PROTEOMICS IN HEART FAILURE: RISK PREDICTION AND BIOMARKER DISCOVERY

Abstract

Heart failure remains an important clinical question despite significant advances in treatment. 1 Healthcare providers can intervene early and personalize treatment plans by predicting a patient's risk of developing heart failure or worsening symptoms. 2 Identifying new biomarkers can help guide drug development and improve the effectiveness of treatment options for heart failure patients. 3 Previous studies have revealed that machine learning (ML) -based models that consider proteomics as predictors can make more precise time-to-event predictions than traditional regression models and facilitate biomarker identification. 4,5 In our first manuscript, we investigated whether machine learning will improve risk prediction of cardiovascular death (CVD) and heart failure hospitalization (HFH) using proteomic predictors in heart failure with reduced ejection fraction (HFrEF) patients and generalize to heart failure with preserved ejection fraction (HFpEF). We then worked towards discover important proteins and pathways involved in the high performing models. In our second manuscript, we aimed to determine if a proteomic risk score derived in HFpEF patients is superior to a previous proteomic score derived in HFrEF, as well as clinical risk factors, NT-proBNP, or high-sensitivity cardiac troponin, and identify serum proteins associated with the risk of HFH and CVD in patients with HFpEF in the proteomic risk predictive model. We then analyzed the most influential prognostic proteins in HFpEF for clinical outcome prediction.