Background: Venous congestion (VC) sets in weeks before visible clinical decompensation, progressively increasing cardiac strain and leading to acute heart failure (HF) decompensation. Currently, the field lacks a universally acknowledged gold standard and early detection methods for VC.
Methods: Using data from the GEO database, we identified VC's impact on HF through key genes using Limma and STRING databases. The potential mechanisms of HF exacerbation were explored via GO and KEGG enrichment analyses. Diagnostic genes for acute decompensated HF were discovered using LASSO, RF, and SVM-REF machine learning algorithms, complemented by single-gene GSEA analysis. A nomogram tool was developed for the diagnostic model's evaluation and application, with validation conducted on external datasets.
Results: Our findings reveal that VC influences 37 genes impacting HF via 8 genes, primarily affecting oxygen transport, binding, and extracellular matrix stability. Four diagnostic genes for HF's pre-decompensation phase were identified: SMOC2, OGN, FCN3, and SERPINA3. These genes showed high diagnostic potential, with AUCs for each gene exceeding 0.9 and a genomic AUC of 0.942.
Conclusions: Our study identifies four critical diagnostic genes for HF's pre-decompensated phase using bioinformatics and machine learning, shedding light on the molecular mechanisms through which VC worsens HF. It offers a novel approach for clinical evaluation of acute decompensated HF patient congestion status, presenting fresh insights into its pathogenesis, diagnosis, and treatment.
Keywords: bioinformatics; heart failure; machine learning; nomogram; venous congestion.
© 2024 Yu, Yuan, Han, Shi, Liu, Xue and Li.