Development of Circulating Protein–Metabolite Signaling Network

Understanding how circulating proteins and metabolites co-vary at the population level is essential for characterising systemic physiology and for interpreting how diseases perturb biological signalling. Yet, most studies analyse proteomics and metabolomics separately, leaving the structure of their joint variation—and the signalling modules that connect them—largely unresolved. This thesis addresses the problem of how to construct a statistically principled, demography-adjusted circulating protein–metabolite signalling network that can serve as a reference framework for future biomedical and epidemiological research. We integrate large-scale affinity-based plasma proteomics measured using the Olink platform (approximately 3,000 proteins) together with nuclear magnetic resonance(NMR)–based plasma metabolomics (covering lipids, lipoprotein subclasses, amino acids,and related metabolic traits) obtained from a population-based cohort (blood samples) comprising over ten thousand individuals after quality control. Using these data, we develop a fully transparent statistical workflow that combines: (i) rigorous preprocessing and demographic adjustment; (ii) large-scale association discovery with false-discovery rate control; (iii) bipartite network construction and community detection; and (iv) multivariate integration through classical, ridge-regularised, and sparse canonical correlation analysis. The methods are chosen to maximise statistical robustness, interpretability, and reproducibility in a high-dimensional setting. The resulting network reveals coherent biological modules—most prominently lipid transport, inflammatory signalling, and amino-acid metabolism—and identifies multivariate axes of shared protein–metabolite variation that are stable across CCA formulations. Collectively, the thesis provides a statistically grounded blueprint for quantifying cross omic structure in human populations and offers a reproducible foundation for future work on mechanistic pathways, risk stratification, and disease-related perturbations.