Serum Metabolomics Reveals Time-Dynamic Metabolic Changes and Candidate Markers Associated with Treatment Response in Acute Myeloid Leukemia

Abstract Acute myeloid leukemia (AML) is a complex disease in which genetic and molecular alterations do not always reflect functional cellular states, leading to an incomplete view of disease progression. Metabolomic analysis provides a more direct measure of these states, yet most studies remain cross-sectional and lack dynamic insight. In this study, a longitudinal serum metabolomics analysis based on proton nuclear magnetic resonance was conducted in AML patients throughout chemotherapy and in healthy controls. Temporal and response-related metabolic trajectories were analyzed using regression models, and an elastic net logistic regression model was applied to identify metabolite signatures distinguishing remission and relapse. Pathway enrichment analysis was further performed to explore the biological relevance of the metabolic alterations. Longitudinal serum metabolomics of 45 AML patients revealed chemotherapy-induced metabolic reprogramming. Linear and nonlinear temporal trajectories of 16 metabolites distinguished remission status between patient groups (p < 0.05). A subset of these metabolites was consistently selected by machine learning as candidate features that discriminated patients with remission from those with relapse (AUC = 0.909), including alanine, taurine, and valine. Overall, longitudinal serum metabolomics revealed dynamic metabolic patterns in response to chemotherapy in AML, including metabolic signatures that distinguished remission from those of relapse. These findings support the potential of dynamic metabolic profiling for monitoring treatment response and characterizing metabolic alterations associated with disease relapse.