hB-PAC: A non-invasive aging clock for quantifying individual differences in aging

Aging progresses heterogeneously among individuals, and aging clocks that estimate biological age have been developed to quantify this heterogeneity. However, existing methods depend on invasive tissue sampling or long-term longitudinal data, and a mathematical framework that explicitly quantifies individual differences in aging has not yet been established. Here, we employed a progeroid zebrafish model (klotho mutant; kl-/-) and developed a hierarchical Bayesian framework, termed the hierarchical Bayesian-Multimodal Aging Clock (hB-MAC), which integrates non-invasive snapshot data of behavior and morphology with intestinal transcriptomic profiles. This model enables the simultaneous estimation of individual biological age while explicitly quantifying deviation from chronological age. Comparative analyses with biological datasets demonstrated that hB-MAC provides biologically meaningful age estimates while capturing inter-individual variability. Based on this framework, we further developed a new aging clock, the hierarchical Bayesian-Phenotypic Aging Clock (hB-PAC), which predicts individual biological age using only non-invasive data. The biological age estimated by hB-PAC was strongly associated with multiple tissue-level aging processes, including metabolic decline, intestinal barrier dysfunction, and abnormalities in mucosal immunity, which were not fully explained by chronological age. These results demonstrate that biologically relevant aging states incorporating individual variability can be inferred solely from short-term non-invasive observations. The proposed framework enables scalable inference of aging dynamics from macroscopic, non-invasive phenotypic data, providing a foundation for high-throughput evaluation of personalized anti-aging interventions.