Interpretable Machine Learning reveals the Role of PANoptosis in the Diagnosis and Subtyping of NAFLD

Non-alcoholic fatty liver disease (NAFLD) is a global health challenge characterized by complex pathogenesis and limited therapeutic options. Emerging evidence highlights PANoptosis-a coordinated interplay of pyroptosis, apoptosis, and necroptosis-as a critical driver of metabolic and immune dysregulation in NAFLD. Here, we integrated multiple datasets and interpretable machine learning to unravel the role of PANoptosis in NAFLD diagnosis, subtyping, and immune microenvironment remodeling. By intersecting differentially expressed genes and PANoptosis-related genes, we identified 9 hub genes (e.g., TRADD, CASP6, TNFRSF1A and TNFAIP3) and constructed a robust diagnostic model (AUC = 0.976) validated via SHAP analysis and nomogram. Unsupervised consensus clustering stratified NAFLD patients into two PANoptosis-driven subtypes (C1/C2 and CA/CB), revealing distinct immune cell infiltration patterns and pathway activation. Single-cell sequencing further localized hub genes to immune cells and revealed their cell communication, implicating their roles in the progression of NAFLD. Molecular docking studies identified fostamatinib and minocycline as potential therapeutic candidates, while pan-cancer analysis revealed that TNFRSF1A overexpression is associated with poor prognosis across multiple cancer types. This study enhances the understanding of PANoptosis as a crucial diagnostic and therapeutic target in NAFLD, providing novel insights into immune-mediated pathogenesis and paving the way for treatment strategies.