0.188–0.229). The feature importance analysis revealed that while routine clinical variables (total protein, albumin, INR, bilirubin) contributed substantially to the combined model, spectral albumin peaks representing the cysteinylated and glycated isoform regions (m/z ∶66,230–66,600 Da) ranked among the top instrumental predictors, validating the biological rationale for our isoform profiling approach. The model's lower accuracy for the F4A class, consistently misclassified into adjacent stages, reaffirms the biological continuum argument: Child-Pugh A cirrhosis represents a transition state where the albumin molecular fingerprint overlaps significantly with advanced fibrosis (F2/F3) and this ambiguity is not a model artifact, but a reflection of the underlying pathophysiology. A key novel contribution of this work is the formal demonstration that the HSA spectral signature is reproducible across LC-HR-MS platforms from distinct manufacturers. The statistical equivalence of classification decisions (McNemar p = 0.149) and the high proportion of shared errors (Jaccard index 0.696) collectively argue that the diagnostic information is encoded in the biology of the sample — specifically in the molecular modifications of albumin — rather than in instrument-specific signal characteristics. This finding is essential for clinical translation: in a multi-center deployment scenario, patients would inevitably be analyzed on different platforms across participating institutions. Our data demonstrate that the our classifier would deliver consistent results regardless of whether a Bruker or Sciex instrument is used, provided that the same preprocessing pipeline is applied. This places our approach within the formal framework of method comparison studies, a standard required by clinical laboratory accreditation bodies. The slight performance advantage of Platform 2 over Platform 1 appears to arise from a technical difference in baseline correction algorithms rather than from a fundamental biological or clinical distinction and can likely be eliminated by optimizing peak detection parameters for the albumin high-mass region in Platform 1 implementations. The search for reliable non-invasive biomarkers of liver injury is a central goal in hepatology, and it is important to contextualize our findings. While markers of hepatocellular necrosis like glutamate dehydrogenase (GLDH) or keratin-18 (K18) can indicate acute hepatocyte damage, they often lack specificity and may not accurately reflect the chronic, cumulative processes of fibrosis ( 29 , 30 ). Similarly, microRNA-122 is a highly sensitive marker of liver injury but can be influenced by etiology and shows considerable inter-individual variability ( 31 – 34 ). In contrast, the HSA isoforms profile offers a unique advantage. As HSA is synthesized exclusively by the liver and has a long half-life, its modified forms represent an integrated, cumulative record of the systemic metabolic and inflammatory environment over weeks to months, making it an ideal candidate biomarker for a chronic, progressive disease. Other second-line tests with a combination of different biomarkers such as FibroTest R , FibroMeter R are often used but even if more performant, they also demonstrated limitations ( 25 ). In the present study, the ‘LC-TOF + Clinical’ model outperformed FIB-4 by 26 percentage points in 3-class triage accuracy (81.5% vs. 59.3%), and provided actionable classification in 62.5% of patients falling within the FIB-4 indeterminate gray zone (FIB-4: 1.30–2.67), a population for whom current non-invasive tools systematically fail to provide guidance These findings invite a re-evaluation of current paradigms in fibrosis classification. For decades, the METAVIR score from liver biopsy has been the gold standard, but its invasive nature and susceptibility to sampling error are significant drawbacks ( 35 ). Non-invasive methods like transient elastography have revolutionized clinical practice but measure a physical property (stiffness) that is an indirect and sometimes confounded surrogate for the complex biological activity of inflammation and fibrosis ( 25 ). Our data suggest that HSA isoforms profiling provides a direct window into the systemic biochemical consequences of liver disease, reflecting the cumulative impact of oxidative stress, inflammation, and metabolic dysregulation. This shifts the diagnostic paradigm from a structural or anatomical classification to a functional and molecular one. The ultimate goal of such a biomarker is not merely to stratify patients into existing categories but to provide predictive information ( 36 ). A specific HSA isoforms signature may, in the future, prove more effective at predicting the risk of fibrosis progression, clinical decompensation, response to therapy, or development of hepatocellular carcinoma than a simple fibrosis stage, thereby enabling a move towards a more proactive and personalized management of CLD. This study is not without limitations. First, its cross-sectional design allows for the characterization of stage-specific differences but does not permit the analysis of intra-individual changes over time; a longitudinal, multicentric study (MALAHBAR; NCT06318949) is now underway to confirm these findings and establish their predictive capacity. Second, fibrosis might have been misclassified, particularly for the F2 and F3 stages. The assessment of fibrosis stages has been mainly based on the Fibroscan R , whose limitations are known for deciding on intermediate scores F2 and F3. The dispersion of our data points in these groups likely reflects the known limitations of transient elastography ( 25 ). Even if the liver biopsy remains the gold standard, it can also be flawed depending on the size of the sample and the heterogeneity of the distribution of lesions in the liver. Third, a selection bias exists within our cohort, as the distribution of etiologies (mainly MASH and ALD) was not uniform across all disease severity groups, which could be a potential confounder. Future studies should address etiology-specific isoforms profiles. Fourth, while the demonstration of cross-platform reproducibility presented here represents a significant step toward clinical implementation, several analytical challenges remain. Specifically, the impact of different baseline correction algorithms between instruments on the quantification of high-mass glycated albumin species (> 67,500 Da) must be systematically characterized and controlled, and a formal proficiency testing scheme across participating laboratories will be required prior to multi-center deployment. Finally, the gray zone FIB-4 analysis (n = 8 patients) is preliminary and should be considered hypothesis-generating only. In conclusion, this study demonstrates that the HSA isoforms profile is dynamically and profoundly altered throughout the progression of chronic liver disease. We have established that an integrated, multivariate assessment of this "molecular fingerprint" provides a more biologically coherent and potentially more clinically useful measure of this disease severity than traditional methods or the analysis of single isoforms. These findings lay the groundwork for developing a new class of biomarkers aiming not just at diagnosing the present state of the liver, but at predicting its future course. Materials and methods Study design and patients This prospective, single-centre study was conducted between January 2021 and January 2023, enrolling patients with CLDs at various fibrosis stages who were managed at the Department of Hepatology, Limoges University Hospital, France. For each enrolled participant, demographic data (age, sex), etiology of liver disease, fibrosis stage, Child-Pugh score for cirrhosis, Model for End-Stage Liver Disease (MELD) score were prospectively collected. Inclusion criteria were: ( 1 ) age > 18 years; ( 2 ) confirmed diagnosis of CLD; and ( 3 ) availability of a routine blood sample collected within the preceding 24 hours of inclusion. Exclusion criteria comprised: patients undergoing dialysis or with a history oftransplantation; administration of contrast media, chelation drugs, blood transfusion, blood derivatives, or albumin infusion within the month prior to enrolment to rule out potential interference with albumin isoform analysis, specifically the Serum Enhanced Binding test ( 14 ). The control group consisted of healthy individuals recruited during the same period, defined by the absence of clinical evidence of liver dysfunction and liver function tests aspartate aminotransferase [AST], alanine aminotransferase [ALT], alkaline phosphatase [ALP], gamma-glutamyl transferase [GGT], total and conjugated bilirubin, within normal reference ranges. Clinical and biochemical assessments Routine liver function tests, coagulation parameters, and complete blood counts, were recorded from the patients' medical records at the time of sample collection. HSA isoforms were analyzed on leftovers of plasma samples obtained from routine blood collections drawn into lithium heparin Vacutainer® tubes (Becton Dickinson). Shortly after collection, samples were centrifuged at 3000 rpm (1500 g ) for 10 minutes and plasma aliquots were stored at -20°C until analysis. Liver fibrosis and cirrhosis staging Patients were categorized into liver fibrosis stages F0/F1, F2, F3 and F4 based on a hierarchical approach incorporating non-invasive and invasive methods. The primary method for fibrosis staging was transient elastography (TE), performed by an experienced examinator using the FibroScan® 502 Touch (Echosens, Paris, France) with the M and XL probes. Liver stiffness measurements (LSM) were considered reliable if a minimum of 10 successful acquisitions were achieved, with an interquartile range to median (IQR/M) ratio of less than 30%. Fibrosis stage was subsequently estimated using established LSM cut-off values according to the underlying liver disease etiology ( 22 – 26 ). Globally, the cut-offs used were 13.1 kPa for F4 (cirrhosis). For the purposes of the machine learning classification model and cross-platform analysis, patients originally staged as F3 were pooled with the F2 group, given the recognized overlap and limited reproducibility of intermediate fibrosis staging by transient elastography, and consistent with the approach recommended in EASL 2021 guidelines for non-invasive test development. When clinically indicated and if it was available within 12 months of study inclusion, liver biopsy specimens were reviewed by an experienced pathologist and histopathological staging of fibrosis performed according to the METAVIR scoring system ( 27 ). In cases where LSM or biopsy was not feasible or indicated in the patient for the diagnosis of cirrhosis, we established a consensus based on the combination of: (i) clinical signs, including a firm or nodular liver on palpation and evidence of portal hypertension and/or hepatocellular insufficiency (ii) biochemical markers, including but not limited to decreased prothrombin time or Factor V levels, hypoalbuminemia, hyperbilirubinemia, and thrombocytopenia; and (iii) classical morphological features on imaging, such as liver surface nodularity, parenchymal heterogeneity, and signs of portal hypertension ( 26 , 28 ). Human serum albumin isoform quantification Absolute quantification of HSA isoforms was performed using a previously validated and published top-down liquid chromatography-quadrupole time-of-flight mass spectrometry (LC-QTOF-MS) method ( 18 ). Briefly, the method involves a simple 1:50 (v/v) dilution of 20 µL patient serum in 0.9% NaCl after spiking with equine myoglobin (Mb, 4 g/L final concentration in the diluted sample before injection, corresponding to 0.08 g/L in the original serum) as an internal standard (IS) for mass recalibration and quantification. Sample preparation, LC-MS analysis conditions (including C4 column chromatography, gradient elution, ESI-QTOF settings), data acquisition, and data processing involving spectral deconvolution (mass range 66,000–68,000 Da) and mass recalibration using the Mb IS were performed exactly as described previously ( 18 ). All analyses were performed using the same validated top-down LC-HR-MS method. Two independent instrument platforms from distinct manufacturers were used across the study cohorts: Platform 1 (Bruker timsTOF Pro2) and Platform 2 (Sciex TripleTOF 5600+). Both platforms were operated under identical chromatographic and ionization conditions. Raw spectra were processed using platform-specific deconvolution software prior to normalization. Data from both platforms were subsequently subjected to identical preprocessing (Total Ion Current normalization followed by Probabilistic Quotient Normalization), feature selection, and classification procedures, as described below. For specific high-molecular-weight glycated isoforms (e.g., HSA + 2Glyc, HSA + Cys + 2Glyc) potentially present at low abundance or absent in the commercial HSAc standard used for primary calibration, their concentrations were estimated. This estimation was based on applying the calibration slope derived from the most abundant, structurally related glycated isoform quantified in the HSAc standard (e.g., the HSA + Glyc slope was used for HSA + 2Glyc), assuming comparable ionization efficiencies between closely related isoforms, as previously demonstrated ( 18 ). Spectral Preprocessing and Feature Selection For the multivariate classification analysis, the full albumin spectral region (m/z 66,000–68,000 Da) was used. Variables with more than 20% missing values were excluded. Spectra were normalized sequentially using Total Ion Current (TIC) normalization to account for instrument-to-instrument intensity differences, followed by Probabilistic Quotient Normalization (PQN) to correct for dilution effects. Feature selection was performed exclusively within the training partition to prevent data leakage. A Random Forest model was first fitted to the full normalized feature matrix, and feature importance was estimated by permutation. The optimal number of features (k) was determined by 5-fold cross-validated Quadratic Weighted Kappa (QWK) over a grid of k values (20, 30, 40, 50, 75, 100). The value k = 75 spectral variables was selected as providing the highest cross-validated QWK without overfitting. Standard routine clinical variables (total protein, serum albumin, prothrombin time expressed as INR, and total bilirubin) were incorporated into a combined model (LCTOF+ Clinical) to evaluate their incremental contribution. Classification Model Three ordinal classification architectures were evaluated: ( 1 ) a standard Forest (RF) with ordinal outcome treated as a factor; ( 2 ) a Hierarchical Random Forest (HRF) using a sequential Random binary decomposition of the six fibrosis classes; and ( 3 ) a Hierarchical OrdinalForest (HOF), which explicitly incorporates the ordinal structure of the class labels into the optimization criterion. All models were trained on an 80% stratified training split (stratified by fibrosis class) and evaluated on the held out 20% test set. Primary performance metric was the Quadratic Weighted Kappa (QWK), which penalizes predictions proportionally to their ordinal distance from the true class, making it the appropriate metric for this staged-disease classification. Balanced accuracy was reported as a secondary metric to account for class imbalance. All analyses were performed in R version 4.4.2 using the ordinalForest, ranger, and yardstick packages. Cross-Platform Validation To assess the reproducibility of the ALBOM diagnostic signature across LC-HR-MS platforms, an independent classification model was trained and evaluated on samples analyzed on each platform separately. The same preprocessing, feature selection, and classification pipeline was applied to both datasets (Bruker and Sciex) independently. Cross-platform equivalence was assessed using two complementary approaches. First, McNemar’s test was applied to paired predictions from patients whose samples were analyzed on both instruments, testing whether the two classifiers produced statistically different classification decisions. Second, the Jaccard Similarity Index (JSI) was calculated on the error matrices of both platforms to quantify the proportion of shared misclassifications: JSI = |errors_FR ∩ errors_AL| / |errors_FR ∪ errors_AL|. A JSI > 0.5 was interpreted as indicating that errors are primarily driven by patient-level biological ambiguity rather than instrument-specific variability. Statistical Analysis Quantitative data are presented as mean ± standard error of the mean (SEM) unless otherwise stated. One-way analysis of variance (ANOVA) followed by Tukey's multiple comparisons test was used to assess differences in HSA isoforms concentrations between patient groups defined by fibrosis stage (Control, F0/F1, F2, F3, F4-Child A, F4-Child B, F4-Child C). Associations between HSA isoforms and disease etiology were also explored using ANOVA. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the discriminatory performance of individual HSA isoforms in distinguishing between different clinically relevant stages of liver disease (e.g., controls vs. F0/F1, F3 vs. F4, Child-Pugh A vs. B/C). Optimal diagnostic thresholds were determined using the Youden index. Multivariate analysis of the HSA isoforms profile was conducted using principal component analysis (PCA). PCA was performed using custom scripts in RStudio software (version 2025.05.1 + 513, Posit Software) applied to the relative abundance data derived from the entire deconvoluted albumin mass spectrum between 66,000 and 67,500 Da to visualize overall profile differences between patient groups. Data analysis and figure generation were performed using R software and GraphPad Prism® (version 9 for Mac OS, GraphPad Software, USA). For the cross-platform comparison, statistical equivalence was assessed using McNemar’s test (two-sided, α = 0.05) and bootstrap confidence intervals for QWK were estimated from 1,000 resampling iterations. Ethics statement This study was conducted on residual biological material collected during routine clinical care at the University Hospital of Limoges (CHU de Limoges), France. Patients were informed of the potential use of their residual samples for research purposes and did not object, in accordance with the institutional information and consent procedures in place at CHU de Limoges. Under French law (Code de la Santé Publique, Art. L.1121-1), research conducted exclusively on residual biological material from informed patients who have not objected is classified as Recherche Impliquant la Personne Humaine de catégorie 3 (RIPH3). This regulatory category does not require review or approval by a Comité de Protection des Personnes (CPP), and no CPP submission was therefore made. The biocollection from which the samples were drawn is formally registered with the French Ministry of Health under declarations DC 2010 − 1074 and AC-2016-2758. All procedures were conducted in accordance with the principles of the Declaration of Helsinki and the French Bioethics Act 2011 − 814 (Loi relative à la bioéthique). Declarations Conflicts of interest statement The authors have no conflicts of interest to declare. Funding statement: This work received a local funding form CHU Limoges Author Contribution Souleiman El Balkhi: Conceptualization, Methodology, data collection, data analysis and interpretation, drafting the article. Racym Berrah: Cross-platform data analysis, machine learning model development, statistical validation, figure generation. *The first two authors contributed equally to this work. Léa Le Du: Patients inclusion, data collection, analysis and interpretationMohamad Ali Rahali and Roy Lakis: Sample analysis, data collection, analysis and interpretationFrançois Ludovic Sauvage: Sample analysis, data collection, mass spectrometry analysis and interpretationPierre Marquet and Franck Saint-Marcoux: Methodology, data analysis and interpretation, critical revision of the article.Paul Carrier and Veronique Loustaud-Ratti: Conceptualization, Methodology, data collection, data analysis and interpretation, drafting the article. Acknowledgement The authors are grateful to BISCEm unit (Univ. Limoges, UAR 2015 CNRS, US 42 Inserm, CHU Limoges) and Emilie Pinault for technical support regarding mass spectrometry analyses. Data Availability The datasets generated and analysed during the current study contain patient-level clinical and proteomic data and cannot be made publicly available, in accordance with applicable patient privacy regulations (French Data Protection Act and GDPR) and the terms governing the biocollection registrations DC 2010-1074 and AC-2016-2758 (French Ministry of Health). Anonymised data supporting the findings of this study are available upon reasonable request to the corresponding author, subject to approval by the institutional data access committee of the University Hospital of Limoges (CHU de Limoges). The R scripts used for spectral preprocessing, feature selection, and cross-platform ordinal classification are available from the corresponding author upon reasonable request. References Spinella, R., Sawhney, R. & Jalan, R. Albumin in chronic liver disease: structure, functions and therapeutic implications. Hepatol. Int. 10 (1), 124–132 (2016). Baldassarre, M. et al. Albumin Homodimers in Patients with Cirrhosis: Clinical and Prognostic Relevance of a Novel Identified Structural Alteration of the Molecule. Sci. Rep. 6 , 35987 (2016). Das, S. et al. Hyperoxidized albumin modulates neutrophils to induce oxidative stress and inflammation in severe alcoholic hepatitis. Hepatology 65 (2), 631–646 (2017). Oettl, K. et al. Oxidative albumin damage in chronic liver failure: relation to albumin binding capacity, liver dysfunction and survival. J. Hepatol. 59 (5), 978–983 (2013). Alcaraz-Quiles, J. et al. Oxidized Albumin Triggers a Cytokine Storm in Leukocytes Through P38 Mitogen-Activated Protein Kinase: Role in Systemic Inflammation in Decompensated Cirrhosis. Hepatology 68 (5), 1937–1952 (2018). Stauber, R. E. et al. Human nonmercaptalbumin-2: a novel prognostic marker in chronic liver failure. Ther. Apher Dial . 18 (1), 74–78 (2014). Jalan, R. et al. Alterations in the functional capacity of albumin in patients with decompensated cirrhosis is associated with increased mortality. Hepatology 50 (2), 555–564 (2009). Cales, P. et al. A new generation of non-invasive tests of liver fibrosis with improved accuracy in MASLD. J. Hepatol. 82 (5), 794–804 (2025). Domenicali, M. et al. Posttranscriptional changes of serum albumin: clinical and prognostic significance in hospitalized patients with cirrhosis. Hepatology 60 (6), 1851–1860 (2014). Baldassarre, M. et al. Determination of Effective Albumin in Patients With Decompensated Cirrhosis: Clinical and Prognostic Implications. Hepatology 74 (4), 2058–2073 (2021). Bernardi, M. Effective albumin - A novel paradigm in the management of decompensated liver cirrhosis. J. Transl Int. Med. 11 (1), 11–14 (2023). Garcia-Martinez, R. et al. Albumin: pathophysiologic basis of its role in the treatment of cirrhosis and its complications. Hepatology 58 (5), 1836–1846 (2013). Naldi, M., Baldassarre, M., Domenicali, M., Bartolini, M. & Caraceni, P. Structural and functional integrity of human serum albumin: Analytical approaches and clinical relevance in patients with liver cirrhosis. J. Pharm. Biomed. Anal. 144 , 138–153 (2017). El Balkhi, S. et al. Early detection of liver injuries by the Serum enhanced binding test sensitive to albumin post-transcriptional modifications. Sci. Rep. 14 (1), 1434 (2024). Lakis, R. et al. Semi-synthetic human albumin isoforms: Production, structure, binding capacities and influence on a routine laboratory test. Int. J. Biol. Macromol. 250 , 126239 (2023). Fanali, G. et al. Human serum albumin: from bench to bedside. Mol. Aspects Med. 33 (3), 209–290 (2012). Rahali, M. A. et al. Posttranslational-modifications of human-serum-albumin analysis by a top-down approach validated by a comprehensive bottom-up analysis. J. Chromatogr. B Analyt Technol. Biomed. Life Sci. 1224 , 123740 (2023). Lakis, R. et al. Absolute Quantification of Human Serum Albumin Isoforms by Internal Calibration Based on a Top-Down LC-MS Approach. Anal. Chem. 96 (2), 746–755 (2024). Naldi, M. et al. Mass spectrometry characterization of circulating human serum albumin microheterogeneity in patients with alcoholic hepatitis. J. Pharm. Biomed. Anal. 122 , 141–147 (2016). Naldi, M. et al. Mass spectrometric characterization of human serum albumin dimer: A new potential biomarker in chronic liver diseases. J. Pharm. Biomed. Anal. 112 , 169–175 (2015). Paar, M. et al. Albumin in patients with liver disease shows an altered conformation. Commun. Biol. 4 (1), 731 (2021). European Association for Study of L. Asociacion Latinoamericana para el Estudio del H. EASL-ALEH Clinical Practice Guidelines: Non-invasive tests for evaluation of liver disease severity and prognosis. J. Hepatol. 63 (1), 237–264 (2015). Boursier, J. et al. An extension of STARD statements for reporting diagnostic accuracy studies on liver fibrosis tests: the Liver-FibroSTARD standards. J. Hepatol. 62 (4), 807–815 (2015). Wong, G. L. et al. Liver stiffness-based optimization of hepatocellular carcinoma risk score in patients with chronic hepatitis B. J. Hepatol. 60 (2), 339–345 (2014). Dietrich, C. et al. EFSUMB Guidelines and Recommendations on the Clinical Use of Liver Ultrasound Elastography, Update 2017 (Short Version). Ultraschall in der Medizin -. Eur. J. Ultrasound . 38 (04), 377–394 (2017). Berzigotti, A. et al. EASL Clinical Practice Guidelines on non-invasive tests for evaluation of liver disease severity and prognosis – 2021 update. J. Hepatol. 75 (3), 659–689 (2021). Bedossa, P. & Poynard, T. An Algorithm for the Grading of Activity in Chronic Hepatitis C. Hepatology 24 (2), 289–293 (1996). Tapper, E. B. & Parikh, N. D. Diagnosis and Management of Cirrhosis and Its Complications: A Review. JAMA 329 (18), 1589–1602 (2023). Schomaker, S. et al. Assessment of Emerging Biomarkers of Liver Injury in Human Subjects. Toxicol. Sci. 132 (2), 276–283 (2013). Zhang, B. et al. Cytokeratin 18 knockdown decreases cell migration and increases chemosensitivity in non-small cell lung cancer. J. Cancer Res. Clin. Oncol. 142 (12), 2479–2487 (2016). Cione, E. et al. Liver Damage and microRNAs: An Update. Curr. Issues. Mol. Biol. 45 (1), 78–91 (2022). Jampoka, K. et al. Serum miR-29a and miR-122 as Potential Biomarkers for Non-Alcoholic Fatty Liver Disease (NAFLD). MicroRNA 7 (3), 215–222 (2018). Starckx, S. et al. Evaluation of miR-122 and Other Biomarkers in Distinct Acute Liver Injury in Rats. Toxicol. Pathol. 41 (5), 795–804 (2012). Wang, X., He, Y., Mackowiak, B. & Gao, B. MicroRNAs as regulators, biomarkers and therapeutic targets in liver diseases. Gut 70 (4), 784–795 (2021). Sebastiani, G. Non-invasive assessment of liver fibrosis in chronic liver diseases: Implementation in clinical practice and decisional algorithms. World J. Gastroenterol. ; 15 (18). (2009). Boursier, J. et al. Non-invasive tests accurately stratify patients with NAFLD based on their risk of liver-related events. J. Hepatol. 76 (5), 1013–1020 (2022). Additional Declarations No competing interests reported. Supplementary Files Supplementaldata.docx Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 18 May, 2026 Reviews received at journal 16 May, 2026 Reviews received at journal 07 May, 2026 Reviewers agreed at journal 06 May, 2026 Reviewers agreed at journal 28 Apr, 2026 Reviewers invited by journal 28 Apr, 2026 Editor assigned by journal 28 Apr, 2026 Editor invited by journal 16 Apr, 2026 Submission checks completed at journal 14 Apr, 2026 First submitted to journal 14 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9371068","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":635480358,"identity":"4866ec74-382f-45fb-8d47-1cec568018c4","order_by":0,"name":"Souleiman El Balkhi","email":"data:image/png;base64,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","orcid":"","institution":"P\u0026T, UMR1248, Inserm, Univ. Limoges","correspondingAuthor":true,"prefix":"","firstName":"Souleiman","middleName":"El","lastName":"Balkhi","suffix":""},{"id":635480359,"identity":"a8b20022-e80a-4358-bb1c-8095dcd882aa","order_by":1,"name":"Racym Berrah","email":"","orcid":"","institution":"P\u0026T, UMR1248, Inserm, Univ. Limoges","correspondingAuthor":false,"prefix":"","firstName":"Racym","middleName":"","lastName":"Berrah","suffix":""},{"id":635480360,"identity":"bb4d16b8-6211-46ae-86ee-75113a112e2a","order_by":2,"name":"François Ludovic Sauvage","email":"","orcid":"","institution":"P\u0026T, UMR1248, Inserm, Univ. Limoges","correspondingAuthor":false,"prefix":"","firstName":"François","middleName":"Ludovic","lastName":"Sauvage","suffix":""},{"id":635480361,"identity":"52203947-bc22-43a7-a6a3-48fffac327a3","order_by":3,"name":"Léa Le Du","email":"","orcid":"","institution":"University of Limoges","correspondingAuthor":false,"prefix":"","firstName":"Léa","middleName":"Le","lastName":"Du","suffix":""},{"id":635480362,"identity":"ad946722-434f-4403-aa28-13753c771de1","order_by":4,"name":"Mohamad Ali Rahali","email":"","orcid":"","institution":"P\u0026T, UMR1248, Inserm, Univ. Limoges","correspondingAuthor":false,"prefix":"","firstName":"Mohamad","middleName":"Ali","lastName":"Rahali","suffix":""},{"id":635480363,"identity":"71f6d9c1-95af-40a3-a63b-0f0bb1d662ec","order_by":5,"name":"Roy Lakis","email":"","orcid":"","institution":"P\u0026T, UMR1248, Inserm, Univ. Limoges","correspondingAuthor":false,"prefix":"","firstName":"Roy","middleName":"","lastName":"Lakis","suffix":""},{"id":635480364,"identity":"e88bcd46-d449-4b09-8387-6f6e70f48ca5","order_by":6,"name":"Pierre Marquet","email":"","orcid":"","institution":"P\u0026T, UMR1248, Inserm, Univ. Limoges","correspondingAuthor":false,"prefix":"","firstName":"Pierre","middleName":"","lastName":"Marquet","suffix":""},{"id":635480365,"identity":"7641a7a7-cba3-4a38-a943-93394c75165e","order_by":7,"name":"Franck Saint-Marcoux","email":"","orcid":"","institution":"P\u0026T, UMR1248, Inserm, Univ. Limoges","correspondingAuthor":false,"prefix":"","firstName":"Franck","middleName":"","lastName":"Saint-Marcoux","suffix":""},{"id":635480366,"identity":"03a62fe0-8927-4792-9821-eef531da28ec","order_by":8,"name":"Véronique Loustaud-Ratti","email":"","orcid":"","institution":"P\u0026T, UMR1248, Inserm, Univ. Limoges","correspondingAuthor":false,"prefix":"","firstName":"Véronique","middleName":"","lastName":"Loustaud-Ratti","suffix":""},{"id":635480367,"identity":"03801dc8-f1bc-4b0f-b8ba-71df0ce56196","order_by":9,"name":"Paul Carrier","email":"","orcid":"","institution":"P\u0026T, UMR1248, Inserm, Univ. Limoges","correspondingAuthor":false,"prefix":"","firstName":"Paul","middleName":"","lastName":"Carrier","suffix":""}],"badges":[],"createdAt":"2026-04-09 16:39:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9371068/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9371068/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108945552,"identity":"1f970b6d-61c0-4ebd-bbf3-e464b9749e34","added_by":"auto","created_at":"2026-05-11 06:16:00","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":8335715,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAbsolute concentration and diagnostic performance of native human serum albumin across chronic liver disease stages.\u003c/strong\u003e Dot plots showing the absolute concentration (g/L) of native HSA in healthy controls and patients stratified by fibrosis stage (F0, F2, F4A, F4B, F4C) on Platform 1 (left panel) and Platform 2 (right panel). Horizontal bars represent the mean ± SEM. Significant pairwise differences between groups are indicated by brackets (one-way ANOVA followed by Tukey's post-hoc test; *p \u0026lt; 0.05, ****p \u0026lt; 0.0001). Lower panels show receiver operating characteristic (ROC) curves for native HSA discriminating each fibrosis stage from healthy controls on each platform. HSA: human serum albumin; F4A: Child-Pugh A cirrhosis; F4B: Child-Pugh B cirrhosis; F4C: Child-Pugh C cirrhosis.\u003c/p\u003e","description":"","filename":"Fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-9371068/v1/f1b72fbacf49046cf1ec448c.png"},{"id":108978079,"identity":"5be1fb83-06c8-4339-9279-caf6e1d563b8","added_by":"auto","created_at":"2026-05-11 11:33:58","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":12001292,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStage-specific concentrations of modified human serum albumin isoforms across chronic liver disease.\u003c/strong\u003e Dot plots showing absolute concentrations (g/L) of nine modified HSA isoforms in healthy controls and patients stratified by fibrosis stage. \u003cstrong\u003e(A)\u003c/strong\u003e N-terminally truncated albumin (HSA-DA). \u003cstrong\u003e(B)\u003c/strong\u003eC-terminally truncated albumin (HSA-L). \u003cstrong\u003e(C)\u003c/strong\u003eN-terminally truncated and cysteinylated albumin (HSA-DA+CYS). \u003cstrong\u003e(D)\u003c/strong\u003e Sulfonated albumin (HSA+SO₃H). \u003cstrong\u003e(E)\u003c/strong\u003e Cysteinylated albumin (HSA+CYS). \u003cstrong\u003e(F)\u003c/strong\u003eMono-glycated albumin (HSA+GLYC). \u003cstrong\u003e(G)\u003c/strong\u003eCysteinylated and mono-glycated albumin (HSA+CYS+GLYC). \u003cstrong\u003e(H)\u003c/strong\u003e Doubly glycated albumin (HSA+2GLYC). \u003cstrong\u003e(I)\u003c/strong\u003eCysteinylated and doubly glycated albumin (HSA+CYS+2GLYC). Horizontal bars represent the mean ± SEM. Significant pairwise comparisons are indicated (*p \u0026lt; 0.05, **p \u0026lt; 0.01, ***p \u0026lt; 0.001, ****p \u0026lt; 0.0001; one-way ANOVA with Tukey's post-hoc test).\u003c/p\u003e","description":"","filename":"Fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-9371068/v1/12c62f7596195674d884e9cf.png"},{"id":108977961,"identity":"bc1ad83b-0fea-4dc7-814c-6483d9e13f92","added_by":"auto","created_at":"2026-05-11 11:33:31","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":13085544,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDiagnostic performance of normalized HSA isoform ratios for liver disease staging.\u003c/strong\u003e Each row presents, for a given isoform ratio, a dot plot of the ratio values stratified by fibrosis stage (left) and the corresponding receiver operating characteristic (ROC) curves discriminating each stage from healthy controls (right). \u003cstrong\u003e(A)\u003c/strong\u003e Cysteinylated-to-native HSA ratio (HSA+CYS / Native). \u003cstrong\u003e(B)\u003c/strong\u003e Mono-glycated-to-native HSA ratio (HSA+GLYC / Native). \u003cstrong\u003e(C)\u003c/strong\u003e Cysteinylated mono-glycated-to-native HSA ratio (HSA+CYS+GLYC / Native). Horizontal bars represent mean ± SEM. Statistical significance is indicated (*p \u0026lt; 0.05, **p \u0026lt; 0.01, ****p \u0026lt; 0.0001; one-way ANOVA with Tukey's post-hoc test). ROC curves are color-coded by fibrosis stage; the diagonal dashed line represents chance-level discrimination.\u003c/p\u003e","description":"","filename":"Fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-9371068/v1/7aca033fd61156e344e23717.png"},{"id":108978140,"identity":"088bb4c5-dfc7-46c3-8c43-53a1fecce22c","added_by":"auto","created_at":"2026-05-11 11:34:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":125125,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrincipal component analysis of the full HSA isoform spectral profile across chronic liver disease stages.\u003c/strong\u003e PCA was performed on the entire deconvoluted albumin isoform profile spanning m/z 66,000–68,000 Da. Each panel shows the progressive displacement of patient clusters from the control group as fibrosis severity advances, from early fibrosis \u003cstrong\u003e(A, B)\u003c/strong\u003e through compensated cirrhosis \u003cstrong\u003e(C, D)\u003c/strong\u003e to decompensated cirrhosis \u003cstrong\u003e(E, F)\u003c/strong\u003e. Each point represents one patient; colors indicate fibrosis stage. Ellipses represent 95% confidence regions. The progressive separation of clusters in PCA space reflects the stage-dependent evolution of the global albumin molecular fingerprint.\u003c/p\u003e","description":"","filename":"Fig4PCA.png","url":"https://assets-eu.researchsquare.com/files/rs-9371068/v1/1af3add476d25da9cdc62c48.png"},{"id":108945555,"identity":"c9858018-4d2c-479e-b974-0f0e9b9a0d16","added_by":"auto","created_at":"2026-05-11 06:16:00","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":147797,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCross-platform classification performance of the ALBOM OrdinalForest model.\u003c/strong\u003e \u003cstrong\u003e(A, B)\u003c/strong\u003e Confusion matrices for the hierarchical OrdinalForest classifier evaluated on the held-out test set (n = 46) from Platform 1 — Bruker (A) and Platform 2 — Sciex (B). Rows represent true biopsy-confirmed fibrosis classes; columns represent model predictions. Color intensity is proportional to cell frequency. Diagonal cells indicate correct classifications. \u003cstrong\u003e(C)\u003c/strong\u003e Patient-level cross-platform agreement plot. Each point represents one patient from the common test set, with the Bruker prediction on the x-axis and the Sciex prediction on the y-axis. Points are color-coded by true biopsy class. The dashed diagonal indicates perfect inter-platform agreement; off-diagonal points represent discordant predictions. \u003cstrong\u003e(D)\u003c/strong\u003e Bootstrap distribution of the Quadratic Weighted Kappa (QWK) for Platform 1 (blue) and Platform 2 (purple), estimated from 1,000 resampling iterations. Shaded areas show the full distribution; vertical lines mark the 95% confidence interval bounds. QWK: Quadratic Weighted Kappa; F4A: Child-Pugh A cirrhosis; F4B: Child-Pugh B cirrhosis; F4C: Child-Pugh C cirrhosis.\u003c/p\u003e","description":"","filename":"Fig5ABConfusionEN.png","url":"https://assets-eu.researchsquare.com/files/rs-9371068/v1/c22e53e8e6c03a8ca9385c5c.png"},{"id":108945556,"identity":"cbd00c39-1ce9-46ff-9e21-f87a9cfed420","added_by":"auto","created_at":"2026-05-11 06:16:00","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":146423,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePatient triage performance of the ALBOM algorithm versus FIB-4.\u003c/strong\u003e Each point represents one patient from the test set (n = 46), positioned on the x-axis according to their FIB-4 score and on the y-axis according to their biopsy-confirmed fibrosis class. Point color indicates whether the MALDI-based triage classification was correct (green) or incorrect (red). The shaded gray band (FIB-4: 1.30–2.67) delineates the FIB-4 indeterminate zone, within which the index cannot provide clinical guidance. Outside this zone, the MALDI algorithm correctly classified the majority of patients. Overall triage accuracy was 81.5% for the combined HSA–clinical model versus 59.3% for FIB-4. FIB-4: Fibrosis-4 index; F4A: Child-Pugh A cirrhosis; F4B: Child-Pugh B cirrhosis; F4C: Child-Pugh C cirrhosis.\u003c/p\u003e","description":"","filename":"Fig6FIB4TriageEN.png","url":"https://assets-eu.researchsquare.com/files/rs-9371068/v1/d9236cdf254885408db91b38.png"},{"id":108979958,"identity":"4e710798-be2a-4de6-9a3d-9eec552da060","added_by":"auto","created_at":"2026-05-11 12:02:40","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":24619084,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9371068/v1/29c43407-0f0d-43a1-8078-beb271fd49da.pdf"},{"id":108945557,"identity":"4063adfc-74a2-4b3e-a94e-b5f2b4e70d17","added_by":"auto","created_at":"2026-05-11 06:16:01","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2530005,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaldata.docx","url":"https://assets-eu.researchsquare.com/files/rs-9371068/v1/f59241b17f63db72b9015800.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Human Serum Albumin Profiling by Top-down Analysis Enables Multi-Class Liver Fibrosis Staging: A Cross-Platform Validation Study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eChronic liver diseases (CLDs) are a global health burden, contributing significantly to morbidity and mortality worldwide (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). The progression of CLDs through various stages, from initial injury to fibrosis, cirrhosis, and eventual decompensation or the development of hepatocellular carcinoma, underscores the critical need for accurate diagnostic and staging tools (\u003cspan additionalcitationids=\"CR3 CR4 CR5 CR6\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eNon-invasive tools for assessing liver fibrosis and significative portal hypertension have been widely developed in recent years, such as liver and more recently splenic elastometry and non-invasive biological fibrosis tests such as Fibrotest\u003csup\u003eR\u003c/sup\u003e and Fibrometer\u003csup\u003eR\u003c/sup\u003e. However, while they have made a definite contribution in practice, they do not meet the objective of EASL with an accuracy\u0026thinsp;\u0026gt;\u0026thinsp;80%, except for advanced fibrosis diagnosis. Their limitation lies primarily in the difficulty of predicting intermediate fibrosis scores and the time to onset of liver-related events (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). While prognostic scores like the Model for End-Stage Liver Disease (MELD) and Child-Pugh scores are widely used (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e), they also have limitations, and accurately predicting the disease outcomes, especially the transition between compensated and decompensated states or the risk of Acute on Chronic Liver Failure (ACLF) development, remains challenging (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). Improved biomarkers are needed to better stratify patients, monitor disease progression, help therapeutic decisions and interpret treatment response.\u003c/p\u003e \u003cp\u003eHuman serum albumin (HSA), the most abundant plasma protein (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e), is synthesized exclusively by the liver (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). Traditionally recognized for its primary role in maintaining plasma oncotic pressure and modulating fluid distribution (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e), HSA is now understood to possess a wide array of critical non-oncotic biological functions (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan additionalcitationids=\"CR10 CR11 CR12\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). These include binding, transport, and detoxification of numerous endogenous and exogenous substances (such as bilirubin, bile acids, fatty acids, metals, and drugs) (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e), potent antioxidant and free-radical scavenging activities, immunomodulatory effects, and contributions to endothelial stabilization and homeostasis. The structural integrity of the HSA molecule, a 66.5 kDa globular protein organized into three homologous domains, is crucial for these non-oncotic functions (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn healthy individuals, the majority (70\u0026ndash;80%) of circulating HSA exists in the reduced form, known as human mercaptalbumin (HMA) (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e), a smaller fraction (20\u0026ndash;30%) as reversibly oxidized human nonmercaptalbumin-1 (HNA1), and a very minor fraction (\u0026lt;\u0026thinsp;5%) as irreversibly oxidized human nonmercaptalbumin-2 (HNA2)(\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e). However, HSA is susceptible to a variety of other post-transcriptional modifications (PTMs) (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e), especially under conditions of systemic inflammation and oxidative stress characteristic of advanced CLDs (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan additionalcitationids=\"CR11\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). The accumulation of these modified HSA molecules, or proteoforms, constitutes the microheterogeneity of circulating albumin (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn patients with CLDs, particularly those with decompensated cirrhosis and ACLF, significant alterations in circulating HSA are observed, encompassing both quantitative reduction (hypoalbuminemia) and profound qualitative changes in its structure and function (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). Numerous studies utilizing high-performance liquid chromatography coupled mass spectrometry have documented extensive PTMs in HSA from cirrhotic patients (\u003cspan additionalcitationids=\"CR3\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e) with a significant decrease in the proportion of native, structurally intact HMA and a corresponding increase in oxidized forms, HNA1 and particularly the irreversibly oxidized HNA2. Specific isoforms, such as cysteinylated, glycated, N-terminal truncated, C-terminal truncated, sulfinylated forms, as well as HSA homodimers, have been identified and found to be significantly more abundant in patients with decompensated cirrhosis compared to healthy controls or patients with compensated disease.\u003c/p\u003e \u003cp\u003eCrucially, these qualitative changes impair HSA's non-oncotic functions and reduce binding capacity for various ligands (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e), diminish antioxidant potential (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e), and alter transport efficiency (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e) of HSA in advanced liver disease (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). This functional impairment has led to the concept of \"effective albumin concentration,\" suggesting that the overall biological function of albumin relates not just to its total concentration but also significantly to the preservation of its structural and functional integrity. Indeed, several studies have highlighted the prognostic significance of specific HSA isoforms: the level of native HMA, the irreversibly oxidized HNA2 fraction, the ischemia-modified albumin ratio (IMAR), certain cysteinylated/truncated isoforms, and specific homodimers might be independent predictors of CLD complications and short-term mortality in hospitalized patients with cirrhosis, suggesting their superiority compared to total serum albumin concentration (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eHowever, the utility of specific HSA isoforms as diagnostic markers for differentiating stages of CLD has not been systematically investigated. Most studies have focused on patients with established decompensation or ACLF, often comparing them to healthy controls or less well-defined cirrhotic groups (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan additionalcitationids=\"CR5\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). Furthermore, while various analytical methods have been employed (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e), a standardized approach to profile and quantify a comprehensive panel of relevant HSA isoforms for diagnostic staging purposes is lacking. There is therefore a clear need: (i) to systematically determine whether the distinctive patterns of HSA isoform distribution can reliably discriminate CLD stages using a comprehensive quantitative approach; and (ii) to establish whether such a diagnostic signature is sufficiently robust to be reproducible across different analytical platforms, a prerequisite for multi-center clinical implementation.\u003c/p\u003e \u003cp\u003eTherefore, this study pursued two co-primary objectives. First, to systematically characterize the complete profile of circulating HSA isoforms using LC-HR-MS across well-defined cohorts representing the full spectrum of chronic liver disease, including pre-cirrhotic fibrosis, compensated cirrhosis, and decompensated cirrhosis, compared to healthy controls. The primary aim was to evaluate the diagnostic relevance of individual and combined HSA isoforms in differentiating CLD stages, and to determine whether their multivariate signature provides superior performance to available non-invasive tools. Second, to validate the analytical reproducibility and instrument-independence of the ALBOM signature by applying the same classification model to samples analysed on two independent LC-HR-MS platforms from distinct manufacturers, and to quantify the degree of cross-platform agreement using formal method comparison statistics.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eBetween January 2021 and January 2023, 172 patients diagnosed with CLD at various stages and 82 control subjects met the initial eligibility criteria.\u003c/p\u003e \u003cp\u003eDetailed demographic and clinical characteristics of the patient cohort, stratified by liver fibrosis stage and cirrhosis severity, are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The patient group consisted predominantly of males (116/172, 67.4%), with a median age at inclusion of 61 years.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003ePopulation\u0026rsquo;s characteristics\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cb\u003eNumber of patients\u003c/b\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eF0/F1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eF2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eF3\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eChild Pugh A\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eChild Pugh B\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eChild Pugh C\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36(21%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e23(13%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e30(17%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e37(22%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e26(15%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e20(12%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003e172\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eGender\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e56\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e116\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge (year)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAverage age\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e59\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e60\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedian age\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e53\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e61\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEtiology\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMASH\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e62 (36%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e40(23,2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHBV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e15(8,7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHCV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e11(6,3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHemochromatosis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3(1,7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCardiac\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1(0,6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAIH\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e11(6,3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAutoimmune cholangitis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3(1,7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePBC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3(1,7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePSC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2(1,1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCrytogenic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e4(2,3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMixed\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e17(9,9%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMixed etiology\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol\u0026thinsp;+\u0026thinsp;MASH\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e11(6,3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePBC\u0026thinsp;+\u0026thinsp;MASH\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2(1,1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol\u0026thinsp;+\u0026thinsp;HBV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1(0,6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol\u0026thinsp;+\u0026thinsp;HCV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2(1,1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAverage MELD\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eHepatic encephalopathy\u003c/b\u003e\u003c/p\u003e \u003cp\u003eAll stages\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModerate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDisabling\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAscite\u003c/b\u003e\u003c/p\u003e \u003cp\u003eAll stages\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e30\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAverage abundance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTense or refractory\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003eNS: Not suitable, MASH: Metabolic associated steatohepatitis, AIH: Autoimmune hepatitis, HCV: Hepatitis C virus, HBV: Hepatitis B virus, PBC: primary biliary cholangitis, PSC: primary sclerosing cholangitis, MELD: Model for end stage liver disease.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe distribution across fibrosis stages was as follows: 36 (20.9%) F0/F1, 23 (13%) F2, 30 (17%) F3, and 83 (48%) F4 (cirrhosis). Fibrosis staging was mainly based on transient elastography.\u003c/p\u003e \u003cp\u003eAmong the 83 patients with cirrhosis (F4), 37 (44.6% of F4; 21.5% of total cohort) were classified as Child-Pugh Class A, 26 (31.3% of F4; 15.1% of total) as Class B, and 20 (24.1% of F4; 11.6% of total) as Class C. The Meld score averaged 15 for Child Pugh B and 21 for Child Pugh C patients respectively (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe primary etiologies of CLD were non-alcoholic steatohepatitis (MASH; 62 patients, 36.0%) and alcohol-related liver disease (ALD; 40 patients, 23.3%). A combination of ALD and MASH was identified in 11 patients (6.4%). Other etiologies included viral hepatitis B (HBV; 15 patients, 8.7%), viral hepatitis C (HCV; 11 patients, 6.4%), autoimmune hepatitis (AIH; 11 patients, 6.4%), and less common causes such as hemochromatosis, primary biliary cholangitis (PBC), primary sclerosing cholangitis (PSC), cardiac hepatopathy, and cryptogenic cirrhosis (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Ascites was present in 30 of the 83 (36.1%) individuals with cirrhosis. It was graded as moderate in 18 patients and was tense or refractory to diuretics in 12 patients. Hepatic encephalopathy (West Haven grade II or III) was documented in 18 (21.7%) cirrhotic patients. Further details on the distribution of clinical complications according to Child-Pugh score are provided in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u0026ndash; Diagnosis methods and Child-Pugh classification for F4 patients.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFibrosis stage\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFibroscan\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLiver biopsy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFibroscan\u0026thinsp;+\u0026thinsp;liver biopsy\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eF0/F1\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e35 (0.97)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4 (0.11)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3 (0.08)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eF2\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23 (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5 (0.22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5 (0.22)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eF3\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e28 (0.93)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e12 (0.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e10 (0.33)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eF4\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e32 (0.45)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e19 (0.23)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4 (0.05)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eF4_Child Pugh A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e25 (0.68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5 (0.13)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2 (0.05)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eF4_Child Pugh B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3 (0.11)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e6 (0.23)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1 (0.04)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eF4_Child Pugh C\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8 (0.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1 (0.05)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eHSA isoforms profiling across liver disease stages\u003c/h2\u003e \u003cp\u003eThe absolute concentrations of native HSA and its major isoforms, stratified by fibrosis stage and Child-Pugh class for cirrhotic patients are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe average value of the native albumin isoform (native HSA) in healthy controls was 12.2 g/L. Native HSA was slightly lower but not significantly in patients with F0/F1 fibrosis (10.6 g/L), F2 (9.9 g/L), and F3 (10.6 g/L), while it was significantly reduced in patients diagnosed with Child-Pugh A cirrhosis (F4_A: 10.2 g/L), Child-Pugh B cirrhosis (F4_B: 4.1 g/L), and Child-Pugh C cirrhosis (F4_C: 4.2 g/L) compared to healthy controls (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). However, by combining F2 and F3 (F2/F3), a discrimination between F2/F3 and the control group was observed (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). A significant difference was found between the F4_B/F4_C groups compared to the other groups. It is important to note that F4_A was distinguished from F4_B. However, no significant difference was detected between F4_B and F4_C.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eConcentrations of other modified HSA isoforms\u003c/h3\u003e\n\u003cp\u003eIn total, up 10 non-native albumin isoforms were quantified in most patients, including N-terminally truncated (HSA-DA) and C-terminally truncated (HSA-L) albumin, cysteinylated (HSA\u0026thinsp;+\u0026thinsp;CYS), oxidized (HSA+SO\u003csub\u003e3\u003c/sub\u003eH), and glycated (HSA+Glyc) forms, as well as combination isoforms such as glycated cysteinylated (HSA\u0026thinsp;+\u0026thinsp;CYS+GLYC) and truncated cysteinylated (HSA-DA\u0026thinsp;+\u0026thinsp;CYS). We compared the mean concentration (g/L) of these albumin isoforms according to the stage of cirrhosis (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe isoforms revealed three distinct patterns of evolution corresponding to the progression of liver disease. First, several isoforms involved in moderate oxidative stress and glycation exhibited a biphasic trajectory. Specifically, the concentrations of cysteinylated (HSA\u0026thinsp;+\u0026thinsp;CYS), singly glycated (HSA+GLYC), and doubly glycated (HSA+2GLYC) isoforms generally increased from the control state through the stages of compensated cirrhosis. The mean concentration of HSA\u0026thinsp;+\u0026thinsp;Cys, for example, rose from 8.7 g/L in controls to a peak of 11.1 g/L in patients with Child-Pugh A cirrhosis. However, as the disease progressed into severe decompensation, the concentrations of these isoforms paradoxically declined, with HSA\u0026thinsp;+\u0026thinsp;CYS falling to 7.9 g/L and 6.8 g/L in Child-Pugh B and C patients, respectively (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.E, 2.F, 2.H).\u003c/p\u003e \u003cp\u003eIn contrast, the more complex, multiply-modified cysteinylated and doubly glycosylated isoform (HSA\u0026thinsp;+\u0026thinsp;CYS+2GLYC) showed a progressive and significant increase with advancing cirrhosis. Its concentration was markedly higher in Child-Pugh A (0.12 g/L), B (0.18 g/L), and C (0.2 g/L) stages compared to almost undetectable levels in healthy controls, suggesting it serves as a marker of cumulative, end-stage protein damage (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.I).\u003c/p\u003e \u003cp\u003eFinally, a third group of isoforms, primarily those involving truncation or irreversible oxidation, showed a general trend of reduction, particularly in the most advanced disease stages. The N-terminally truncated isoform (HSA-DA) was significantly lower across all disease stages compared to its level in controls (0.2 g/L). Similarly, the irreversibly oxidized (HSA+SO\u003csub\u003e3\u003c/sub\u003eH) and the combined truncated-cysteinylated (HSA-DA\u0026thinsp;+\u0026thinsp;CYS) isoforms were notably decreased in patients with Child-Pugh B and C cirrhosis (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.A, 2.C, 2.D). The C-terminally truncated isoform (HSA-L) showed no significant variation across the disease spectrum (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.B).\u003c/p\u003e\n\u003ch3\u003eDiscriminatory behavior of HSA isoforms\u003c/h3\u003e\n\u003cp\u003eTo better assess the discriminant capacities of HSA isoforms, we normalized the average concentration of some clinically significant isoforms. The ratios of cysteinylated, glycated, and cysteinylated-glycated isoforms to native albumin (HSA\u0026thinsp;+\u0026thinsp;CYS/Native, HSA+GLYC/Native, and HSA\u0026thinsp;+\u0026thinsp;CYS+GLYC/Native) showed a marked and statistically significant increase with the evolution of the CLD stage. Notably, a significant increase in the concentration ratio of these isoforms was found in patients with decompensated cirrhosis B and C, whereas the concentration remained comparable in the earlier fibrotic stages (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). The glycated isoform was the only one capable of discriminating between the control group and F2 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA.2). Moreover, the GLYC and CYS+GLYC isoforms were able to distinguish between F2 and F4_B (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA.2 and 3A.3), which was not the case with the cysteinylated isoform (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA.1). We analysed the discriminant potential of these specific isoforms between the control group and the different fibrotic stages using ROC curves. For the cysteinylated isoform, a discrimination with a sensitivity of 65% and a specificity of 99% between the F4_C group of 20 patients and the control group was observed. For the glycated and cysteinylated-glycated isoforms, better discrimination of the F4_B and the F4_C group was observed. For the GLYC isoform, a diagnostic sensitivity of 85% and a specificity of 100% were observed for F4_B versus a sensitivity of 70% and a specificity of 99% for F4_C. For the CYS+GLYC isoform, a diagnostic sensitivity of 77% and a specificity of 99% was observed for F4_B versus a sensitivity of 70% and a specificity of 99% for F4_C (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn comparison, analysis of classical laboratory parameters confirmed the progressive deterioration of liver function across the patient cohorts. Markers of hepatic synthesis and excretion, such as serum albumin and bilirubin, remained stable through early fibrosis and compensated cirrhosis (Child-Pugh A). However, they showed significant deterioration in advanced decompensation, with albumin levels progressively decreasing while bilirubin markedly increased in Child-Pugh B and C stages. In parallel, the FIB-4 index and AST levels demonstrated a clear, stepwise increase corresponding to advancing disease severity, while ALT lacked discriminatory power (Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eHowever, no single isoform was sufficient to discriminate all six CLD stages simultaneously, motivating the development of a multivariate classification approach incorporating the full spectral profile.\u003c/p\u003e\n\u003ch3\u003eCross-Platform Validation\u003c/h3\u003e\n\u003cp\u003eTo evaluate whether the diagnostic signature encoded in the albumin spectral profile is platform-independent, classification models were independently trained and evaluated on samples acquired on two distinct LC-HR-MS instruments. Despite differences in instrument architecture, software-based baseline correction algorithms, and spectral resolution between Platform 1 and Platform 2, the QWK of both classifiers fell within the range defined as 'substantial to near-perfect agreement', and their 95% bootstrap confidence intervals showed substantial overlap (Platform 1: [0.735\u0026ndash;0.923]; Platform 2: [0.822\u0026ndash;0.964]; Figure S4).Cross-platform equivalence was formally confirmed by McNemar's test applied to paired patient predictions (p\u0026thinsp;=\u0026thinsp;0.149), indicating no statistically significant difference in classification decisions between platforms. The Jaccard Similarity Index of prediction errors between platforms was 0.696, meaning that approximately 70% of misclassified patients were identically misclassified by both instruments (Figure S5). This convergence of errors is a key finding: it indicates that the sources of misclassification reside in patient-level biological ambiguity at transitional fibrosis stages, particularly at the F2/F4A and F4A/F4B boundaries, rather than in instrument-specific noise or systematic bias. The slightly lower performance of Platform 1 compared to Platform 2 is consistent with a technical attenuation of high-mass albumin peaks (\u0026gt;\u0026thinsp;67,500 Da, corresponding to poly-glycated adducts) by the Platform 1 baseline correction algorithm, which geometrically misidentifies these broad, disease-specific peaks as baseline drift, partially suppressing the diagnostic signal in the F4B and F4C classes. This observation does not affect the validity of the Platform 1 classifier but provides an actionable optimization target for future implementations.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study provides the first comprehensive, quantitative analysis of a wide panel of human serum albumin (HSA) isoforms across the full spectrum of CLD, from early-stage fibrosis to end-stage decompensated cirrhosis. While previous research has established that structural and functional alterations of HSA occur in advanced liver disease, these investigations have predominantly focused on patients with advanced liver disease and have primarily investigated prognostic, rather than diagnostic, value. By employing a robust and sensitive top-down liquid chromatography high-resolution mass spectrometry (LC-HR-MS) method applied across two independent instrument platforms (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e), our work moves beyond relative measurements to map the dynamic, stage-specific changes in the HSA isoforms landscape. This approach offers a novel \"molecular staging\" system that reflects the underlying pathophysiology of CLD progression, providing a more granular view than is achievable with conventional biomarkers or elastography alone.\u003c/p\u003e \u003cp\u003eOur findings both confirm and substantially extend the existing literature. The observed progressive decline in native HSA concentration with increasing disease severity, particularly the precipitous drop in patients with Child-Pugh class B and C cirrhosis, is in strong agreement with the concept of a diminishing \"effective albumin concentration\" in advanced liver disease (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). This loss of the structurally and functionally intact HSA pool is a key pathophysiological feature, as native albumin is critical for mitigating the systemic inflammation and oxidative stress that drive disease progression (\u003cspan additionalcitationids=\"CR5\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). However, our data clarify that while native HSA is a reliable biomarker for predicting and identifying decompensated cirrhosis (Child-Pugh B and C), its utility for diagnosing the initial stages of fibrosis is limited, showing low sensitivity and specificity in these earlier phases.\u003c/p\u003e \u003cp\u003eThe analysis of modified isoforms reveals a highly dynamic and nuanced picture. For instance, the concentrations of cysteinylated (HSA\u0026thinsp;+\u0026thinsp;Cys) and glycated (HSA+Glyc) isoforms exhibit a biphasic pattern: they increase during the transition from a healthy state to compensated cirrhosis (F4_A), likely reflecting the escalating systemic oxidative and glycative stress, but subsequently decline in decompensated cirrhosis (F4_B and F4_C). This is a novel and critical observation that suggests a complex interplay of factors in end-stage disease. The decline in the absolute concentration of these modified forms may be due to the profound depletion of the native HSA as the substrate for modification, an accelerated clearance of these moderately modified isoforms, or their conversion into more complex, multiply-modified species, such as the doubly- and triply-modified isoforms (e.g., HSA\u0026thinsp;+\u0026thinsp;CYS+2GLYC) that we observed to increase in the most advanced stages. This apparent paradox, where the absolute concentration of a modified isoform decreases while the disease worsens, is resolved when considering the isoform ratios. The use of ratios, such as (HSA\u0026thinsp;+\u0026thinsp;Cys)/Native, normalizes for the overall drop in albumin synthesis and directly reflects the escalating proportion of damaged albumin, thus powerfully amplifying the diagnostic signal for advanced disease. This underscores that the dynamic flux of specific isoforms offers a far richer dataset for staging than a simple monotonic change in a single biomarker.\u003c/p\u003e \u003cp\u003eThe distinct behavior of individual isoforms suggests specific diagnostic utilities. The finding that the N-terminally truncated isoform (HSA-DA) was significantly reduced across all disease stages compared to controls is an unexpected result, as truncation is often considered a damage product. This may suggest that either the control population has a higher baseline for this specific proteoform, or that HSA-DA is subject to accelerated clearance or further modification even in early stages of liver injury, a hypothesis that warrants further investigation. Furthermore, the ROC curve analysis revealed that while cysteinylated isoforms were highly effective at distinguishing the most severe stage (F4_C) from controls, glycated and cysteinylated-glycated isoforms appeared superior in discriminating the transition to Child-Pugh B cirrhosis. However, these observations also make it clear that no single isoform is sufficient to accurately classify all disease stages. This highlights that HSA must be considered as an interdependent network of isoforms, where the concentration of one isoform is influenced by the availability and flux of others.\u003c/p\u003e \u003cp\u003eConsequently, the true strength of this approach lies in the integrative analysis of the entire HSA isoform profile. While individual isoforms or classical biomarkers like FIB-4 (Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e and S2 in supplemental data) can distinguish between distant stages (e.g., F0 vs. F4_C), they often fail to provide clear separation between contiguous intermediate stages. Our principal component analysis (PCA) compellingly demonstrates that the global HSA isoforms signature creates a distinct and progressively shifting pattern that maps directly to disease progression.\u003c/p\u003e \u003cp\u003eThe OrdinalForest classification model, trained on 75 spectral features within the albumin m/z 66,000\u0026ndash;68,000 Da region and four routine clinical variables, achieved QWK values of 0.862 and 0.916 on two independent platforms, a performance levels that substantially exceed what is achievable by FIB-4 alone in multiclass staging (QWK\u0026thinsp;=\u0026thinsp;0.188\u0026ndash;0.229). The feature importance analysis revealed that while routine clinical variables (total protein, albumin, INR, bilirubin) contributed substantially to the combined model, spectral albumin peaks representing the cysteinylated and glycated isoform regions (m/z ∶66,230\u0026ndash;66,600 Da) ranked among the top instrumental predictors, validating the biological rationale for our isoform profiling approach. The model's lower accuracy for the F4A class, consistently misclassified into adjacent stages, reaffirms the biological continuum argument: Child-Pugh A cirrhosis represents a transition state where the albumin molecular fingerprint overlaps significantly with advanced fibrosis (F2/F3) and this ambiguity is not a model artifact, but a reflection of the underlying pathophysiology.\u003c/p\u003e \u003cp\u003eA key novel contribution of this work is the formal demonstration that the HSA spectral signature is reproducible across LC-HR-MS platforms from distinct manufacturers. The statistical equivalence of classification decisions (McNemar p\u0026thinsp;=\u0026thinsp;0.149) and the high proportion of shared errors (Jaccard index 0.696) collectively argue that the diagnostic information is encoded in the biology of the sample \u0026mdash; specifically in the molecular modifications of albumin \u0026mdash; rather than in instrument-specific signal characteristics. This finding is essential for clinical translation: in a multi-center deployment scenario, patients would inevitably be analyzed on different platforms across participating institutions. Our data demonstrate that the our classifier would deliver consistent results regardless of whether a Bruker or Sciex instrument is used, provided that the same preprocessing pipeline is applied. This places our approach within the formal framework of method comparison studies, a standard required by clinical laboratory accreditation bodies. The slight performance advantage of Platform 2 over Platform 1 appears to arise from a technical difference in baseline correction algorithms rather than from a fundamental biological or clinical distinction and can likely be eliminated by optimizing peak detection parameters for the albumin high-mass region in Platform 1 implementations.\u003c/p\u003e \u003cp\u003eThe search for reliable non-invasive biomarkers of liver injury is a central goal in hepatology, and it is important to contextualize our findings. While markers of hepatocellular necrosis like glutamate dehydrogenase (GLDH) or keratin-18 (K18) can indicate acute hepatocyte damage, they often lack specificity and may not accurately reflect the chronic, cumulative processes of fibrosis (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e). Similarly, microRNA-122 is a highly sensitive marker of liver injury but can be influenced by etiology and shows considerable inter-individual variability (\u003cspan additionalcitationids=\"CR32 CR33\" citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e). In contrast, the HSA isoforms profile offers a unique advantage. As HSA is synthesized exclusively by the liver and has a long half-life, its modified forms represent an integrated, cumulative record of the systemic metabolic and inflammatory environment over weeks to months, making it an ideal candidate biomarker for a chronic, progressive disease. Other second-line tests with a combination of different biomarkers such as FibroTest\u003csup\u003eR\u003c/sup\u003e, FibroMeter\u003csup\u003eR\u003c/sup\u003e are often used but even if more performant, they also demonstrated limitations (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). In the present study, the \u0026lsquo;LC-TOF\u0026thinsp;+\u0026thinsp;Clinical\u0026rsquo; model outperformed FIB-4 by 26 percentage points in 3-class triage accuracy (81.5% vs. 59.3%), and provided actionable classification in 62.5% of patients falling within the FIB-4 indeterminate gray zone (FIB-4: 1.30\u0026ndash;2.67), a population for whom current non-invasive tools systematically fail to provide guidance\u003c/p\u003e \u003cp\u003eThese findings invite a re-evaluation of current paradigms in fibrosis classification. For decades, the METAVIR score from liver biopsy has been the gold standard, but its invasive nature and susceptibility to sampling error are significant drawbacks (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e). Non-invasive methods like transient elastography have revolutionized clinical practice but measure a physical property (stiffness) that is an indirect and sometimes confounded surrogate for the complex biological activity of inflammation and fibrosis (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). Our data suggest that HSA isoforms profiling provides a direct window into the systemic biochemical consequences of liver disease, reflecting the cumulative impact of oxidative stress, inflammation, and metabolic dysregulation. This shifts the diagnostic paradigm from a structural or anatomical classification to a functional and molecular one. The ultimate goal of such a biomarker is not merely to stratify patients into existing categories but to provide predictive information (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e). A specific HSA isoforms signature may, in the future, prove more effective at predicting the risk of fibrosis progression, clinical decompensation, response to therapy, or development of hepatocellular carcinoma than a simple fibrosis stage, thereby enabling a move towards a more proactive and personalized management of CLD.\u003c/p\u003e \u003cp\u003eThis study is not without limitations. First, its cross-sectional design allows for the characterization of stage-specific differences but does not permit the analysis of intra-individual changes over time; a longitudinal, multicentric study (MALAHBAR; NCT06318949) is now underway to confirm these findings and establish their predictive capacity. Second, fibrosis might have been misclassified, particularly for the F2 and F3 stages. The assessment of fibrosis stages has been mainly based on the Fibroscan\u003csup\u003eR\u003c/sup\u003e, whose limitations are known for deciding on intermediate scores F2 and F3. The dispersion of our data points in these groups likely reflects the known limitations of transient elastography (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). Even if the liver biopsy remains the gold standard, it can also be flawed depending on the size of the sample and the heterogeneity of the distribution of lesions in the liver. Third, a selection bias exists within our cohort, as the distribution of etiologies (mainly MASH and ALD) was not uniform across all disease severity groups, which could be a potential confounder. Future studies should address etiology-specific isoforms profiles. Fourth, while the demonstration of cross-platform reproducibility presented here represents a significant step toward clinical implementation, several analytical challenges remain. Specifically, the impact of different baseline correction algorithms between instruments on the quantification of high-mass glycated albumin species (\u0026gt;\u0026thinsp;67,500 Da) must be systematically characterized and controlled, and a formal proficiency testing scheme across participating laboratories will be required prior to multi-center deployment. Finally, the gray zone FIB-4 analysis (n\u0026thinsp;=\u0026thinsp;8 patients) is preliminary and should be considered hypothesis-generating only.\u003c/p\u003e \u003cp\u003eIn conclusion, this study demonstrates that the HSA isoforms profile is dynamically and profoundly altered throughout the progression of chronic liver disease. We have established that an integrated, multivariate assessment of this \"molecular fingerprint\" provides a more biologically coherent and potentially more clinically useful measure of this disease severity than traditional methods or the analysis of single isoforms. These findings lay the groundwork for developing a new class of biomarkers aiming not just at diagnosing the present state of the liver, but at predicting its future course.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eStudy design and patients\u003c/h2\u003e \u003cp\u003eThis prospective, single-centre study was conducted between January 2021 and January 2023, enrolling patients with CLDs at various fibrosis stages who were managed at the Department of Hepatology, Limoges University Hospital, France. For each enrolled participant, demographic data (age, sex), etiology of liver disease, fibrosis stage, Child-Pugh score for cirrhosis, Model for End-Stage Liver Disease (MELD) score were prospectively collected.\u003c/p\u003e \u003cp\u003eInclusion criteria were: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) age\u0026thinsp;\u0026gt;\u0026thinsp;18 years; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) confirmed diagnosis of CLD; and (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) availability of a routine blood sample collected within the preceding 24 hours of inclusion. Exclusion criteria comprised: patients undergoing dialysis or with a history oftransplantation; administration of contrast media, chelation drugs, blood transfusion, blood derivatives, or albumin infusion within the month prior to enrolment to rule out potential interference with albumin isoform analysis, specifically the Serum Enhanced Binding test (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe control group consisted of healthy individuals recruited during the same period, defined by the absence of clinical evidence of liver dysfunction and liver function tests aspartate aminotransferase [AST], alanine aminotransferase [ALT], alkaline phosphatase [ALP], gamma-glutamyl transferase [GGT], total and conjugated bilirubin, within normal reference ranges.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eClinical and biochemical assessments\u003c/h2\u003e \u003cp\u003eRoutine liver function tests, coagulation parameters, and complete blood counts, were recorded from the patients' medical records at the time of sample collection. HSA isoforms were analyzed on leftovers of plasma samples obtained from routine blood collections drawn into lithium heparin Vacutainer\u0026reg; tubes (Becton Dickinson). Shortly after collection, samples were centrifuged at 3000 rpm (1500 \u003cem\u003eg\u003c/em\u003e) for 10 minutes and plasma aliquots were stored at -20\u0026deg;C until analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eLiver fibrosis and cirrhosis staging\u003c/h2\u003e \u003cp\u003ePatients were categorized into liver fibrosis stages F0/F1, F2, F3 and F4 based on a hierarchical approach incorporating non-invasive and invasive methods. The primary method for fibrosis staging was transient elastography (TE), performed by an experienced examinator using the FibroScan\u0026reg; 502 Touch (Echosens, Paris, France) with the M and XL probes. Liver stiffness measurements (LSM) were considered reliable if a minimum of 10 successful acquisitions were achieved, with an interquartile range to median (IQR/M) ratio of less than 30%. Fibrosis stage was subsequently estimated using established LSM cut-off values according to the underlying liver disease etiology (\u003cspan additionalcitationids=\"CR23 CR24 CR25\" citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e). Globally, the cut-offs used were \u0026lt;\u0026thinsp;7.0 kPa for F0/F1, 7.1\u0026ndash;9.5 kPa for F2, 9.6\u0026ndash;13.0 kPa for F3, and \u0026gt;\u0026thinsp;13.1 kPa for F4 (cirrhosis).\u003c/p\u003e \u003cp\u003e For the purposes of the machine learning classification model and cross-platform analysis, patients originally staged as F3 were pooled with the F2 group, given the recognized overlap and limited reproducibility of intermediate fibrosis staging by transient elastography, and consistent with the approach recommended in EASL 2021 guidelines for non-invasive test development.\u003c/p\u003e \u003cp\u003eWhen clinically indicated and if it was available within 12 months of study inclusion, liver biopsy specimens were reviewed by an experienced pathologist and histopathological staging of fibrosis performed according to the METAVIR scoring system (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn cases where LSM or biopsy was not feasible or indicated in the patient for the diagnosis of cirrhosis, we established a consensus based on the combination of: (i) clinical signs, including a firm or nodular liver on palpation and evidence of portal hypertension and/or hepatocellular insufficiency (ii) biochemical markers, including but not limited to decreased prothrombin time or Factor V levels, hypoalbuminemia, hyperbilirubinemia, and thrombocytopenia; and (iii) classical morphological features on imaging, such as liver surface nodularity, parenchymal heterogeneity, and signs of portal hypertension (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eHuman serum albumin isoform quantification\u003c/h2\u003e \u003cp\u003eAbsolute quantification of HSA isoforms was performed using a previously validated and published top-down liquid chromatography-quadrupole time-of-flight mass spectrometry (LC-QTOF-MS) method (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e). Briefly, the method involves a simple 1:50 (v/v) dilution of 20 \u0026micro;L patient serum in 0.9% NaCl after spiking with equine myoglobin (Mb, 4 g/L final concentration in the diluted sample before injection, corresponding to 0.08 g/L in the original serum) as an internal standard (IS) for mass recalibration and quantification.\u003c/p\u003e \u003cp\u003eSample preparation, LC-MS analysis conditions (including C4 column chromatography, gradient elution, ESI-QTOF settings), data acquisition, and data processing involving spectral deconvolution (mass range 66,000\u0026ndash;68,000 Da) and mass recalibration using the Mb IS were performed exactly as described previously (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAll analyses were performed using the same validated top-down LC-HR-MS method. Two independent instrument platforms from distinct manufacturers were used across the study cohorts: Platform 1 (Bruker timsTOF Pro2) and Platform 2 (Sciex TripleTOF 5600+). Both platforms were operated under identical chromatographic and ionization conditions. Raw spectra were processed using platform-specific deconvolution software prior to normalization. Data from both platforms were subsequently subjected to identical preprocessing (Total Ion Current normalization followed by Probabilistic Quotient Normalization), feature selection, and classification procedures, as described below.\u003c/p\u003e \u003cp\u003eFor specific high-molecular-weight glycated isoforms (e.g., HSA\u0026thinsp;+\u0026thinsp;2Glyc, HSA\u0026thinsp;+\u0026thinsp;Cys\u0026thinsp;+\u0026thinsp;2Glyc) potentially present at low abundance or absent in the commercial HSAc standard used for primary calibration, their concentrations were estimated. This estimation was based on applying the calibration slope derived from the most abundant, structurally related glycated isoform quantified in the HSAc standard (e.g., the HSA\u0026thinsp;+\u0026thinsp;Glyc slope was used for HSA\u0026thinsp;+\u0026thinsp;2Glyc), assuming comparable ionization efficiencies between closely related isoforms, as previously demonstrated (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eSpectral Preprocessing and Feature Selection\u003c/h2\u003e \u003cp\u003eFor the multivariate classification analysis, the full albumin spectral region (m/z 66,000\u0026ndash;68,000 Da) was used. Variables with more than 20% missing values were excluded. Spectra were normalized sequentially using Total Ion Current (TIC) normalization to account for instrument-to-instrument intensity differences, followed by Probabilistic Quotient Normalization (PQN) to correct for dilution effects. Feature selection was performed exclusively within the training partition to prevent data leakage. A Random Forest model was first fitted to the full normalized feature matrix, and feature importance was estimated by permutation. The optimal number of features (k) was determined by 5-fold cross-validated Quadratic Weighted Kappa (QWK) over a grid of k values (20, 30, 40, 50, 75, 100). The value k\u0026thinsp;=\u0026thinsp;75 spectral variables was selected as providing the highest cross-validated QWK without overfitting. Standard routine clinical variables (total protein, serum albumin, prothrombin time expressed as INR, and total bilirubin) were incorporated into a combined model (LCTOF+ Clinical) to evaluate their incremental contribution.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eClassification Model\u003c/h2\u003e \u003cp\u003eThree ordinal classification architectures were evaluated: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) a standard Forest (RF) with ordinal outcome treated as a factor; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) a Hierarchical Random Forest (HRF) using a sequential Random binary decomposition of the six fibrosis classes; and (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) a Hierarchical OrdinalForest (HOF), which explicitly incorporates the ordinal structure of the class labels into the optimization criterion. All models were trained on an 80% stratified training split (stratified by fibrosis class) and evaluated on the held out 20% test set. Primary performance metric was the Quadratic Weighted Kappa (QWK), which penalizes predictions proportionally to their ordinal distance from the true class, making it the appropriate metric for this staged-disease classification. Balanced accuracy was reported as a secondary metric to account for class imbalance. All analyses were performed in R version 4.4.2 using the ordinalForest, ranger, and yardstick packages.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eCross-Platform Validation\u003c/h2\u003e \u003cp\u003eTo assess the reproducibility of the ALBOM diagnostic signature across LC-HR-MS platforms, an independent classification model was trained and evaluated on samples analyzed on each platform separately. The same preprocessing, feature selection, and classification pipeline was applied to both datasets (Bruker and Sciex) independently. Cross-platform equivalence was assessed using two complementary approaches. First, McNemar\u0026rsquo;s test was applied to paired predictions from patients whose samples were analyzed on both instruments, testing whether the two classifiers produced statistically different classification decisions. Second, the Jaccard Similarity Index (JSI) was calculated on the error matrices of both platforms to quantify the proportion of shared misclassifications: JSI = |errors_FR \u0026cap; errors_AL| / |errors_FR \u0026cup; errors_AL|. A JSI\u0026thinsp;\u0026gt;\u0026thinsp;0.5 was interpreted as indicating that errors are primarily driven by patient-level biological ambiguity rather than instrument-specific variability.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eQuantitative data are presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard error of the mean (SEM) unless otherwise stated. One-way analysis of variance (ANOVA) followed by Tukey's multiple comparisons test was used to assess differences in HSA isoforms concentrations between patient groups defined by fibrosis stage (Control, F0/F1, F2, F3, F4-Child A, F4-Child B, F4-Child C). Associations between HSA isoforms and disease etiology were also explored using ANOVA. All statistical tests were two-sided, and a \u003cem\u003eP\u003c/em\u003e value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003cp\u003eReceiver operating characteristic (ROC) curve analysis was performed to evaluate the discriminatory performance of individual HSA isoforms in distinguishing between different clinically relevant stages of liver disease (e.g., controls vs. F0/F1, F3 vs. F4, Child-Pugh A vs. B/C). Optimal diagnostic thresholds were determined using the Youden index.\u003c/p\u003e \u003cp\u003eMultivariate analysis of the HSA isoforms profile was conducted using principal component analysis (PCA). PCA was performed using custom scripts in RStudio software (version 2025.05.1\u0026thinsp;+\u0026thinsp;513, Posit Software) applied to the relative abundance data derived from the entire deconvoluted albumin mass spectrum between 66,000 and 67,500 Da to visualize overall profile differences between patient groups. Data analysis and figure generation were performed using R software and GraphPad Prism\u0026reg; (version 9 for Mac OS, GraphPad Software, USA). For the cross-platform comparison, statistical equivalence was assessed using McNemar\u0026rsquo;s test (two-sided, α\u0026thinsp;=\u0026thinsp;0.05) and bootstrap confidence intervals for QWK were estimated from 1,000 resampling iterations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eEthics statement\u003c/h2\u003e \u003cp\u003e This study was conducted on residual biological material collected during routine clinical care at the University Hospital of Limoges (CHU de Limoges), France. Patients were informed of the potential use of their residual samples for research purposes and did not object, in accordance with the institutional information and consent procedures in place at CHU de Limoges. Under French law (Code de la Sant\u0026eacute; Publique, Art. L.1121-1), research conducted exclusively on residual biological material from informed patients who have not objected is classified as Recherche Impliquant la Personne Humaine de cat\u0026eacute;gorie 3 (RIPH3). This regulatory category does not require review or approval by a Comit\u0026eacute; de Protection des Personnes (CPP), and no CPP submission was therefore made. The biocollection from which the samples were drawn is formally registered with the French Ministry of Health under declarations DC 2010\u0026thinsp;\u0026minus;\u0026thinsp;1074 and AC-2016-2758. All procedures were conducted in accordance with the principles of the Declaration of Helsinki and the French Bioethics Act 2011\u0026thinsp;\u0026minus;\u0026thinsp;814 (Loi relative \u0026agrave; la bio\u0026eacute;thique).\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eConflicts of interest statement\u003c/h2\u003e \u003cp\u003eThe authors have no conflicts of interest to declare.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding statement:\u003c/h2\u003e \u003cp\u003eThis work received a local funding form CHU Limoges\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eSouleiman El Balkhi: Conceptualization, Methodology, data collection, data analysis and interpretation, drafting the article. Racym Berrah: Cross-platform data analysis, machine learning model development, statistical validation, figure generation. *The first two authors contributed equally to this work. L\u0026eacute;a Le Du: Patients inclusion, data collection, analysis and interpretationMohamad Ali Rahali and Roy Lakis: Sample analysis, data collection, analysis and interpretationFran\u0026ccedil;ois Ludovic Sauvage: Sample analysis, data collection, mass spectrometry analysis and interpretationPierre Marquet and Franck Saint-Marcoux: Methodology, data analysis and interpretation, critical revision of the article.Paul Carrier and Veronique Loustaud-Ratti: Conceptualization, Methodology, data collection, data analysis and interpretation, drafting the article.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThe authors are grateful to BISCEm unit (Univ. Limoges, UAR 2015 CNRS, US 42 Inserm, CHU Limoges) and Emilie Pinault for technical support regarding mass spectrometry analyses.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets generated and analysed during the current study contain patient-level clinical and proteomic data and cannot be made publicly available, in accordance with applicable patient privacy regulations (French Data Protection Act and GDPR) and the terms governing the biocollection registrations DC 2010-1074 and AC-2016-2758 (French Ministry of Health). Anonymised data supporting the findings of this study are available upon reasonable request to the corresponding author, subject to approval by the institutional data access committee of the University Hospital of Limoges (CHU de Limoges). The R scripts used for spectral preprocessing, feature selection, and cross-platform ordinal classification are available from the corresponding author upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSpinella, R., Sawhney, R. \u0026amp; Jalan, R. Albumin in chronic liver disease: structure, functions and therapeutic implications. \u003cem\u003eHepatol. Int.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e (1), 124\u0026ndash;132 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaldassarre, M. et al. Albumin Homodimers in Patients with Cirrhosis: Clinical and Prognostic Relevance of a Novel Identified Structural Alteration of the Molecule. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e6\u003c/b\u003e, 35987 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDas, S. et al. Hyperoxidized albumin modulates neutrophils to induce oxidative stress and inflammation in severe alcoholic hepatitis. \u003cem\u003eHepatology\u003c/em\u003e \u003cb\u003e65\u003c/b\u003e (2), 631\u0026ndash;646 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOettl, K. et al. Oxidative albumin damage in chronic liver failure: relation to albumin binding capacity, liver dysfunction and survival. \u003cem\u003eJ. Hepatol.\u003c/em\u003e \u003cb\u003e59\u003c/b\u003e (5), 978\u0026ndash;983 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlcaraz-Quiles, J. et al. Oxidized Albumin Triggers a Cytokine Storm in Leukocytes Through P38 Mitogen-Activated Protein Kinase: Role in Systemic Inflammation in Decompensated Cirrhosis. \u003cem\u003eHepatology\u003c/em\u003e \u003cb\u003e68\u003c/b\u003e (5), 1937\u0026ndash;1952 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStauber, R. E. et al. Human nonmercaptalbumin-2: a novel prognostic marker in chronic liver failure. \u003cem\u003eTher. Apher Dial\u003c/em\u003e. \u003cb\u003e18\u003c/b\u003e (1), 74\u0026ndash;78 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJalan, R. et al. Alterations in the functional capacity of albumin in patients with decompensated cirrhosis is associated with increased mortality. \u003cem\u003eHepatology\u003c/em\u003e \u003cb\u003e50\u003c/b\u003e (2), 555\u0026ndash;564 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCales, P. et al. A new generation of non-invasive tests of liver fibrosis with improved accuracy in MASLD. \u003cem\u003eJ. Hepatol.\u003c/em\u003e \u003cb\u003e82\u003c/b\u003e (5), 794\u0026ndash;804 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDomenicali, M. et al. Posttranscriptional changes of serum albumin: clinical and prognostic significance in hospitalized patients with cirrhosis. \u003cem\u003eHepatology\u003c/em\u003e \u003cb\u003e60\u003c/b\u003e (6), 1851\u0026ndash;1860 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaldassarre, M. et al. Determination of Effective Albumin in Patients With Decompensated Cirrhosis: Clinical and Prognostic Implications. \u003cem\u003eHepatology\u003c/em\u003e \u003cb\u003e74\u003c/b\u003e (4), 2058\u0026ndash;2073 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBernardi, M. Effective albumin - A novel paradigm in the management of decompensated liver cirrhosis. \u003cem\u003eJ. Transl Int. Med.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e (1), 11\u0026ndash;14 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarcia-Martinez, R. et al. Albumin: pathophysiologic basis of its role in the treatment of cirrhosis and its complications. \u003cem\u003eHepatology\u003c/em\u003e \u003cb\u003e58\u003c/b\u003e (5), 1836\u0026ndash;1846 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNaldi, M., Baldassarre, M., Domenicali, M., Bartolini, M. \u0026amp; Caraceni, P. Structural and functional integrity of human serum albumin: Analytical approaches and clinical relevance in patients with liver cirrhosis. \u003cem\u003eJ. Pharm. Biomed. Anal.\u003c/em\u003e \u003cb\u003e144\u003c/b\u003e, 138\u0026ndash;153 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEl Balkhi, S. et al. Early detection of liver injuries by the Serum enhanced binding test sensitive to albumin post-transcriptional modifications. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e (1), 1434 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLakis, R. et al. Semi-synthetic human albumin isoforms: Production, structure, binding capacities and influence on a routine laboratory test. \u003cem\u003eInt. J. Biol. Macromol.\u003c/em\u003e \u003cb\u003e250\u003c/b\u003e, 126239 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFanali, G. et al. Human serum albumin: from bench to bedside. \u003cem\u003eMol. Aspects Med.\u003c/em\u003e \u003cb\u003e33\u003c/b\u003e (3), 209\u0026ndash;290 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRahali, M. A. et al. Posttranslational-modifications of human-serum-albumin analysis by a top-down approach validated by a comprehensive bottom-up analysis. \u003cem\u003eJ. Chromatogr. B Analyt Technol. Biomed. Life Sci.\u003c/em\u003e \u003cb\u003e1224\u003c/b\u003e, 123740 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLakis, R. et al. Absolute Quantification of Human Serum Albumin Isoforms by Internal Calibration Based on a Top-Down LC-MS Approach. \u003cem\u003eAnal. Chem.\u003c/em\u003e \u003cb\u003e96\u003c/b\u003e (2), 746\u0026ndash;755 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNaldi, M. et al. Mass spectrometry characterization of circulating human serum albumin microheterogeneity in patients with alcoholic hepatitis. \u003cem\u003eJ. Pharm. Biomed. Anal.\u003c/em\u003e \u003cb\u003e122\u003c/b\u003e, 141\u0026ndash;147 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNaldi, M. et al. Mass spectrometric characterization of human serum albumin dimer: A new potential biomarker in chronic liver diseases. \u003cem\u003eJ. Pharm. Biomed. Anal.\u003c/em\u003e \u003cb\u003e112\u003c/b\u003e, 169\u0026ndash;175 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePaar, M. et al. Albumin in patients with liver disease shows an altered conformation. \u003cem\u003eCommun. Biol.\u003c/em\u003e \u003cb\u003e4\u003c/b\u003e (1), 731 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEuropean Association for Study of L. Asociacion Latinoamericana para el Estudio del H. EASL-ALEH Clinical Practice Guidelines: Non-invasive tests for evaluation of liver disease severity and prognosis. \u003cem\u003eJ. Hepatol.\u003c/em\u003e \u003cb\u003e63\u003c/b\u003e (1), 237\u0026ndash;264 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoursier, J. et al. An extension of STARD statements for reporting diagnostic accuracy studies on liver fibrosis tests: the Liver-FibroSTARD standards. \u003cem\u003eJ. Hepatol.\u003c/em\u003e \u003cb\u003e62\u003c/b\u003e (4), 807\u0026ndash;815 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWong, G. L. et al. Liver stiffness-based optimization of hepatocellular carcinoma risk score in patients with chronic hepatitis B. \u003cem\u003eJ. Hepatol.\u003c/em\u003e \u003cb\u003e60\u003c/b\u003e (2), 339\u0026ndash;345 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDietrich, C. et al. EFSUMB Guidelines and Recommendations on the Clinical Use of Liver Ultrasound Elastography, Update 2017 (Short Version). Ultraschall in der Medizin -. \u003cem\u003eEur. J. Ultrasound\u003c/em\u003e. \u003cb\u003e38\u003c/b\u003e (04), 377\u0026ndash;394 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerzigotti, A. et al. EASL Clinical Practice Guidelines on non-invasive tests for evaluation of liver disease severity and prognosis \u0026ndash; 2021 update. \u003cem\u003eJ. Hepatol.\u003c/em\u003e \u003cb\u003e75\u003c/b\u003e (3), 659\u0026ndash;689 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBedossa, P. \u0026amp; Poynard, T. An Algorithm for the Grading of Activity in Chronic Hepatitis C. \u003cem\u003eHepatology\u003c/em\u003e \u003cb\u003e24\u003c/b\u003e (2), 289\u0026ndash;293 (1996).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTapper, E. B. \u0026amp; Parikh, N. D. Diagnosis and Management of Cirrhosis and Its Complications: A Review. \u003cem\u003eJAMA\u003c/em\u003e \u003cb\u003e329\u003c/b\u003e (18), 1589\u0026ndash;1602 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchomaker, S. et al. Assessment of Emerging Biomarkers of Liver Injury in Human Subjects. \u003cem\u003eToxicol. Sci.\u003c/em\u003e \u003cb\u003e132\u003c/b\u003e (2), 276\u0026ndash;283 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, B. et al. Cytokeratin 18 knockdown decreases cell migration and increases chemosensitivity in non-small cell lung cancer. \u003cem\u003eJ. Cancer Res. Clin. Oncol.\u003c/em\u003e \u003cb\u003e142\u003c/b\u003e (12), 2479\u0026ndash;2487 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCione, E. et al. Liver Damage and microRNAs: An Update. \u003cem\u003eCurr. Issues. Mol. Biol.\u003c/em\u003e \u003cb\u003e45\u003c/b\u003e (1), 78\u0026ndash;91 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJampoka, K. et al. Serum miR-29a and miR-122 as Potential Biomarkers for Non-Alcoholic Fatty Liver Disease (NAFLD). \u003cem\u003eMicroRNA\u003c/em\u003e \u003cb\u003e7\u003c/b\u003e (3), 215\u0026ndash;222 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStarckx, S. et al. Evaluation of miR-122 and Other Biomarkers in Distinct Acute Liver Injury in Rats. \u003cem\u003eToxicol. Pathol.\u003c/em\u003e \u003cb\u003e41\u003c/b\u003e (5), 795\u0026ndash;804 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, X., He, Y., Mackowiak, B. \u0026amp; Gao, B. MicroRNAs as regulators, biomarkers and therapeutic targets in liver diseases. \u003cem\u003eGut\u003c/em\u003e \u003cb\u003e70\u003c/b\u003e (4), 784\u0026ndash;795 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSebastiani, G. Non-invasive assessment of liver fibrosis in chronic liver diseases: Implementation in clinical practice and decisional algorithms. \u003cem\u003eWorld J. Gastroenterol.\u003c/em\u003e ;\u003cb\u003e15\u003c/b\u003e(18). (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoursier, J. et al. Non-invasive tests accurately stratify patients with NAFLD based on their risk of liver-related events. \u003cem\u003eJ. Hepatol.\u003c/em\u003e \u003cb\u003e76\u003c/b\u003e (5), 1013\u0026ndash;1020 (2022).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Chronic liver disease, biomarker, Post-transcriptional modification, albumin, Machine learning, Cross-platform validation","lastPublishedDoi":"10.21203/rs.3.rs-9371068/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9371068/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eChronic liver disease (CLD) affects millions worldwide, yet accurately staging its progression without liver biopsy remains a major clinical challenge. Human serum albumin (HSA), the most abundant blood protein synthesized exclusively by the liver, undergoes measurable structural modifications as liver disease advances, making it a potential molecular marker of disease severity. Using high-resolution liquid chromatography\u0026ndash;mass spectrometry (LC-HR-MS), we quantified native HSA and nine modified isoforms in plasma from 172 CLD patients spanning all fibrosis stages and 82 healthy controls. Native HSA declined markedly with disease severity, reaching 4.1\u0026ndash;4.2 g/L in decompensated cirrhosis versus 12.2 g/L in controls. Modified isoforms showed stage-specific patterns, and their ratios to native HSA amplified the diagnostic signal for advanced disease. A machine learning classifier trained on the full albumin spectral profile achieved substantial agreement with biopsy-based staging and outperformed the widely-used FIB-4 index in clinical triage (81.5% vs. 59.3% accuracy). Critically, these results were fully reproduced on two independent instruments from different manufacturers (McNemar p\u0026thinsp;=\u0026thinsp;0.149), confirming the platform-independence of the albumin signature. These findings establish HSA spectral profiling as a promising non-invasive staging tool for CLD, with cross-platform reproducibility supporting its translation to multicenter clinical practice.\u003c/p\u003e","manuscriptTitle":"Human Serum Albumin Profiling by Top-down Analysis Enables Multi-Class Liver Fibrosis Staging: A Cross-Platform Validation Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-11 06:15:51","doi":"10.21203/rs.3.rs-9371068/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-05-18T09:44:57+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-16T19:07:31+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-07T20:36:43+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"183945785630273974360718098191888344284","date":"2026-05-06T08:03:49+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"41081735852388601279124604264797570237","date":"2026-04-28T19:27:01+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-28T16:58:33+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-28T16:41:56+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-04-16T14:35:04+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-14T16:11:10+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-04-14T15:09:36+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c2bc5318-c538-4435-a0e6-95524e417368","owner":[],"postedDate":"May 11th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-18T09:44:57+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-16T19:07:31+00:00","index":128,"fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-07T20:36:43+00:00","index":127,"fulltext":""},{"type":"reviewerAgreed","content":"183945785630273974360718098191888344284","date":"2026-05-06T08:03:49+00:00","index":125,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[{"id":67632715,"name":"Health sciences/Biomarkers"},{"id":67632716,"name":"Health sciences/Diseases"},{"id":67632717,"name":"Health sciences/Gastroenterology"},{"id":67632718,"name":"Health sciences/Medical research"}],"tags":[],"updatedAt":"2026-05-18T09:54:27+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-11 06:15:51","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9371068","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9371068","identity":"rs-9371068","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}