Metabolomic network reveals novel biomarkers for type 2 diabetes mellitus in the UK Biobank study

Aims: To identify hub metabolic biomarkers that constructively shape the type 2 diabetes mellitus (T2DM) risk network.

Materials and methods: We analysed data from 98 831 UK Biobank participants, confirming T2DM diagnoses via medical records and International Classification of Diseases codes. Totally 168 circulating metabolites were quantified by nuclear magnetic resonance at baseline. Metabolome-wide association studies with Cox proportional hazards models were performed to identify statistically significant metabolites. Network analysis was applied to compute topological attributes (degree, betweenness, closeness and eigencentrality) and to detect small-world features (high clustering, short path lengths). Identified metabolites were used with XGBoost models to assess risk prediction performance.

Results: Over a median 12-year follow-up, 114 metabolites were significantly associated with T2DM risk and clustered into three distinct small-world modules. Total triglycerides and large high-density lipoprotein (HDL) cholesterol emerged as the pivotal biomarkers in the 'risk' and 'protective' modules, respectively, as evidenced by their high eigencentrality. Moreover, total branched-chain amino acids (BCAAs) exhibited small-world network characteristics exclusively in pre-T2DM individuals, suggesting them as a potent early indicators. GlycA demonstrated high closeness centrality in females, implying a female-specific risk biomarker.

Conclusions: By constructing a metabolic network that captures the complex interrelationships among circulating metabolites, our study identified total triglycerides and large HDL cholesterol as central hubs in the T2DM risk metabolome network. BCAA and GlycA emerged as alarm indicators for pre-T2DM individuals and females, respectively. Network analysis not only elucidates the topological functional roles of biomarkers but also addresses the limitations of false positives and collinearity in single-metabolite studies, offering insights for metabolic pathway research and precision interventions.