Abstract WP289: Plasma Proteomics Nominate Biomarker Candidates Of Atrial Fibrillation In Stroke Patients

Introduction: Atrial fibrillation (AFib) is a major risk factor for ischemic stroke (IS). Early AFib diagnosis is critical to optimize secondary prevention and reduce recurrent stroke risk. Objectives: We performed a proteomics study to identify plasma biomarkers associated with a history of AFib in patients with IS or TIA. Methods: We used clinical and proteomics data of stroke patients aged ≥18 years lodged within prospective plasma repositories at Grady Memorial Hospital (GMH) and Yale New Haven Hospital (YNHH) from 2010 to 2020. Plasma was collected at emergency room presentation before any intervention at GMH and within 72 hours of stroke onset at YNHH. We measured differentially abundant proteins (DAPs) between patients with or without AFib history using aptamer-based 7K SomaScan proteomics. Batch effects between centers were corrected using the ComBat function ( sva package in R). The DAPs were identified using ±1.5-fold change and unadjusted p-value <0.05 (Welch’s t-test) and Boruta random forest-based feature selection algorithm. We then adjusted for clinical covariates in multivariable logistic regression. Next, regularized least absolute shrinkage and selection operator (LASSO) logistic regression was applied to the adjusted DAPs to identify a panel that discriminates AFib from no AFib, with its performance assessed using the receiver operating characteristic (ROC) curves. Pathway analysis was conducted using the Ingenuity Pathway Analysis tool. Results: Among 124 patients with IS/TIA (mean age 65.5 years, 54.8% males), 29 had AFib and 95 had no AFib (Figure 1). SomaScan quantified 7301 proteins. We identified 45 DAPs, with 37 upregulated and 8 downregulated in AFib compared to no AFib (Figure 2A, B). Of these, 40 DAPs were significantly associated with AFib in IS/TIA patients after adjusting for age, sex, and coronary artery disease in the multivariable analysis (adjusted p<0.05). LASSO regression identified a panel of 12 proteins that distinguished AFib from no AFib (AUC: 0.95, sensitivity: 79%, specificity: 88%, negative predictive value: 93%) (Figure 2C-F). Pathways associated with collagen degradation and fibrin formation were activated in AFib (Figure 3). Conclusions: Our study highlights the potential of plasma proteomics as a valuable tool for discovering protein biomarkers to discriminate IS/TIA patients with AFib history compared to no AFib. Further longitudinal studies with adequate sample sizes are needed to support these findings.