Leveraging epigenetic biomarker proxies for precision medicine

Background. Clinical and molecular biomarkers are central to disease risk stratification, but their direct measurement remains fragmented across cohorts and analytical platforms, limiting scalability and reuse in population-scale studies. Results. We developed DNAm-based surrogates, termed epigenetic biomarker proxies (EBPs), in 4,418 participants from the Massachusetts General Brigham Aging Biobank Cohort (MGB-ABC). Using mutual-information feature selection followed by elastic net regression, with gradient boosting for poorly fitted targets, we retained 1,694 EBPs with a Spearman correlation > 0.2 with their target in held-out data: 42 clinical laboratory traits, 689 metabolites and 963 proteins. We validated EBPs in seven independent cohorts totalling 57,454 participants, spanning paediatric to older-adult populations, differing disease burdens and distinct biospecimen and molecular profiling platforms; biomarker correlations, disease associations and clinical-range detection were broadly replicated. Conclusions. EBPs can be constructed at scale from a single DNAm assay and capture biologically and clinically relevant biological variation across clinical, proteomic and metabolomic domains. Their combination of modest cross-sectional concordance with robust disease association suggests that EBPs may capture biological information that is related to, but not identical to, contemporaneous biomarker concentrations. EBPs provide complementary harmonised proxies for molecular phenotyping and population-scale research, rather than as replacements for direct biomarker measurement.