Differentiation between Descending Thoracic Aortic Diseases using Machine Learning and Plasma Proteomic Signatures

Area of Science:

• Cardiovascular Medicine
• Proteomics
• Machine Learning

Background:

• Descending thoracic aortic aneurysms and dissections are severe conditions lacking effective screening tools.
• Few clinical indices exist to predict the risk of aortic dissection.

Approach:

• A plasma proteomic dataset was generated from patients with type B dissection and descending thoracic aortic aneurysm.
• Supervised machine learning algorithms were compared with standard statistical methods.
• Proteins were clustered, and machine learning models were trained and optimized using cross-validation.

Key Points:

• Standard statistics identified only one significantly different protein (hemopexin) between the two conditions.
• Machine learning achieved a precision-recall area under the curve of 0.7 for classification.
• Top predictive proteins included immunoglobulin heavy variable genes, lecithin-cholesterol acyltransferase, coagulation factor 12, and hemopexin.

Conclusions:

• Machine learning can differentiate highly similar disease states based on plasma proteomic data.
• This approach aids in prioritizing important proteins for predictive modeling.
• Machine learning offers a novel strategy for identifying biomarkers in complex cardiovascular diseases.