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