Baseline-referenced plasma metabolomic trajectories during early post-exercise recovery: A secondary analysis of public human exercise metabolomics data

BACKGROUND: Post-exercise metabolomic changes are difficult to interpret without reference to the pre-exercise state. A baseline-referenced approach may help distinguish transient exercise responses from residual metabolic deviation during early recovery. METHODS: We performed a secondary analysis of two public human plasma metabolomics datasets. ST001789 was used as the primary discovery dataset and included 57 paired plasma samples from 19 participants collected before exercise, immediately after a 90-min endurance run, and after 60 min of recovery. ST004303 was used as an external support dataset and included moderate-intensity and sprint-interval exercise conditions. We assessed paired baseline-referenced contrasts, recovery-pattern classes, residual baseline deviation, cross-dataset support, intensity-related behavior, and threshold sensitivity to prioritize candidate metabolites for future validation. RESULTS: In ST001789, 198 metabolites differed from baseline immediately after exercise, 140 differed after 60 min of recovery, and 116 changed between the immediate post-exercise and recovery samples. Recovery-pattern classification showed heterogeneous early responses. Of 77 discovery features matched in ST004303, 74 had at least one direction-consistent external comparison and 66 had at least one FDR-supported external change. Staged prioritization identified 54 primary candidates. Corrected sensitivity schemes retained 136, 25, and 11 features, with 11 retained across all three schemes. These candidates spanned fatty-acid transport and oxidation, amino acid metabolism, lipid remodeling or signaling, and purine/nucleoside turnover. CONCLUSIONS: Baseline return, residual deviation, external support, and intensity-related behavior help organize early post-exercise metabolomic responses. The identified metabolites should be interpreted as hypothesis-generating candidates for future validation, not as validated recovery biomarkers.