Differentiation between descending thoracic aortic diseases using machine learning and plasma proteomic signatures

Area of Science:

• Cardiovascular Research
• Proteomics
• Machine Learning Applications

Background:

• Descending thoracic aortic aneurysms and dissections pose significant health risks.
• Limited clinical tools exist for early screening or risk prediction of these conditions.

Purpose of the Study:

• To investigate the utility of plasma proteomic analysis combined with machine learning.
• To differentiate between descending thoracic aortic aneurysm (DTAA) and type B aortic dissection (Type B) cases.

Main Methods:

• Generated plasma proteomic data from 75 Type B and 62 DTAA patients.
• Compared statistical methods with supervised machine learning (ML) algorithms.
• Utilized hierarchical clustering and ML models with hyperparameter optimization.
• Employed permutation importance (PI) to rank predictor proteins and pathway analysis.

Main Results:

• Only one protein, hemopexin (HPX), showed significant difference (p < 0.01) between groups using standard statistics.
• A Support Vector Classifier achieved a precision-recall area under the curve of 0.7 for classification.
• Top PI proteins included immunoglobulin heavy variable genes, LCAT, F12, and HPX.
• Enriched pathways in Type B patients involved complement activation, humoral immune response, and blood coagulation.

Conclusions:

• Machine learning can distinguish highly similar disease states based on plasma proteomic data.
• ML aids in prioritizing key proteins for diagnostic model development.
• Identified pathways suggest roles for immune response and coagulation in Type B dissection.