Subtyping strokes using blood-based biomarkers: A proteomics approach

Background and Objectives: Rapid diagnosis of stroke and its subtypes is critical in early stages. We aimed to discover and validate blood-based protein biomarkers to differentiate ischemic stroke (IS) from intracerebral hemorrhage (ICH) within 24 hours using high-throughput proteomics. Methods: We collected serum samples within 24 hours from acute stroke (IS & ICH) and mimics patients. In the discovery phase, SWATH-MS proteomics identified differentially expressed proteins (fold change: 1.5, p<0.05, and confirmed/tentative selection using Boruta random forest) between IS and ICH which were validated using Multiple Reaction Monitoring (MRM) proteomics in the validation phase. Protein-protein interactions and pathway analysis were conducted using STRING version 11 and Cytoscape 3.9.0. Cut-off points were determined using Youden Index. Prediction models were developed using backward stepwise multivariable logistic regression analysis. Hanley-McNeil test, Integrated discrimination improvement index, and likelihood ratio test determined the improved discrimination ability of biomarkers added to clinical models. Results: Discovery phase included 20 IS and 20 ICH while validation phase included 150 IS, 150 ICH, and 6 stroke mimics. We quantified 365 proteins in the discovery phase. Between IS and ICH, we identified 20 differentially expressed proteins. In the validation phase, combined prediction model including three biomarkers: GFAP (OR 0.04; 95%CI 0.02-0.11), MMP9 (OR 0.09; 95%CI 0.03-0.28), APO-C1 (OR 5.76; 95%CI 2.66-12.47) and clinical variables independently differentiated IS from ICH (accuracy: 92%, sensitivity: 96%, specificity: 69%). Addition of biomarkers to clinical variables improved the discrimination capacity by 26% (p<0.001). Subgroup analysis within 6 hours identified that GFAP and MMP9 differentiated IS from ICH with a sensitivity> 90%. Conclusions: Our study identified that GFAP, MMP, and APO-C1 biomarkers independently differentiated IS from ICH within 24 hours and significantly improved the discrimination ability to predict IS. Temporal profiling of these biomarkers in the acute phase of stroke is urgently warranted.