A Multi-Omics Approach Combining Causal Inference and In Vivo Validation Identifies Key Protein Drivers of Alcohol-Associated Liver Disease

Qingyi Zhou^1,2xuan Ma³Qianqian Cui²Lei Zhang¹Chao Yao¹Zilu Zhang²Xiaoli Wang^2*Liang Chu^1*

• ¹Second Affiliated Hospital of Bengbu Medical College, Bengbu, China
• ²Bengbu Medical University, Bengbu, China
• ³Shandong Second Medical University, Weifang, China

Background: Alcohol-associated liver disease (ALD) constitutes a global health crisis, yet the molecular mechanisms driving its pathogenesis remain unresolved, hindering the development of effective therapeutics. A key challenge is differentiating correlational biomarkers from causal drivers. Here, we systematically characterize the ALD causal proteome to uncover novel pathogenic mediators and prioritize therapeutic targets. Methods: We implemented a multi-stage pipeline integrating human genetics with experimental validation. A two-sample Mendelian randomization (MR) framework, leveraging large-scale plasma proteomics and ALD genome-wide association study (GWAS) data, was employed to nominate proteins causally linked to ALD. These candidates underwent functional enrichment analysis. To control for genetic linkage confounding, findings were validated using Summary Mendelian Randomization (SMR). Single-cell RNA sequencing from a murine ALD model was employed to identify the hepatic cellular origins of the validated targets. Finally, their functional relevance was confirmed in vivo using a chronic-plus-binge ethanol feeding mouse model. Results: MR analysis identified 17 proteins with a putative causal association with ALD. Functional enrichment implicated these proteins in inflammatory and immune response pathways. After SMR validation, four high-confidence causal proteins were identified: TREML2 (Triggering Receptor Expressed on Myeloid cells-like 2) and MMP12 (Matrix Metallopeptidase 12) as risk factors, and PLA2R1 (Phospholipase A2 Receptor 1) and MAX (MYC Associated Factor X) as protective factors. Single-cell transcriptomics resolved their hepatic cellular sources. Treml2 and Mmp12 were upregulated in hepatic myeloid cells (macrophages and monocytes), whereas Pla2r1 and Max were downregulated in hepatocytes. These findings were corroborated in our ALD mouse model. Conclusion: By integrating genetic causal inference with multi-level functional genomics, we identify and validate four causal protein drivers of ALD. Our findings reveal a novel pathogenic axis where ALD risk is governed by a balance between pro-inflammatory myeloid cells (TREML2/MMP12) and hepatocyte functions (PLA2R1/MAX). These genetically validated targets provide critical insights into ALD pathophysiology and represent promising therapeutic avenues.