Abstract WMP32: Cross-Platform Proteomics and Machine Learning Algorithms Nominate Biomarkers of Stroke Diagnosis: An Exploratory Study

• DOI: 10.1161/str.56.suppl_1.wmp32
• Publication Date: 2025-01-30T10:18:44Z
• Authors: Shubham Misra; Aditya Natu; Prateek Kumar; Caroline Watson; Michael Frankel; Srikant RANGARAJU
• Keywords: NOMINATE; Stroke

Introduction: Rapid stroke diagnosis is critical in the early stages to initiate subtype-specific treatment soon after symptom onset. Objectives: We undertook a cross-platform proteomics study discover plasma biomarkers of emergency room (ER) settings. Methods: analyzed clinical and data from patients aged ≥18 years using prospective repository 2010 2014. Blood samples were collected at admission ER before any therapeutic intervention. Our outcomes differentially expressed protein (DEP) levels between with acute ischemic (AIS), intracerebral hemorrhage (ICH), transient attack (TIA), mimics (MIM). performed aptamer-based 7K SomaScan assay. For pairwise comparisons, we identified DEPs ±1.5-fold change unadjusted p-value <0.05 cut-offs, Boruta random forest feature selection, variance partitioning analyses. top proteins conducted multivariable logistic regression multigroup selection by sparse partial least squares discriminant analysis (sPLS-DA) mixOmics R package. internal validation on same PeptiQuant Plus biomarker assessment kits (BAK-270) for targeted quantitation (Figure 1). Results: included 100 (mean age 58.6 years, 43% males) classified into four subgroups: 40 AIS, 20 ICH, TIA, MIM 2). quantified 7307 somamers targeting 6373 unique proteins. Using nominated 58 that differentiated subtypes. panel 7 as AIS classifiers (area under curve (AUC): 0.82, negative predictive value (NPV): 74%), 5 ICH (AUC 0.88, NPV: 90%), 8 0.94, 94%), 6 TIA 91%) 3). In phase, validated VTN PLG against TIA. Conclusions: exploratory highlights valuable tool discovering diagnosis. Further research warranted validate these findings larger multi-center cohorts elucidate their utility settings guiding decision-making improving patient outcomes.