Identification of biomarkers for right ventricular dysfunction in dilated cardiomyopathy using urinary proteomics and machine learning

Background Right ventricular dysfunction (RVD) is prevalent in patients with dilated cardiomyopathy (DCM), and it is associated with a poor prognosis. However, no biomarker indicative of RVD is currently available for routine clinical use. Purpose This study aimed to identify urinary biomarkers associated with RVD in idiopathic DCM patients, utilizing urinary proteomics and machine learning techniques within a prospective cohort. Methods A total of 147 patients with idiopathic DCM were included in this study, with a median follow-up of 19.3 months (interquartile range: 10.9–28.1 months). Urinary specimens from 64 RVD patients and 83 non-RVD controls were analyzed using data-independent acquisition (DIA) high-resolution mass spectrometry (MS) to identify candidate biomarker proteins. RVD was defined as a right ventricular ejection fraction (RVEF) <45% by cardiac magnetic resonance (CMR). The least absolute shrinkage and selection operator (LASSO) combined with 10-fold cross-validation were employed to develop novel urinary proteomic biomarker panels. Results A total of 3,579 urinary proteins were identified and quantified through MS analysis. Forty-six proteins exhibited significant differential expression (fold change >1.5 or <0.67, false discovery rate [FDR] <0.05) between RVD and non-RVD patients. After adjusting for clinical variables using multiple logistic regression and selecting variables with LASSO, a final 3-biomarker diagnostic panel (RARRES1, MVB12B, GSK3A) was developed, which could distinguish RVD from non-RVD in DCM. This 3-biomarker panel demonstrated an area under the curve (AUC) of 0.946 (95% confidence interval [CI], 0.908–0.983) in the training dataset (n=109), compared to an AUC of 0.676 (95% CI, 0.574–0.777) for N-terminal pro B-type natriuretic peptide (NT-proBNP). In the validation dataset (n=38), the AUC for the biomarker panel was 0.935 (95% CI, 0.878–0.996), significantly outperforming NT-proBNP (AUC = 0.662, 95% CI, 0.483–0.841). Reduced urinary levels of these 3 biomarkers were significantly associated with a higher risk of cardiovascular death or heart failure admission (hazard ratio, 3.24 [95% CI, 1.56–6.71]; P = 0.001). Conclusion A 3-biomarker urinary proteomic panel can reliably distinguish RVD from non-RVD in DCM with high sensitivity and specificity, and it can predict adverse clinical outcomes, offering potential for improved diagnostic and prognostic assessment in this patient population.