August 4th, 2026
A broad variety of aging clocks have been created in recent years. Such a clock is produced via machine learning techniques applied to any sufficiently complex set of biological data that tends to change with age. Everything from imaging to blood chemistry to omics data sets can and has been used for this purpose. A reference data set is processed to derive combinations of measurements that predict chronological age, or mortality risk, or some other output. A good clock then produces similar results in other data sets. A potentially useful clock also has the characteristic that a predicted clock age higher than chronological age correlates with a greater risk of mortality and age-related disease.
It is proposed that aging clocks are a measurement of biological age. If starting out with the very simple concept that biological aging is an increase in the risk of mortality due to intrinsic causes, then clocks that show correlation between clock age and mortality risk can reasonably be considered a first step in that direction. At any more detailed level of inquiry, however, it becomes a great deal less clear as to whether any given clock is actually decent measure of biological age. It is also difficult to gain consensus on how exactly to define biological age in any more detailed way. This is in part because aging is very complex. Any given clock is probably only sampling the consequences of some of the mechanisms involved. Can we trust that any given clock will correctly predict the outcome of a therapy that only affects one mechanism of aging, such as a senolytic drug that clears senescent cells from aged tissues? Not without actually running a lengthy study to find out.
