The diagnostic and therapeutic potential of palmitoylation-related genes in Alzheimer’s disease

Alzheimer’s disease (AD) is a highly heterogeneous disorder, for which early accurate diagnosis and subtype stratification remain clinically challenging. Palmitoylation, a critical reversible post‑translational modification, has been implicated in the pathological progression of AD. However, a comprehensive investigation of palmitoylation‑related pathway genes in the context of AD diagnosis and molecular subtyping is still lacking. Transcriptomic data from three AD cohorts in the GEO database were integrated and corrected for batch effects. Palmitoylation‑related differentially expressed genes (PR‑DEGs) were screened through differential expression analysis combined with weighted gene co‑expression network analysis (WGCNA). Based on these PR‑DEGs, molecular subtyping, pathway enrichment, and immune infiltration analyses were performed in AD. Twelve machine learning algorithms were employed to construct 113 classification models, and core palmitoylation‑related biomarkers with stable cross‑cohort performance were identified via multi‑algorithm cross‑screening. Model generalizability was validated using two independent external GEO datasets (GSE132903 and GSE138260). In addition, SHAP analysis, single‑cell transcriptomics, regulatory network construction, and drug prediction were integrated to elucidate the biological functions and clinical potential of the core genes. In this study, 15 core PR‑DEGs were identified, and AD was stratified into two heterogeneous subtypes: subtype C1, whose palmitoylation profile resembled that of normal aging, and subtype C2, which exhibited pronounced palmitoylation suppression, synaptic impairment, and immune dysregulation. A random forest model constructed based on the PR‑DEGs demonstrated stable and excellent diagnostic performance, with consistently high area under the curve (AUC) values across multiple cohorts. Six core diagnostic genes were further screened, among which CHRM1 showed the highest diagnostic value and model contribution, along with a neuron‑specific aberrant expression pattern. These core genes exhibited coordinated dysregulation and were implicated in multiple AD‑relevant pathways, including synaptic homeostasis, lipid metabolism, Wnt signaling, and neuroinflammation. In addition, several potential targeted therapeutic agents were predicted. In this study, a preliminary machine learning‑based diagnostic and subtype stratification analytic framework for AD was constructed based on the palmitoylation pathway. Bioinformatics analyses suggested that core genes, including CHRM1, PORCN, CRMP1, GNG3, CLIP3and ACOT7, may hold promise as potential biomarkers for precise diagnosis and targeted intervention in AD, thereby providing a theoretical reference for early screening, personalized mechanistic investigation, and therapeutic exploration.