Identification of estrogen-related biomarkers in metabolic dysfunction-associated steatotic liver disease (MASLD) through machine learning and single-cell analysis

Background: Metabolic dysfunction-associated steatotic liver disease (MASLD), formerly known as non-alcoholic fatty liver disease (NAFLD), is a prevalent chronic liver disease linked to metabolic risk factors. Sex hormones, particularly estrogen, appear to influence MASLD development, as premenopausal women are relatively protected. This study aims to identify estrogen-related biomarkers of MASLD using integrated bioinformatics, machine learning, and single-cell RNA sequencing (scRNA-seq) approaches. Methods: Six public transcriptomic datasets (five training, one validation) were analyzed to find differentially expressed genes (DEGs) in MASLD versus controls. Estrogen-related genes (ERGs) were obtained from the Molecular Signatures Database (MSigDB) and intersected with DEGs to define estrogen-associated DEGs (DEERGs). Machine learning algorithms—least absolute shrinkage and selection operator (LASSO) regression and extreme gradient boosting (XGBoost)—were applied to select key feature genes, which were then validated in an independent cohort. Immune cell infiltration was evaluated by CIBERSORT and single-sample gene set enrichment analysis (ssGSEA). Consensus clustering based on feature genes defined molecular subtypes. Additionally, a human liver scRNA-seq dataset was analyzed to map cell-type-specific expression and pathway activity. A competing endogenous RNA (ceRNA) network of lncRNA-miRNA-mRNA interactions was constructed for the diagnostic genes. Results: Fourteen DEERGs were identified, and two diagnostic biomarkers—IGFBP2 and P4HA1—showed high diagnostic accuracy. Both genes were associated with immune infiltration and defined two estrogen-related molecular subtypes with distinct immune microenvironment features. Single-cell analysis localized IGFBP2 and P4HA1 to hepatocytes, implicating estrogen-related regulation in metabolic and fibrotic remodeling. CellChat analysis revealed weakened hepatocyte-centered signaling in MASLD. The ceRNA network suggested that multiple lncRNAs (e.g., NEAT1, MALAT1, XIST) may modulate these biomarkers via shared miRNAs. Conclusion: Through integrated multi-cohort analysis, we identified IGFBP2 and P4HA1 as novel estrogen-related tissue biomarkers of MASLD. Preliminary exploration suggests their potential detectability in peripheral blood, though large-scale tissue and protein-level validations are required. Our findings provide insight into sex hormone-linked mechanisms in MASLD and propose candidate molecular targets for improved risk stratification and therapy.