A290: Multi-Omics Data Fusion-Driven Predictive Model of Exercise Load, Metabolite Profiles, and Cardiovascular Effects

This study aims to construct a predictive model that integrates multi-omics data, including metabolomics, transcriptomics, and clinical exercise metrics, to elucidate the causal relationships between exercise load, dynamic metabolite profiles, and cardiovascular outcomes. This research addresses the existing knowledge gap in the optimization of personalized exercise prescriptions for metabolic and cardiovascular diseases. A longitudinal cohort of 500 participants underwent cardiopulmonary exercise testing (CPET) alongside blood metabolomics profiling, utilizing targeted lipid and amino acid panels at rest, at the anaerobic threshold, and at peak exercise. The multi-omics fusion approach included: (1) LASSO regression to identify exercise-responsive metabolites (e.g., succinate, β-hydroxybutyrate); (2) Mendelian randomization to validate causal pathways between metabolites and cardiovascular disease (e.g., branched-chain amino acids and endothelial dysfunction); and (3) graph neural networks (GNNs) to integrate metabolic pathways (e.g., PPAR/AMPK), transcriptomic signatures (e.g., PGC-1α), and echocardiographic indices (e.g., left ventricular ejection fraction, CAVI). Cross-species validation was conducted using ApoE−/− mice subjected to exercise preconditioning. The GNN model demonstrated superior predictive accuracy for cardiovascular risk stratification, achieving an area under the curve (AUC) of 0.92, compared to Framingham scores, which yielded an AUC of 0.76. Key findings include: (1) Exercise-induced elevation of β-hydroxybutyrate (β-HB) mediated 38% of the improvement in cardiac output through inhibition of HDAC3 (p < 0.001); (2) High-intensity interval training (HIIT) activated a conserved PPARα-CPT1 axis, which contributed to a reduction in atherosclerotic plaque instability (odds ratio = 0.62, 95% confidence interval: 0.51–0.74); and (3) The temporal dynamics of metabolites measured from 0 to 24 hours post-exercise outperformed single timepoint assessments in predicting endothelial function (ΔR ² = 0.21). This multi-omics framework pioneers causal inference in exercise medicine, identifying β-hydroxybutyrate and the PPARα-CPT1 axis as dual targets for cardiovascular disease prevention. It facilitates precision exercise dosing by quantifying metabolic reprogramming thresholds, thereby supporting clinical guidelines for the management of metabolic syndrome. Future research should focus on validating these thresholds in diabetic cohorts through proteomic-epigenomic integration.