Approved research

Deep Learning Model to Predict Physiological Ageing Rate

Lay summary

We aim to discover ageing related factors from the multitude of variables profiled in the UK Biobank. Aim 1. Reapply our Deep Learning methodology to an independent electronic medical records (EMR) dataset. We will employ our original Deep Learning model to predict chronological age as a proxy for physiological age utilizing the biomarker data from the UK Biobank. We will interpret the predictive model by investigating the contribution of biomarkers to the prediction of physiological ageing, and compare these results to our original study. Aim 2: Perform association study to discover how physiological, clinical and genetic factors affect ageing rate. Ageing is a risk factor for many complex diseases, whereby ageing rate varies across individuals. This research will use the UK Biobank's data to: 1) Validate and improve our previous physiological ageing predictive model; 2) Identify physiological, clinical and genetic factors that affect physiological ageing rates across individuals. The predictive model will be helpful for indicating general health status for individual patients. All together, the new knowledge gained from this study can inform new opportunities in personalized healthcare. We will use biomarker data from all individuals to build a predictor capable of estimating the physiological age of patients. Such estimated physiological age will enable us to identify individuals with abnormal ageing rates. We will then perform association analyses to identify genetic, clinical and environmental factors affecting the divergent physiological ageing rates for these individuals. The full cohort will be needed for our analysis/project.