Building upon these insights, we constructed 36 histone modification-based epigenetic clocks, which exhibited robust predictive accuracy (mean Pearson’s r = 0.91) across multiple tissues and marks. Among these, the blood-derived H3K27ac clock emerged as a particularly powerful model, outperforming several established DNA methylation clocks under matched conditions. This performance is remarkable considering that DNA methylation clocks have undergone extensive optimization over the past decade (9, 16, 18), while our histone-based approach represents a first-generation effort.
A distinctive advantage of our histone-based clocks is their resilience to technical and biological noise. When exposed to artificial Gaussian noise, the histone-based clock maintained stable predictive performance, in contrast to the sharp degradation observed in many methylation-based models. This robustness is likely attributable to the broader, structural nature of histone mark signals, which may be less sensitive to local fluctuations than single CpG methylation values. This characteristic makes histone clocks potentially more suitable for noisy, heterogeneous, or clinically derived datasets where sample quality may vary.
The practical utility of our histone-based clocks was further demonstrated by their ability to detect biological age acceleration in leukemia samples and capture age reversal following therapeutic interventions. These applications highlight the potential of histone-based clocks as biomarkers for disease states and treatment responses, offering a complementary approach to existing clinical tools.
