Machine learning integrated clinical-proteomics data identifies a 6-protein panel signature for atherosclerotic...

Area of Science:

• Cardiovascular Research
• Biomarker Discovery
• Machine Learning in Medicine

Background:

• Atherosclerosis is a leading cause of global mortality, driven by lipid accumulation and arterial wall inflammation.
• Early identification of at-risk patients is critical for preventing thrombotic events and enabling personalized treatments.

Purpose of the Study:

• To leverage machine learning (ML) for enhanced diagnostics and biomarker discovery in atherosclerosis.
• To identify a robust protein signature for discriminating atherosclerosis severity using clinical and proteomic data.

Main Methods:

• Applied five ML classification algorithms to clinical and serum proteomic data from patients with carotid atherosclerotic stenosis (AT), dyslipidemic patients (DLP), and healthy controls (HC).
• Integrated clinical and proteomic data for improved patient stratification compared to individual analyses.
• Validated the identified protein panel in an independent cohort of patients with acute atherothrombotic stroke.

Main Results:

• Identified a 6-protein panel (B2M, GPV, MMP9, PLF4, TSP1, and FB isoforms) with high diagnostic accuracy (ROC-AUC > 0.9) for discriminating AT patients.
• The combined clinical-proteomic ML approach demonstrated superior patient stratification capabilities.
• Validated the panel's potential as a biomarker for atherosclerosis severity in an external cohort.

Conclusions:

• The 6-protein panel serves as a promising biomarker for assessing atherosclerosis severity and identifying at-risk individuals.
• The identified biomarkers implicate platelet activation, angiogenesis, and intraplaque hemorrhage in atherosclerosis.
• Highlights the need for multipathway therapeutic strategies to prevent adverse thrombotic events in atherosclerosis.