Identification of key signature genes and construction of a diagnostic model for heart failure based on integrative transcriptomic and eQTL-based analyses and single‑cell eQTL Mendelian randomization

Heart failure (HF) is a life-threatening end-stage syndrome of cardiovascular disease with a rapidly increasing global incidence. Its underlying biological mechanisms remain unclear, and current biomarkers lack sufficient diagnostic accuracy. Single-cell eQTLs and Mendelian randomization (MR) can reveal cell-specific causal relationships between gene expression and disease. This study aimed to combine integrative transcriptomic and eQTL-based analyses with single-cell eQTLs MR to identify key signature gene(s), establish a diagnostic model for HF, and facilitate accurate clinical application. Differentially expressed genes (DEGs) were identified from the GSE5406 dataset ( P < 0.05, and |logFC| > 0.585), followed by functional enrichment analysis. Two-sample MR analysis was performed using cis-eQTL data from GTEx V10 and eQTLGen to estimate the causal relationship between gene expression and HF. Genes overlapping between MR and DEG analyses were identified as signature genes. An artificial neural network (ANN) model and a dynamic nomogram for HF diagnosis were developed and validated using GSE84796 dataset. Immune microenvironment analysis was performed using CIBERSORT. Single-cell MR analysis was conducted using sc-eQTL data from the OneK1K cohort. A set of DEGs was identified, and overlap between MR and DEG analyses resulted in three signature genes: CORIN , MTHFD2 , and TRIM22 . The ANN model based on the three signature genes demonstrated good diagnostic performance (area under the curve = 0.862 in the training set). Immune correlation analysis demonstrated that CORIN , MTHFD2 , and TRIM22 were closely associated with macrophages, dendritic cells, and T cells, respectively. Single-cell MR analysis revealed that MTHFD2 expression in CD4 + SOX4 + T cells increased the risk of HF, whereas TRIM22 expression in monocytes exerted a protective effect. This study identified CORIN , MTHFD2 , and TRIM22 as core signature genes for HF and developed a dynamic nomogram-based diagnostic model with good performance. The findings describe cell-type-specific regulatory mechanisms and provide new insights into the immune and metabolic regulatory pathways involved in HF. These results offer mechanistic insights and a preliminary transcriptomic model for HF research. Further clinical validation incorporating routine markers such as BNP/NT-proBNP and LVEF is needed to verify the model’s incremental value before its clinical application in early diagnosis and risk stratification.