HSA Isoform Profile as Biomarker of Chronic Liver Disease Staging
Biomarker description
- Target: Panel of 10 circulating HSA isoforms; multivariate spectral fingerprint (66,000–68,000 Da)
- Disease context: Liver fibrosis / Chronic Liver Disease (CLD) — staging F0/F1 through F4_C
- Analytical method: Top-down LC-MS (LC-HR-MS); absolute quantification (g/L) via internal calibration
- Source study: el-balkhi-2025 — El Balkhi et al. 2025 (ALBOM study)
Biomarker panel — 10 isoforms
The 10 quantified isoforms
| Isoform | PTM(s) | Δmass (Da) | Pattern with CLD progression |
|---|---|---|---|
| Native HSA | — | 0 | ↓ progressive; precipitous F4_B/C |
| HSA-DA | N-term truncation | −115 | ↓ all stages (paradoxically lower in patients) |
| HSA-L | C-term truncation | variable | No significant change |
| HSA+CYS | Cysteinylation Cys34 | +119 | Biphasic — peaks F4_A (11.1 g/L), falls F4_B/C |
| HSA+SO₃H | Sulfonylation Cys34 | +48 | ↓ F4_B/C |
| HSA+GLYC | Glycation (mono) | +162 | Biphasic — peaks F3, falls F4_B/C |
| HSA-DA+CYS | Truncation + CYS | −115 +119 | ↓ F4_B/C |
| HSA+CYS+GLYC | CYS + glycation | +281 | Biphasic — peaks F3 (1.9 g/L) |
| HSA+2GLYC | Double glycation | +324 | Biphasic — peaks F4_A |
| HSA+CYS+2GLYC | CYS + 2× glycation | +443 | Progressive ↑ — end-stage marker |
The three patterns (biological interpretation)
- Decrease group: Reflects loss of hepatocyte synthetic capacity and depletion of functional HSA substrate (native, truncated, irreversibly-oxidized forms all reduced in advanced disease)
- Biphasic group: Reflects escalating oxidative/glycative stress in compensated disease followed by exhaustion of the substrate pool in decompensation. Absolute ↓ but ratio-to-native ↑ — the ratio is the correct metric
- Increase group: HSA+CYS+2GLYC — cumulative end-stage damage marker; measures the proportion of HSA that has undergone maximal modification; near-zero in healthy individuals
Clinical performance
Native HSA (single marker)
| Comparison | AUC |
|---|---|
| Controls vs F0/F1 | 0.67 |
| Controls vs F4_B | 0.99 |
| Controls vs F4_C | 0.89 |
Best for decompensated cirrhosis; limited for early fibrosis.
Isoform ratios (normalized to native HSA)
| Ratio | Comparison | Sensitivity | Specificity |
|---|---|---|---|
| HSA+CYS/Native | Controls vs F4_C | 65% | 99% |
| HSA+GLYC/Native | Controls vs F2 | detects earliest ← | — |
| HSA+GLYC/Native | Controls vs F4_B | 85% | 100% |
| HSA+GLYC/Native | Controls vs F4_C | 70% | 99% |
| HSA+CYS+GLYC/Native | Controls vs F4_B | 77% | 99% |
| HSA+CYS+GLYC/Native | Controls vs F4_C | 70% | 99% |
HSA+GLYC/Native is the best single-ratio discriminator; also uniquely capable of discriminating controls from F2 (early disease).
Multivariate OrdinalForest model (75 spectral features + 4 clinical variables)
| Platform | n_test | QWK | 95% CI | Accuracy (3-class) |
|---|---|---|---|---|
| Bruker timsTOF Pro2 | 46 | 0.862 | 0.735–0.923 | 81.5% |
| Sciex TripleTOF 5600+ | 49 | 0.916 | 0.822–0.964 | — |
| FIB-4 index (comparator) | same | 0.188–0.229 | — | 59.3% |
Improvement over FIB-4: +26 percentage points accuracy. FIB-4 gray zone (1.30–2.67): ALBOM correctly classifies 62.5% (⚠️ n=8, preliminary).
Cross-platform reproducibility
- McNemar’s test: p = 0.149 → not significantly different between Bruker and Sciex
- Jaccard index of errors: 0.696 → errors are biologically driven (boundary stages), not instrument artifacts
- Both platforms: QWK within “substantial to near-perfect agreement” range
Validation status
- ☑ Discovery — el-balkhi-2025 (published, Scientific Reports 2026, DOI 10.1038/s41598-026-57614-y)
- ☑ Cross-instrument reproducibility (within single center)
- ☐ Replicated (independent cohort)
- ◧ Externally validated (multicenter) — MALAHBAR ongoing: 560 patients / 8 French CHUs (target >700), readout expected 2027
- ◧ Prospective longitudinal validation (MALAHBAR NCT06318949)
- ☐ Clinically approved
Advantages over existing markers
| Feature | FIB-4 | Elastography | HSA isoform profile |
|---|---|---|---|
| Invasive | No | No | No |
| Multi-class staging | Weak | Weak | Yes (6 classes) |
| Gray zone problem | Severe (30–40%) | Moderate | Reduced (62.5% classified) |
| Biological insight | Indirect | Structural proxy | Direct molecular readout |
| Etiology-specific | No | No | ⚠️ To be established |
| Platform portability | Universal | Device-specific | Cross-manufacturer confirmed |
| Clinical deployment | Immediate | Moderate barrier | High barrier (LC-HR-MS) |
HSA isoforms provide a functional and molecular measure of CLD severity rather than anatomical or enzymatic surrogates.
The biphasic paradox — key interpretive concept
Modified isoforms (HSA+CYS, HSA+GLYC) peak in absolute concentration at compensated cirrhosis (F4_A), then fall in decompensation (F4_B/C). This can be mistaken for improvement. The correct interpretation:
- Absolute ↓ = substrate (native HSA) depleted faster than modification accrues
- Ratio (isoform/native) ↑ = the proportion of damaged albumin continues to rise
- Always use ratios, not absolute values, for diagnostic interpretation
Limitations and gaps
- ⚠️ Single-center cross-sectional cohort (n=254 total) — limited generalizability
- ⚠️ Etiology bias: MASH + ALD predominate; etiology-specific profiles not yet characterized
- ⚠️ F2/F3 overlap: FibroScan staging limitations cause cluster dispersion; biopsy validation limited
- ⚠️ Platform 1 (Bruker): baseline correction attenuates high-mass glycated species >67,500 Da — Platform 1 classifier is valid but sub-optimal for poly-glycated endpoints
- ⚠️ Gray zone FIB-4 analysis: n=8, hypothesis-generating only
- ⚠️ LC-HR-MS not universally available in routine clinical labs
Our contribution
This biomarker page documents the primary output of the ALBOM study, generated using the CQFD-PTM pipeline and Top-down proteomics methodology developed at CHU Limoges / INSERM UMR1248. The ALBOM study is the proof-of-concept clinical validation for both the pipeline and the biomarker concept.
Key studies
- el-balkhi-2025 — El Balkhi & Berrah et al. 2025 (ALBOM manuscript, Scientific Reports submission)