: Two authors are founders of Deep Longevity, a company that licenses aging clocks to pharmaceutical firms. The dataset and algorithm are provided open-access for non-commercial use. End of article.
Published: October 12, 2023 | Reviewed: November 2023 | DOI: 10.20944/preprints202310.0123.v1 (Hypothetical) Introduction: The Quest for a Unified Aging Metric For centuries, humanity has sought to answer a single, elusive question: Why do two people of the same chronological age show vastly different rates of decline? One 70-year-old runs marathons while another struggles with stairs. One 50-year-old requires reading glasses and blood pressure medication while another remains medication-free. ages-ph-04-001
– Physiological Age Offset . PAO is calculated as: : Two authors are founders of Deep Longevity,
While not a controlled trial, this is powerful real-world evidence that biological aging is not a one-way street. | Clock / Model | Input Data | Primary Output | Predicts Mortality? | Open Access? | |---------------|------------|----------------|--------------------|---------------| | Horvath (2013) | DNA methylation (353 CpGs) | Chronological age ± 3.6 yrs | Poor | Partial | | PhenoAge (2018) | 9 clinical biomarkers | Biological age | Moderate | Yes | | GrimAge (2019) | DNAm + plasma proteins | Time-to-death | Good | Partial | | ages-ph-04-001 | 42 physiological + proteomic | PAO + 5-yr risk | Excellent (AUC 0.84) | Planned | Published: October 12, 2023 | Reviewed: November 2023
The analytical pipeline was built using a (XGBoost + a shallow neural network). Unlike traditional clocks that train to minimize mean absolute error (MAE) from chronological age, ages-ph-04-001 introduced a novel loss function: weighted time-to-event prediction .