Multi-omics and machine learning-driven identification of PTM-related potential biomarkers and therapeutic insights in MASLD

Abstract

Metabolic dysfunction-associated steatotic liver disease (MASLD) is biologically heterogeneous and lacks sufficiently validated molecular biomarkers for disease characterization. Post-translational modification (PTM)-related genes may link metabolic stress to lipid dysregulation and inflammation, but their relevance remains unclear. We integrated liver transcriptomic datasets and PTM-related gene annotation with machine-learning prioritization, cell-resolved analysis, and experimental validation. PELI1, VPS41, and TRIM32 were prioritized as potential diagnostic candidates based on their disease-associated expression and discriminative performance in the examined transcriptomic cohorts, with concordant expression changes observed in an independent local liver cohort. Single-nucleus RNA sequencing suggested that disease-associated VPS41 upregulation was most prominent in hepatocytes. VPS41 expression also increased in livers from a diet-induced mouse model of MASLD and in lipid-loaded hepatocytes. In primary mouse hepatocytes, Vps41 knockdown aggravated lipid accumulation, lipogenic gene expression, and inflammatory responses, whereas VPS41 overexpression attenuated these changes. These findings distinguish the expression-based evidence supporting the diagnostic potential of PELI1 and TRIM32 from the functional evidence for VPS41 and support a protective role for VPS41 under hepatocellular lipid stress. Prospective clinical validation and in vivo gain- and loss-of-function studies are required to establish diagnostic utility and the causal and therapeutic relevance of VPS41.