M. Darooghegi Mofrad, B. Nosal, J. Bi and OK. Chun (2026). Blood Metabolomic and Proteomic Signatures of Total, Healthy, and Unhealthy Low-Carbohydrate Diets in the UK Biobank. Journal of the American Nutrition Association, 45(5), 397-411. https://doi.org/10.1080/27697061.2025.2588435

PubMed 41609508

[Abstract]
OBJECTIVES: Understanding the biological pathways linking low-carbohydrate diets (LCDs) to cardiovascular diseases (CVDs) is complex due to heterogeneity in LCD patterns, limited clinical biomarkers that reflect underlying mechanisms, and substantial interindividual variability. This study aimed to identify plasma metabolomic and proteomic signatures of total, healthy, and unhealthy LCDs to elucidate biological pathways associated with these LCDs.

METHODS: We conducted a cross-sectional analysis of 57,199 participants with metabolomics data and 9,895 participants with proteomics data without CVD at baseline from the UK Biobank. The LCD adherence was assessed using multiple validated 24-h dietary recalls. Plasma metabolites were profiled using high-throughput nuclear magnetic resonance spectroscopy, and protein expression levels were quantified using the antibody-based Proximity Extension Assay by Olink. Elastic net regression was applied to identify the metabolomic and proteomic signatures of total, healthy, and unhealthy LCDs. Pathway overrepresentation analysis was conducted for LCD-related proteins.

RESULTS: A total of 29, 36, and 89 metabolites were associated with total, healthy, and unhealthy LCDs, respectively, including lipoproteins, fatty acids, lipids, ketone bodies, amino acids, inflammation, fluid balance, and glycolysis metabolites. Pearson correlations between total, healthy, and unhealthy LCD scores and the metabolomic signatures were 0.177, 0.302, 0.254 in the training set and 0.159, 0.293, and 0.250 in the test set, respectively (p < 0.001). For the proteomic analysis, 89, 62, and 163 proteins were associated with total, healthy, and unhealthy LCDs, respectively with corresponding Pearson correlations of 0.343, 0.325, 0.396 in the training set and 0.294, 0.284, and 0.337 in the test set, respectively (p < 0.001). Six unique biological pathways were significantly enriched by LCD-related proteins.

CONCLUSIONS: Distinct plasma metabolomic and proteomic profiles were associated with adherence to total, healthy, and unhealthy LCDs in UK adults. Further study is warranted to clarify the link between LCDs, metabolomic and proteomic profiles, and pathogenesis of CVDs.