Integrating DNA Methylation Surrogates of Clinical, Functional, and Metabolomic Biomarkers into a Systems or Disease-specific Framework for Biological Ageing

This project will develop DNA methylation (DNAm)-based surrogates of clinical, functional, metabolomic, proteomic and other molecular biomarkers, and integrate them into interpretable physiological system- and disease-specific predictors of biological ageing. The objectives are to improve prediction of mortality, frailty, disability, cognitive decline, ADRD/dementia and other age-related diseases, and to identify pathways underlying multisystem ageing.

The study will use an observational cohort design. HRS will serve as the primary training dataset because it links DNAm profiles with clinical, functional, genetic and longitudinal health data. Other DNAm datasets will be used as external validation cohorts to assess generalisability across populations and study settings.

Analyses will include epigenome-wide association and differential methylation analyses to identify CpGs associated with target biomarkers and ageing-related phenotypes. Penalised regression, primarily elastic net with nested cross-validation, will be used to train DNAm surrogate models. Other machine-learning methods may be explored where appropriate. DNAm surrogates will be grouped into physiological systems or disease domains using biological annotation, clustering and model-based approaches. The resulting DNAm proxies and integrated scores will be evaluated in association with mortality, multimorbidity, frailty, disability, functional performance, cognitive performance, cognitive decline, ADRD/dementia and other age-related diseases. Where genotype data are available, genome-wide variants and polygenic risk scores will be assessed in relation to DNAm surrogates and ageing-related phenotypes to evaluate genetic contributions to biological ageing and disease susceptibility.

Restricted data will be analysed only within approved secure environments. Exported results will be aggregate only, such as regression coefficients, hazard ratios, odds ratios, summary statistics and model performance metrics. No individual-level data, direct identifiers or small-cell outputs will be exported.

Planned collaborators include Prof Andrea B. Maier and researchers at the Academy for Healthy Longevity, NUS.

Ageing affects many parts of the body, but current biological ageing tests do not fully capture this complexity. This project aims to develop new DNA methylation-based measures that reflect different aspects of health, including clinical markers, physical function, metabolism and age-related diseases. DNA methylation is a chemical mark on DNA that can provide information about biological ageing.

The Health and Retirement Study will be used as the main dataset to develop these measures because it includes DNA methylation data linked with long-term health information. Other DNA methylation datasets will be used to test whether the measures work well in different populations.

The goal is to create more accurate and interpretable tools for studying ageing, predicting risks such as frailty, disability, dementia and mortality, and understanding how different body systems age together. Findings will be reported only as summary results, with no release of individual-level participant data.