Human Stroke Studies
Plasma protein biomarker discovery for stroke using high-throughput proteomics approaches
Copy Link
Figure 1. Human plasma proteomics to identify biomarkers of stroke diagnostic subgroups. A.1) Heatmap and A.2) PCA plot of the nominated 61 proteins to classify stroke subtypes (AIS, ICH, TIA, MIM). B) Internal validation proteomics study using BAK-270 targeted MS. C) External validation study in an independent cohort of 80 samples using DIA-MS proteomics.
Stroke diagnosis is often challenging in the acute phase. Blood-based biomarkers of stroke diagnostic groups, namely acute ischemic stroke (AIS), intracerebral hemorrhage (ICH), transient ischemic attack (TIA) or stroke mimics (MIM), can be valuable in emergency settings for triage and treatment decisions. Molecular biomarkers that reflect the severity of neuronal injury may also enhance the objectivity and accuracy of stroke severity assessment. Furthermore, protein biomarkers that can detect large vessel occlusion (LVO) strokes and inform stroke onset time can aid in rapid triage to EVT-capable centers and expand thrombolysis eligibility. Additionally, atrial fibrillation (AFib) is a major risk factor for AIS and early AFib diagnosis is critical to optimize secondary prevention and reduce the risk of recurrent stroke.
Our lab utilizes several proteomics platforms for biomarker discovery using human plasma. These range from lower-plex approaches to detect pre-specified inflammatory proteins (Luminex or Mesoscale multiplexed assays of cytokines and chemokines) and high-plex approaches to measure hundreds of proteins per sample (targeted proteomics) to thousands of proteins, such as label-free quantitative mass spectrometry (LFQ-MS) using both data-dependent and data-independent approaches and aptamer-based proteomics (SomaLogic). Using banked plasma samples from adult humans, we have applied these proteomics approaches to nominate biomarkers of stroke diagnosis.
Figure 2. Plasma proteomics for LVO detection. A. Scatter plot of common proteins across two cohorts. B. Box plot of 4 concordant proteins across two cohorts. C. Protein panel of GH2 + ACP2 added to prehospital stroke scales to classify LVO from non-LVO.
In a cross-platform proteomics study, we used aptamer-based SomaScan proteomics to measure >7,000 proteins in plasma from 100 individuals with suspected stroke diagnosis in the ED and nominated 61 proteins that differentiated between AIS, ICH, TIA, and MIM. The same 100 samples were used for a cross-platform validation study using targeted MS (270 protein biomarkers, BAK-270, MRM Proteomics), thereby validating 11 SomaScan-nominated proteins. We identified the strongest classification for stroke mimics from AIS, ICH, and TIA (e.g., VTN) with a high negative predictive value (NPV) of 88%. External validation in an independent cohort of 80 samples validated 32 proteins using untargeted proteomics by mass spectrometry (Figure 9).
In another study, we applied aptamer-based proteomics across two independent cohorts (GMH: 18 LVO, 22 non-LVO and YNHH: 34 LVO, 30 non-LVO) and identified four concordant proteins as potential biomarker candidates (GH2, C1QL2, CD2, and ACP2) that reproducibly distinguished LVO from non-LVO. These concordant proteins classified LVO independent of baseline stroke severity (NIHSS), age, sex, and other comorbidities, despite inherent differences between both cohorts. When incorporated into prediction models, concordant protein pairs including GH2 + ACP2 substantially improved discrimination (AUC ≥0.85) beyond NIHSS alone and enhanced the performance of widely used prehospital stroke scales (RACE, G-FAST, and mG-FAST) (Figure 10).
Key References
Misra S, Jang WE, Sanchez S, Natu A, Kumar P, Liu M, Kaur A, Lopez VT, Caglayan P, Garcia-Milian R, Watson CM, Frankel MR, Falcone GJ, Sansing LH, Rangaraju S. Cross-Platform Proteomics and Machine Learning Algorithms Nominate Plasma Biomarkers of Stroke Diagnosis. Journal of the American Heart Association. 2026 Mar 10:e048249. DOI: https://doi.org/10.1161/JAHA.125.048249. PMID: 41804885.