A gene interaction network–based molecular taxonomy of dilated cardiomyopathy highlighting immune and inflammatory heterogeneity

Dilated cardiomyopathy (DCM) is a highly heterogeneous disease, complicating mechanistic understanding and limiting effective patient stratification. To capture this complexity, a gene interaction network-based taxonomy was developed. A total of 245 DCM samples and 174 controls from four cohorts were included. Gene–gene interactions were quantified for each sample to build interaction networks based on transcriptomic data and Reactome database, and representative interactions were used for consensus clustering to identify subtypes. Subtype stability was assessed with Uniform Manifold Approximation and Projection (UMAP), silhouette, and nearest-template prediction. Weighted gene co-expression network analysis (WGCNA), enrichment analysis, and immune-infiltration profiling were applied to characterize subtype heterogeneity. An immunoadsorption-treated cohort was used to predict therapeutic responsiveness. Finally, subtype-specific differentially expressed genes (DEGs) were validated by qPCR in an Angiotensin II–stimulated cardiomyocyte model. Three reproducible DCM subtypes were consistently identified, with EPS1 comprising 43.2–52.5% of patients, EPS2 40.0–44.2%, and EPS3 7.5–13.5%. These subtypes exhibited distinct molecular characteristics, reflecting heterogeneous biological states. Edge-perturbation subtype 1 (EPS1) showed mitochondrial/TCA-cycle and proteostasis dysregulation and corresponded to responders to immunoadsorption with IgG substitution (IA/IgG). EPS2 was characterized by enrichment of humoral immune and platelet activation pathways, accompanied by broad immune cell infiltration. EPS3 displayed abnormalities in respiratory chain and endocytosis, together with cytotoxic and T follicular helper (Tfh) cell–associated signals. qPCR analysis in Angiotensin II–treated cardiomyocytes validated the subtype-specific genes, further supporting subtype stability. This gene interaction network–based taxonomy identifies distinct DCM subtypes with differential therapeutic responsiveness, particularly highlighting that EPS1 patients may benefit from IA/IgG therapy. The classification reveals immune and inflammatory heterogeneity that can guide personalized treatment strategies and improve patient stratification in DCM management.