Identification of palmitoylated biomarkers in non-alcoholic fatty liver disease via integrated bioinformatics analysis and machine learning.

Non-alcoholic fatty liver disease (NAFLD) is a global health challenge with complex pathogenesis and limited diagnostic biomarkers. Palmitoylation, a post-translational modification, has emerged as a critical regulator in metabolic disorders, yet its role in NAFLD remains underexplored. This study integrated bioinformatics analysis and machine learning to identify palmitoylation-related biomarkers for NAFLD. Transcriptomic datasets from human liver tissues were analyzed to identify differentially expressed genes (DEGs) and co-expression modules via WGCNA. Intersection analysis revealed 60 palmitoylation-related DEGs (PR-DEGs). Seven machine learning models were employed, with Neural Network (NNET) and Decision Tree (DT) outperforming others, identifying three hub genes: TYMS, WNT5A, and ZFP36. A nomogram integrating these genes demonstrated robust diagnostic accuracy (AUC = 0.976). The pivotal role of these genes in diagnosing NAFLD was confirmed using the validation dataset (AUC = 0.903). Functional enrichment linked these genes to TNF signaling, lipid metabolism, and immune pathways. Single-cell RNA-seq analysis highlighted their expression in hepatocytes and immune cells, with altered intercellular communication patterns. Immune infiltration analysis revealed significant shifts in monocytes, dendritic cells, and macrophages in NAFLD. Regulatory network analysis highlighted that hsa-let-7b-5p might be pivotal co-regulator of the three hub gene expressions. Finally, the top 10 potential gene-targeted drugs were screened. This study unveils novel palmitoylation-related biomarkers and provides insights into NAFLD pathogenesis, offering diagnostic and therapeutic avenues.