The paper introduces scAgeClock, a sophisticated human aging clock model designed to predict biological age at the single-cell level. Utilizing a gated multi-head attention neural network, this tool was trained on a massive dataset of over 16 million transcriptomes across 44 different human tissues. Researchers can use the model to identify transcriptomic age acceleration in various diseases, including Alzheimer’s and COVID-19, and to track cell dynamics in cancer. The study also proposes a new metric called the Aging Deviation Index (ADI), which quantifies how much a cell's biological age differs from its chronological age. By focusing on cell-type heterogeneity, scAgeClock provides more precise insights into the aging process than traditional bulk-level methods. This open-source software serves as a valuable resource for evaluating anti-aging interventions and assessing long-term health risks.

References:

• Xie G. scAgeClock: a single-cell transcriptome-based human aging clock model using gated multi-head attention neural networks[J]. npj Aging, 2026.