Scientists Build First Multiomic Biological Age Clocks for Mice
Researchers created Mouse PhenoAge, the first composite biological age model for mice, using DNA methylation and metabolomics to predict lifespan.
Summary
Researchers at Harvard Medical School and the VoLo Foundation developed Mouse PhenoAge, the first second-generation biological age clock for mice. Built from frailty assessments, complete blood counts, and mortality data in C57BL/6 mice tracked across their full lifespan, this composite model was then used to train three omic-based predictive clocks: one using DNA methylation (DNAm PhenoAge), one using plasma metabolomics (Mtb PhenoAge), and one combining both (MultiOmic PhenoAge). All three clocks accurately predicted Mouse PhenoAge, and their residuals correlated with remaining lifespan even in mice of the same chronological age. This work could dramatically reduce the need for lengthy, expensive survival studies in preclinical longevity research.
Detailed Summary
Measuring biological age — how old an organism truly is in terms of health and function — is more informative than chronological age alone, yet sophisticated 'second-generation' biological age clocks have previously existed only for humans. This study fills that gap for laboratory mice, a critical step for accelerating preclinical longevity research.
The team followed two cohorts of C57BL/6NIA mice longitudinally across their full lifespans (median 914 days), collecting frailty index scores, complete blood counts (CBC), plasma metabolomics, and genome-wide DNA methylation data from blood at up to five distinct time points. The mice exceeded typical laboratory lifespan benchmarks, confirming healthy husbandry conditions. Only untreated control animals were used for omic model training, excluding mice from an ongoing NMN intervention study.
Mouse PhenoAge was constructed in a three-step process: (1) a univariate linear regression model predicted days-to-death from chronological age; (2) an elastic net model predicted days-to-death from frailty index items, CBC measures, and age; (3) the predicted survival was mathematically back-transformed into an age-equivalent value. This composite outcome captures both phenotypic health and mortality risk, analogous to the human PhenoAge framework developed by Levine and colleagues.
Three omic clocks were then trained using leave-one-out cross-validation to predict Mouse PhenoAge: DNAm PhenoAge (genome-wide DNA methylation via the Infinium Mouse Methylation BeadChip), Mtb PhenoAge (global plasma metabolomics via UPLC-MS/MS through Metabolon), and MultiOmic PhenoAge (combining both data types). All models incorporated sex as a binary covariate. Critically, the residuals of these clocks — the gap between predicted and actual Mouse PhenoAge — were significantly associated with remaining lifespan even among mice of the same chronological age, validated with Kaplan–Meier survival analyses and log-rank tests stratified by clock-residual quantiles.
These clocks represent a practical advance for the field: instead of waiting years for survival endpoints in preclinical studies, researchers could use multiomic blood measurements at a single time point to estimate biological age and stratify animals by mortality risk. This could meaningfully shorten and reduce the cost of interventional aging studies in mice, enabling faster translation of longevity interventions toward clinical development.
Key Findings
- Mouse PhenoAge is the first composite biological age outcome for mice, integrating frailty, blood counts, and mortality risk.
- DNA methylation, metabolomic, and multiomic clocks all accurately predicted Mouse PhenoAge using blood samples.
- Clock residuals predicted remaining lifespan even in mice matched for chronological age.
- Kaplan–Meier analyses confirmed multiomic clock predictions significantly stratify mice by survival outcome.
- The approach could replace lengthy survival studies with single time-point blood-based biological age assessments.
Methodology
Two cohorts of C57BL/6NIA mice were tracked longitudinally across their full lifespans with frailty assessments, CBC, genome-wide DNA methylation (Infinium Mouse Methylation BeadChip), and global plasma metabolomics (Metabolon UPLC-MS/MS) collected at up to five time points. Elastic net regression models were trained with leave-one-out cross-validation to prevent data leakage, incorporating sex as a covariate.
Study Limitations
Models were trained exclusively on male and female C57BL/6NIA mice, limiting generalizability to other strains or species. One cohort included mice from an NMN treatment study, and only untreated animals were used, potentially reducing sample size. Frailty scoring involved subjective assessments that could introduce inter-rater variability.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: