Exploring biomarkers of orange consumption in human urine through untargeted metabolomics using dual-column UHPLC-MS

Existing biomarkers for fruit intake have not been well validated owing to the limited coverage of metabolomic platforms and lack specificity when used individually. In a highly controlled cross-over intervention study with 19 healthy volunteers on orange biomarker discovery, metabolomics analysis was performed on orange fruit, its in vitro digested samples, and postprandial urine using dual-column (Amide and T3) ultra-high-performance liquid chromatography-mass spectrometry (UHPLC-MS). Based on an intensive statistical strategy that incorporated the urinary kinetic curve (T_{0 h}, T_{2 h}, T_{4 h}, T_{6 h}, T_{24 h}) of over 30,000 features, we confirmed three previously reported biomarkers (achillein, hesperetin, ferulic acid), and identified nine novel candidates that remained elevated after 24 h of consumption. Multi-biomarker panels using XGBoost achieved high predictive performance in classifying orange consumers after 1-2 days free diet background (single: 0.72 vs. panel: 0.85-0.90). Further application of dual-column UHPLC-MS and multi-biomarkers may accelerate the discovery and application of food biomarkers.