Identification of Biomarkers for Right Ventricular Dysfunction in Idiopathic Dilated Cardiomyopathy Via Urinary Proteomics and Machine Learning

Comprehensive Summary

In this study, researchers used machine learning to identify various differentially expressed urinary proteins in patients with idiopathic dilated cardiomyopathy (iDCM). 147 patients who met the diagnostic criteria for iDCM underwent cardiac magnetic resonance (CMR) imaging to identify right ventricular dysfunction (RVD) – a common complication of iDCM associated with all-cause mortality – and urine sampling. 64 patients were identified as having RVD, defined by a right ventricular ejection fraction < 45%. A total of 3579 different urinary proteins were identified with data‐independent acquisition‐based mass spectrometry, and after filtering and differential expression analysis, 46 proteins met the fold change criteria (>1.5 or < 0.67). After binary multivariable logistic regression analysis, 15 of those 46 proteins were found to be independently associated with RVD. Finally, the patient population was divided into training (n = 109) and validation (n = 38) sets to construct a protein panel of three proteins with the best statistical performance. The proteins RARRES1 (AUC = 0.797), GSK3A (AUC = 0.801), and MVB12B (AUC = 0.747) had a combined AUC of 0.964, outperforming other traditional iDCM markers.

Outcomes and Implications

Traditionally, RVD has been associated with extracellular remodeling or hormonal systemic mechanisms rather than intracellular processes. However, the identification of the three-protein panel may reveal an association with intracellular processes, with all three proteins playing significant intracellular roles. Additionally, the combined model AUC outperformed other biomarkers like NT‐proBNP and tricuspid annular plane systolic excursion. The higher accuracy combined with the relatively inexpensive and noninvasive method of urinary proteomics has the potential to aid clinicians in identifying patients with iDCM-RVD and improving clinical risk stratification in the future.