Systems Biology and Machine Learning Decode an Immunometabolic Signature for Post-Thrombotic Syndrome

ObjectivePost-thrombotic syndrome (PTS), a common complication of deep vein thrombosis, lacks objective diagnostic biomarkers and its molecular mechanisms remain poorly understood. This study aimed to identify plasma biomarkers and clarify pathways using integrated multi-omics and machine learning. MethodsProteomic and metabolomic profiling of 75 PTS patients and 75 controls was performed. Differential expression analysis, pathway enrichment, and protein-metabolite network analysis were conducted. A multi-algorithm machine learning with 8 feature selection methods prioritized biomarkers. Validations and 14 models were assessed. Results1,104 proteins and 1,891 metabolites were differentially expressed. Citrate cycle and unsaturated fatty acid biosynthesis were enriched. Three proteins, namely DIP2B, KNG1, and SUCLG2, were consistently selected as core biomarkers. All of these proteins were significantly downregulated in PTS and externally validated. A random forest model utilizing these proteins achieved an accuracy of 97.7% in independent testing, with SUCLG2 being the most influential predictor. ConclusionThis study identifies a novel three - protein biomarker panel for the diagnosis of PTS and reveals an immunometabolic axis in the pathogenesis of PTS, which links inflammatory regulation with mitochondrial energy metabolism. These findings provide valuable insights into the development of diagnostic tools and targeted therapeutic approaches.