Identification and experimental validation of mitochondrial metabolism-associated diagnostic markers for nonalcoholic fatty liver disease using integrated WGCNA and machine learning

The global rise of Non-alcoholic fatty liver disease (NAFLD) necessitates better diagnostic and therapeutic strategies. This study aimed to identify novel diagnostic biomarkers and molecular subtypes for NAFLD. We analyzed Gene Expression Omnibus (GEO) datasets (GSE89632, GSE63067) to pinpoint mitochondrial metabolism-related genes (MRGs) linked to NAFLD. Differential expression analysis and Weighted Gene Co-expression Network Analysis (WGCNA) identified candidate genes. Key diagnostic biomarkers were refined using three machine learning algorithms: Support Vector Machine Recursive Feature Elimination (SVM-RFE), LASSO regression, and Random Forest. These biomarkers were subsequently validated via quantitative PCR and Western blot. We identified four pivotal diagnostic markers: ANGPTL4, PPARGC1A, NDUFA6, and CYP7A1. A diagnostic nomogram constructed with these markers demonstrated high predictive efficacy. Furthermore, consensus clustering revealed two distinct NAFLD subtypes based on mitochondrial metabolism. Significant differences in gene expression, enriched pathways, and immune infiltration landscapes were observed between these subtypes. Experimental validation confirmed the expression patterns of the four diagnostic markers, aligning with our bioinformatics findings. This study identifies four promising diagnostic biomarkers for NAFLD and delineates distinct molecular subtypes, providing valuable insights for early diagnosis, prognosis, and the development of personalized treatment.