Biomarker Discovery for Metabolic Dysfunction-associated Steatotic Liver Disease Utilizing Mendelian Randomization, Machine Learning, and External Validation
Biomarker Discovery for Metabolic Dysfunction-associated Steatotic Liver Disease Utilizing Mendelian Randomization, Machine Learning, and External Validation
Background and Aims: The causal biomarkers for metabolic dysfunction-associated steatotic liver disease (MASLD) and their clinical value remain unclear. In this study, we aimed to identify biomarkers for MASLD and evaluate their diagnostic and prognostic significance. Methods: We conducted a Mendelian randomization analysis to assess the causal effects of 2,925 molecular biomarkers (from proteomics data) and 35 clinical biomarkers on MASLD. Mediation analysis was performed to determine whether clinical biomarkers mediated the effects of molecular biomarkers. The association between key clinical biomarkers and MASLD was externally validated in a hospital-based cohort (n = 415). A machine learning–based diagnostic model for MASLD was developed and validated using the identified molecular biomarkers. Prognostic significance was evaluated for both molecular and clinical biomarkers. Results: Six molecular biomarkers—in-cluding canopy FGF signaling regulator 4 (CNPY4), ectonu-cleoside triphosphate diphosphohydrolase 6 (ENTPD6), and major histocompatibility complex, class I, A (HLA-A)—and eight clinical biomarkers (e.g., serum total protein (STP)) were identified as causally related to MASLD. STP partially mediated the effect of HLA-A on MASLD (23.61%) and was associated with MASLD in the external cohort (odds ratio = 1.080, 95% confidence interval: 1.011–1.155). A random forest model demonstrated high diagnostic performance (AUC = 0.941 in training; 0.875 in validation). High expres-sion levels of CNPY4 and ENTPD6 were associated with the development of and poorer survival from hepatocellular car-cinoma. Low STP (<60 g/L) predicted all-cause mortality (HR = 2.50, 95% confidence interval: 1.22–5.09). Conclusions: This study identifies six causal molecular biomarkers (e.g., CNPY4, ENTPD6, HLA-A) and eight clinical biomarkers for MASLD. Notably, STP mediates the effect of HLA-A on MASLD and is associated with all-cause mortality.
Causal biomarkers, Machine learning, Mediation analysis, Mendelian randomization, Metabolic dysfunction-associated fatty liver disease, Non-invasive diagnosis, Prognosis, Proteomics
723-733
Feng, Gong
01a55a38-0227-43aa-8eba-6ae8ca4665e0
Targher, Giovanni
bd646ac5-baa1-4944-8050-0c001b559283
Byrne, Chrisopher D.
1370b997-cead-4229-83a7-53301ed2a43c
He, Na
d3bd1563-d091-4746-8b84-6901304f87c1
Mi, Man
cf5d1581-a047-4db2-8e6d-cfd64154ceda
Liu, Yi
90fd6b8c-b22a-4910-ae5d-775f137193f1
Zhu, Hongbin
4506ac02-3d92-4233-96bb-9683bea4c8f2
Zheng, Ming-Hua
3a497ced-7e64-4f40-b442-702cc802c310
Ye, Feng
4e4bba72-e531-4131-9974-adfb3d090170
16 July 2025
Feng, Gong
01a55a38-0227-43aa-8eba-6ae8ca4665e0
Targher, Giovanni
bd646ac5-baa1-4944-8050-0c001b559283
Byrne, Chrisopher D.
1370b997-cead-4229-83a7-53301ed2a43c
He, Na
d3bd1563-d091-4746-8b84-6901304f87c1
Mi, Man
cf5d1581-a047-4db2-8e6d-cfd64154ceda
Liu, Yi
90fd6b8c-b22a-4910-ae5d-775f137193f1
Zhu, Hongbin
4506ac02-3d92-4233-96bb-9683bea4c8f2
Zheng, Ming-Hua
3a497ced-7e64-4f40-b442-702cc802c310
Ye, Feng
4e4bba72-e531-4131-9974-adfb3d090170
Feng, Gong, Targher, Giovanni, Byrne, Chrisopher D., He, Na, Mi, Man, Liu, Yi, Zhu, Hongbin, Zheng, Ming-Hua and Ye, Feng
(2025)
Biomarker Discovery for Metabolic Dysfunction-associated Steatotic Liver Disease Utilizing Mendelian Randomization, Machine Learning, and External Validation.
Journal of Clinical and Translational Hepatology, 13 (9), .
(doi:10.14218/JCTH.2025.00270).
Abstract
Background and Aims: The causal biomarkers for metabolic dysfunction-associated steatotic liver disease (MASLD) and their clinical value remain unclear. In this study, we aimed to identify biomarkers for MASLD and evaluate their diagnostic and prognostic significance. Methods: We conducted a Mendelian randomization analysis to assess the causal effects of 2,925 molecular biomarkers (from proteomics data) and 35 clinical biomarkers on MASLD. Mediation analysis was performed to determine whether clinical biomarkers mediated the effects of molecular biomarkers. The association between key clinical biomarkers and MASLD was externally validated in a hospital-based cohort (n = 415). A machine learning–based diagnostic model for MASLD was developed and validated using the identified molecular biomarkers. Prognostic significance was evaluated for both molecular and clinical biomarkers. Results: Six molecular biomarkers—in-cluding canopy FGF signaling regulator 4 (CNPY4), ectonu-cleoside triphosphate diphosphohydrolase 6 (ENTPD6), and major histocompatibility complex, class I, A (HLA-A)—and eight clinical biomarkers (e.g., serum total protein (STP)) were identified as causally related to MASLD. STP partially mediated the effect of HLA-A on MASLD (23.61%) and was associated with MASLD in the external cohort (odds ratio = 1.080, 95% confidence interval: 1.011–1.155). A random forest model demonstrated high diagnostic performance (AUC = 0.941 in training; 0.875 in validation). High expres-sion levels of CNPY4 and ENTPD6 were associated with the development of and poorer survival from hepatocellular car-cinoma. Low STP (<60 g/L) predicted all-cause mortality (HR = 2.50, 95% confidence interval: 1.22–5.09). Conclusions: This study identifies six causal molecular biomarkers (e.g., CNPY4, ENTPD6, HLA-A) and eight clinical biomarkers for MASLD. Notably, STP mediates the effect of HLA-A on MASLD and is associated with all-cause mortality.
Text
R1_MR Identifies Biomarkers for MASLD_JCTH20250622
- Accepted Manuscript
Text
Supplementary Material_20250622
- Accepted Manuscript
Text
Figure 1
- Accepted Manuscript
Text
Figure 2
- Accepted Manuscript
Text
Figure 3
- Accepted Manuscript
Text
Figure 4
- Accepted Manuscript
Text
Figure 5
- Accepted Manuscript
Text
Figure 6
- Accepted Manuscript
Text
Graphical abstract
- Accepted Manuscript
Text
JCTH.2025.0(0).0.00270.Gong Feng
- Version of Record
Show all 10 downloads.
More information
Accepted/In Press date: 2 July 2025
e-pub ahead of print date: 16 July 2025
Published date: 16 July 2025
Additional Information:
Publisher Copyright:
© 2025 The Author(s).
Keywords:
Causal biomarkers, Machine learning, Mediation analysis, Mendelian randomization, Metabolic dysfunction-associated fatty liver disease, Non-invasive diagnosis, Prognosis, Proteomics
Identifiers
Local EPrints ID: 503947
URI: http://eprints.soton.ac.uk/id/eprint/503947
ISSN: 2225-0719
PURE UUID: dd0e1136-05fd-42a6-a62f-fb460f961e84
Catalogue record
Date deposited: 19 Aug 2025 16:31
Last modified: 30 Sep 2025 01:40
Export record
Altmetrics
Contributors
Author:
Gong Feng
Author:
Giovanni Targher
Author:
Na He
Author:
Man Mi
Author:
Yi Liu
Author:
Hongbin Zhu
Author:
Ming-Hua Zheng
Author:
Feng Ye
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics
