Circulating Amino Acid Network Remodeling Reveals Systemic Metabolic Reprogramming Predictive of Colorectal Cancer Recurrence and Metastasis

Abstract

Characterizing the dynamics of systemic metabolic pathways during cancer progression could enable the development of prognostic tools and therapeutic strategies. However, achieving this goal requires analytical platforms optimized for liquid biospecimens together with interpretive frameworks capable of identifying robust serum-based biomarkers. Here, we establish a network-based metabolic profiling framework using 1 9F NMR-based serum amino acid analysis to characterize systemic metabolic remodeling during colorectal cancer (CRC) progression. Using an analytical protocol optimized for clinical serum samples, we quantified circulating amino acids from 152 CRC patients and implemented a ratio-based normalization strategy to mitigate cohort variability in concentration-based approaches. We systematically evaluated both individual amino acid changes and correlation structures across tumor stages. Advanced-stage CRC exhibited a distinct metabolic shift characterized by decreased valine and increased glycine levels. Correlation network analysis further revealed stage-dependent remodeling of circulating amino acid interactions, leading to the emergence of a glycine-centered metabolic architecture. Importantly, machine-learning models integrating individual amino acid levels with network-derived features significantly improved the prediction of recurrence or metastasis compared with models using either feature type alone and outperformed conventional biomarker carcinoembryonic antigen (AUROC = 0.806). These findings highlight the remodeling of circulating amino acid network as a promising strategy for prognostic stratification and postoperative monitoring in CRC.