Subtyping strokes using blood-based protein biomarkers: A high-throughput proteomics and machine learning approach

Area of Science:

• Biochemistry
• Proteomics
• Neurology

Background:

• Accurate and rapid stroke diagnosis is crucial for timely treatment.
• Distinguishing between ischemic stroke (IS) and intracerebral hemorrhage (ICH) is essential for appropriate management.
• High-throughput proteomics offers a promising avenue for discovering blood-based biomarkers.

Purpose of the Study:

• To discover and validate blood protein biomarkers for differentiating IS from ICH.
• To develop predictive models for stroke subtype classification using identified biomarkers.
• To assess the diagnostic accuracy and clinical utility of novel biomarkers.

Main Methods:

• Serum samples from IS, ICH, and stroke mimic patients were analyzed using SWATH-MS and targeted proteomics.
• Differentially expressed proteins were identified in the discovery phase and validated in a larger cohort.
• Multivariable logistic regression models were developed incorporating biomarkers and clinical variables.

Main Results:

• Twenty differentially expressed proteins were identified between IS and ICH in the discovery phase.
• A combined model using GFAP, MMP-9, APO-C1, and clinical variables achieved 92% accuracy in differentiating IS from ICH.
• Biomarkers significantly improved the discrimination ability of prediction models by 26%.

Conclusions:

• GFAP, MMP-9, and APO-C1 are effective biomarkers for differentiating IS from ICH within 24 hours.
• These biomarkers enhance the accuracy of diagnostic models for stroke subtypes.
• Further investigation into the temporal dynamics of these biomarkers in acute stroke is recommended.