June 16th, 2026
Machine learning approaches can be used to create aging clocks from near any set of biological data collected from people of various ages. The techniques are well established and many new clocks are published every year. A clock is really an age predictor (or a mortality predictor, or a predictor of some other outcome) trained on a single dataset. When the clock algorithm is applied to any given individual not in that data set, it is thought that the predicted age or mortality risk or other outcome is some reflection of biological age. It is hard to validate this proposition, as there is very little concrete connection between any easily measured biomarker and mechanisms of aging, and indeed all too little consensus on how to measure biological age in the first place. To my eyes more effort should go towards understanding the clocks we have and less to producing new clocks.
Biological aging is a key determinant of liver disease and mortality, but there is little evidence on noninvasive index for assessment of liver biological aging. We developed the Liver Aging Index (LAI) in the China Kadoorie Biobank (CKB, N = 21,629) using Cox-Gompertz proportional hazards model. The LAI incorporated three clinical factors (body mass index, systolic and diastolic blood pressure), eight plasma biomarkers (glucose, total cholesterol, triglycerides, high-density and low-density lipoprotein cholesterol, alanine aminotransferase, aspartate aminotransferase, and γ-glutamyl transpeptidase), and two imaging biomarkers (fat attenuation parameter and liver stiffness measurement).
