Identification of acetylation-related molecular signatures in knee osteoarthritis patients with significant response to warm-needle acupuncture using machine-learning approaches

Objective: This study aims to characterize the remodeling of acetylation-related gene expression in knee osteoarthritis (KOA) patients with a marked response to warm-needle acupuncture (WNA) and to identify key genes and underlying immunoregulatory mechanisms. Methods: In this prospective self-paired study, 34 KOA patients were recruited, and whole-blood samples were collected before and after WNA treatment for transcriptome sequencing. Differential expression analysis identified acetylation-related genes followed by enrichment and protein–protein interaction analyses. Key genes were extracted via feature selection based on LASSO and SVM-RFE methods and further used to establish and validate a multigene logistic regression model. Consensus clustering was used to identify two acetylation subtypes (ACEcluster A/B), and their immune characteristics were further explored by ssGSEA and immune-cell infiltration profiling. Results: After treatment, acetylation-related genes were globally upregulated and were enriched in protein acetylation and acetyltransferase complex pathways. Further, four genes (SPRED1, HDAC3, NSRP1, and DUSP1) exhibited stable performance and were further used to build a nomogram and achieve high discriminative performance (AUC = 0.908 in the training set; 0.810 in the validation set). Subtype B displayed higher acetylation activity and immune-cell infiltration. Co-expression analysis on 452 acetylation-related genes extracted 107 highly co-regulated candidates and further clustered into two groups coinciding with ACEcluster classification. Conclusion: WNA markedly remodels the peripheral-blood acetylation-related transcriptomic network in KOA and suggests that stronger acetylation activity is associated with immune-cell activation. The four-gene nomogram provides both mechanistic insights and a potential tool for individualized prediction of WNA efficacy.