Refining the generation, interpretation, and application of multi-organ, multi-omics biological aging clocks

Abstract Multi-organ biological aging clocks derived from clinical phenotypes and neuroimaging have emerged as valuable tools for studying human aging and disease 1,2,3,4 . Plasma proteomics provides an additional molecular dimension to enrich these clocks 5 . Building on previous work 1,3 , I developed 11 multi-organ proteome-based biological age gaps (ProtBAGs) using 2448 plasma proteins from 43,498 participants in the UK Biobank. I highlighted key methodological and clinical considerations for developing and using ProtBAG, including age bias correction 6 , and investigated the impact of training data sample size, protein organ specificity, and the underlying pathologies of the training data on model generalizability and clinical interpretability. I then integrated the 11 ProtBAGs with our previously developed 9 multi-organ phenotype-based biological age gaps (PhenoBAG 1 ) to investigate their genetic underpinnings, causal associations with 525 disease endpoints (DE) from FinnGen and PGC, and their clinical potential in predicting 14 disease categories and mortality. Genetic analyses revealed overlap between ProtBAGs and PhenoBAGs via shared loci, genetic correlations, and colocalization signals. A three-layer causal network linked ProtBAG, PhenoBAG, and DE, exemplified by the pathway of obesity→renal PhenoBAG→renal ProtBAG to holistically understand human aging and disease. Combining features across multiple organs improved predictions for disease categories and mortality. These findings provide a framework for integrating multi-organ and multi-omics biological aging clocks in biomedicine.