Single-nucleus cNMF identifies a six-gene classifier predictive of atrial fibrillation recurrence

Atrial fibrillation (AF) reflects intertwined immune–stromal remodeling, yet compact and generalizable diagnostic markers remain limited. Left-atrial single-nucleus RNA-seq was analyzed using SCTransform, PCA/UMAP, and lineage-aware cNMF (K = 4–8) to define transcriptional programs. Immune-related genes were filtered via ImmPort and analyzed using clusterProfiler, ReactomePA, and Enrichr. A 16-model benchmarking framework integrating stepwise GLM and LASSO identified a six-gene panel (EXT1, ATG7, PTPRC, MERTK, IRAK3, RORA). Immune infiltration was estimated using CIBERSORTx, and clinical outcomes were assessed using Andersen–Gill and Fine–Gray models. Experimental validation included RT-qPCR, siRNA-mediated knockdown, wound-healing, colony formation, and immunofluorescence assays in AF-mimicking atrial fibroblasts. Docking-guided drug screening and gene–drug interaction assays were performed to evaluate therapeutic targeting. Luciferase assays confirmed miRNA regulation. cNMF revealed lineage-restricted programs enriched in TGF-β, interferon-α, and inflammatory signaling, with macrophage and T cell activation as dominant features. The six-gene panel achieved AUROC = 0.583 (AUPRC = 0.627) with consistent external validation. RT-qPCR confirmed upregulation of all genes under AF-like conditions. Functional assays demonstrated that MERTK and IRAK3 regulate fibroblast activation, migration, proliferation, and extracellular matrix deposition, with MERTK exerting the strongest effect. Docking-guided screening identified MK-2461 as a potent inhibitor of fibrotic remodeling, significantly reducing gene expression, wound healing, and α-SMA/collagen accumulation. Lack of additive effects with MERTK knockdown indicated target dependency. hsa-miR-155-5p directly repressed all six genes. Persistent AF showed increased recurrence and progression risk. This integrative framework links immune–stromal remodeling, miRNA regulation, functional fibroblast activation, and therapeutic targeting, identifying MERTK as a key driver and actionable target in AF progression. Graphical abstract Overview of the integrated computational and experimental framework linking single-nucleus transcriptomics to a therapeutically actionable six-gene signature in atrial fibrillation (AF).