Plasma Metabolomics and Machine Learning Reveals Metabolic Alterations and Diagnostic Biomarkers for Deep Venous Thrombosis in Hypertensive Patients after Traumatic Fracture

We aimed to explore the metabolic dysregulations and diagnostic biomarkers for post-traumatic deep venous thrombosis (pt-DVT) in hypertensive (HPT) patients after fracture. An untargeted ultraperformance liquid chromatography-mass spectrometry-based metabolomics approach was employed to perform a comprehensive metabolic analysis of plasma samples of 80 patients with post-traumatic deep venous thrombosis and hypertension (pt-DVT&HPT) and 117 patients with hypertension only (HPT). Thirty-seven (37) differential metabolites were identified between pt-DVT&HPT and HPT patients. Purine metabolism, citric acid cycle, sphingolipid metabolism, histidine metabolism, aminoacyl-tRNA biosynthesis, valine, leucine, and isoleucine degradation were the most significantly altered metabolic pathways in pt-DVT in HPT patients. Metabolite-protein interaction network analysis unveils ten (10) proteins/genes that could serve as therapeutic targets. Multivariate methods with Unbiased Variable selection in the R package (MUVR) algorithm identified four metabolites as novel biomarkers for pt-DVT&HPT, and receiver operating characteristics (ROC) analysis showed that a predictive model based on the four biomarkers (l-carnitine, lactic acid, adenine, and l-acetylcarnitine) exhibited better predictive capability for pt-DVT than the D-Dimer. The integration of these diagnostic biomarkers and D-Dimer increased its diagnostic potential. The generalization ability of this biomarker panel was validated in an independent cohort. This study contributed to our understanding of metabolic alterations associated with pt-DVT and paved the way for early diagnosis of pt-DVT in HPT patients.