Aging clocks are produced from machine learning strategies applied to databases of biological data, typically omics data of various sorts, obtained from people of various ages. Patterns that change with age can be identified and an algorithm defined to take any other person's data and predict their age based on comparisons to the reference database. Whether the predicted age is higher or lower than chronological age says something about the individual's biological age, the accumulation of damage and dysfunction in tissues and systems.
The biggest challenge in using these clocks is that the method of production tells us nothing about how exactly the data used in the algorithm is connected to particular processes or dysfunctions of aging. Thus it is hard to trust the outcome, particularly if the intent is to use clock measures to assess potential interventions that might slow or reverse aspects of aging. The clock may underestimate outcomes, overestimate outcomes, or just produce completely irrelevant results for any specific individual, and we have no good way of knowing which of these is the case.
This issue is well understood by the research community, and there are a number of different approaches that might be taken to improve the situation. Researchers have, for example, built clocks based on clinical measures such as blood counts and inflammatory cytokine levels rather than omics data. This is still not ideal, as the details of the connection between clinical measures and mechanisms of aging remain somewhat nebulous in most cases, but one can at least theorize on what is going on under the hood to a greater degree. Another, much harder approach is to start over and develop the means of building new omics clocks that are, from the ground up, manufactured with the intent of providing greater insight into underlying mechanisms. That work continues, but research groups are producing incremental progress along the way, such as the interpretable clock reported in today's open access paper.
Aging is the strongest risk factor for chronic diseases such as cardiovascular disease, Alzheimer's, and cancer. DNA methylation (DNAm) clocks offer a promising measure of biological age, but most rely on linear models that miss non-linear dynamics and CpG interactions. To address this, we developed a deep neural network (DNN)-based DNAm clock trained on 29,167 samples profiled on Illumina EPIC v1.0 and v2.0 arrays. Using 12,234 CpGs selected through sex- and age-stratified correlations, our model achieved high accuracy (1.89 years) and outperformed published deep learning and elastic net based epigenetic clocks in a separate validation cohort.
Using Shapley Additive Explanations (SHAP), we further uncovered phase-structured, wave-like dynamics in age-influential CpGs: an early-life module, a midlife transition, and late-life remodeling, with distinct timings by sex. These epigenetic waves cohere with non-linear, multi-omic "aging waves" reported in proteomics and longitudinal omics. SHAP further enabled interpretable CpG attribution, revealing structured, sex-specific aging phases: early-life male clocks involved developmental pathways, while female clocks emphasized cytoskeletal regulation; late-life divergence included immune activation in males and transcriptional remodeling in females. Our framework thus unites accuracy with mechanistic interpretability, revealing sex-specific windows when molecular aging reconfigures most rapidly.
View the full article at FightAging
