Robust biomarkers are needed both to evaluate interventions that target the aging process and to stratify individuals by disease risk.
Most aging clocks are developed and tested on different datasets with different preprocessing pipelines, which makes direct comparisons difficult.
The CASP (Critical Assessment of Structure Prediction) competitions transformed protein structure prediction and ultimately produced AlphaFold7.
Aging is heterogeneous: no single biomarker or modeling approach captures all relevant aspects of biological aging, and the space of possible feature combinations is vast.
Here, we report on results from the Biomarkers of Aging Challenge, a two-phase open competition organized by the Biomarkers of Aging Consortium (ref.
Chronological age alone cannot capture the heterogeneity in how individuals age1. Robust biomarkers are needed both to evaluate interventions that target the aging process and to stratify individuals by disease risk. Over the past decade, epigenetic clocks have provided quantitative tools for measuring biological age2,3, and mortality-predictive models such as GrimAge4 and PhenoAge5 have extended their clinical relevance. However, the field faces persistent challenges: limited reproducibility across cohorts, inconsistent dataset structures, and insufficient validation against hard clinical end points6. Most aging clocks are developed and tested on different datasets with different preprocessing pipelines, which makes direct comparisons difficult. It remains unclear whether current prediction accuracy reflects a true ceiling or simply a lack of methodological diversity.
Open scientific competitions have overcome analogous barriers in other domains. The CASP (Critical Assessment of Structure Prediction) competitions transformed protein structure prediction and ultimately produced AlphaFold7. Aging is heterogeneous: no single biomarker or modeling approach captures all relevant aspects of biological aging, and the space of possible feature combinations is vast. This makes the field well suited to such a paradigm. Here, we report on results from the Biomarkers of Aging Challenge, a two-phase open competition organized by the Biomarkers of Aging Consortium (ref. 8) to test whether crowdsourcing diverse computational expertise against a shared benchmark could advance the field.
