Prediction of new-onset atrial fibrillation in patients with hypertrophic cardiomyopathy using proteomics profiling

Abstract

Abstract Background Hypertrophic cardiomyopathy (HCM) is one of the most common genetic cardiac disorders and affects 1 in 200–500 individuals. HCM is known to be heterogeneous. Approximately 20–30% of patients with HCM develop atrial fibrillation (AF), which can lead to stroke and worsening of heart failure symptoms. As AF increases the risk of stroke by 8-fold, AF in HCM is a Class 1 indication for anticoagulation. Despite its high prevalence and clinical importance of early AF detection, there are no risk stratification tools available to predict new-onset AF in patients with HCM. Furthermore, it is still unknown which signaling pathways mediate AF in HCM. Proteomics profiling can determine concentrations of thousands of proteins and potentially reveal underlying molecular mechanisms of disease progression. Purpose To develop plasma proteomics-based model to predict new-onset AF in patients with HCM and to determine signaling pathways dysregulated in those who subsequently develop AF. Methods In this prospective, multi-center cohort study, we conducted plasma proteomics profiling of 5,032 proteins on 397 patients with HCM. We developed a proteomics-based random forest machine learning model to predict new-onset AF using samples from one institution (training set, n=278). We tested the predictive ability of the model using independent samples from the other institution (test set, n=119). We estimated the hazard ratio for new-onset AF using a Cox proportional hazards model comparing high- and low-risk groups as determined by the proteomics-based model. We also performed pathway analysis of proteins significantly (i.e., univariable P<0.05) associated with new-onset AF using a false discovery rate (FDR) threshold of 0.05. Results A total of 15 patients in the training set (5.4%) and 7 in the test set (5.9%) developed new-onset AF. Using the proteomics-based model developed in the training set, the area under the receiver-operating characteristic (ROC) curve to predict new-onset AF was 0.87 (95% confidence interval [CI] 0.77–0.98; Figure) in the test set. The sensitivity was 0.86 (95% CI 0.42–0.99) and the specificity was 0.77 (95% CI 0.68–0.84). In the test set, patients categorized as high-risk based on the proteomics model had a significantly higher rate of developing new-onset AF (hazard ratio 8.18; 95% CI 1.55–43.20; P=0.01). Pathway analysis revealed that the Ras-MAPK pathway was dysregulated in patients who subsequently developed AF (FDR=0.01; Table). Pathways involved in inflammation were also dysregulated. Conclusions This study serves as the first to demonstrate the ability of proteomics profiling to predict new-onset AF in patients with HCM, exhibiting dysregulation of both novel (e.g., Ras-MAPK) and known pathways in patients who subsequently experience AF. These results not only exhibit the utility of proteomics profiling for clinical risk stratification but also suggest mechanisms underlying the development of AF in HCM. Funding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): NIH/NHLBI