Machine Learning Identifies Plasma Proteomic Signatures of Descending Thoracic Aortic Disease.

Descending thoracic aortic aneurysms may go undetected until severe and catastrophic complications such as rupture or dissection occur. Few clinical indicators exist to screen for aneurysms or predict risk of dissection or rupture. This study generated a plasma proteomic dataset from 150 patients with descending thoracic aortic disease and 52 controls to identify proteomic signatures capable of differentiating descending thoracic aortic disease from controls, as well as aneurysm from descending aortic dissections. Of the 1,468 peptides and 195 proteins quantified, 853 peptides and 99 proteins were significantly different between disease and controls. Using machine learning to classify disease from control, the highest precision-recall area under the curve (PR AUC) was achieved using significantly different proteins (PR AUC 0.99), followed by significant peptides (PR AUC 0.96). Despite no statistically significant protein signatures between aneurysm and dissection, use of all proteins was able to modestly classify between the disease states (PR AUC 0.77). To address correlation between proteins, a disease versus control classifier was optimized using only seven unique protein clusters, which achieved comparable performance to models trained on all/significant proteins (PR AUC 0.90). Model interpretation with permutation importance revealed that proteins differentiating disease and control function in coagulation, protein-lipid complex remodeling, and acute inflammatory response.