This study reports development and validation of a urinary peptidomic biological ageing clock measured by capillary electrophoresis–mass spectrometry (CE‑MS). Because urine composition reflects renal handling, the authors derived an unconditioned urine peptide clock (UPBioAge) and then applied a kidney‑specific correction, Filtrate‑Aware Calibration (FAC), to generate a kidney‑conditioned clock (k‑UPBioAge). The work aimed to produce a non‑invasive ageing biomarker expected to respond to interventions and to test whether accelerated biological age measured in urine predicts all‑cause mortality and age‑related health outcomes.
Urinary peptides were quantified by CE‑MS across multiple retrospective cohorts. The unconditioned model (UPBioAge) was trained in a derivation cohort of participants with preserved kidney function (n = 1,811). FAC was fitted in an independent, kidney‑diverse cohort (n = 7,798) to condition the model on estimated glomerular filtration rate (eGFR) and urinary albumin‑to‑creatinine ratio (UACR), producing the kidney‑conditioned model k‑UPBioAge. Age prediction performance was assessed in three cohorts that were independent of model development, including one population cohort and one set of samples analysed at an external site.
The final model retained 2,631 peptides; the authors provide anonymised summary statistics, model coefficients, weights and parent‑protein annotations as supplementary material to support independent evaluation. Individual‑level peptidomic data remain proprietary and are available subject to data‑use agreements with the originating company.
Because urine is generated and modified by the kidney, urinary peptide signatures are influenced by renal function. To address this, the team used Filtrate‑Aware Calibration (FAC) to condition the urine peptide clock on two markers of kidney status: eGFR and UACR. FAC was trained in a kidney‑diverse cohort so that the resulting k‑UPBioAge would provide age estimates that are not confounded by differences in kidney filtration or albuminuria across individuals.
After an age‑bias correction step, the kidney‑conditioned clock estimated chronological age with a calibrated holdout mean absolute error (MAE) of 4.91 years and Pearson correlation r = 0.945 in the development/holdout set. In two independent validation cohorts the calibrated MAE was 5.43–5.47 years. These performance metrics indicate that a multi‑peptide urinary signature can estimate chronological age with accuracy comparable to other biological ageing measures, once kidney function is accounted for.
The study quantified kidney‑conditioned age acceleration (k‑UPBioAgeAcc) as the extent to which predicted biological age exceeded chronological age after calibration and kidney conditioning. Associations between k‑UPBioAgeAcc and outcomes were examined in a clinically enriched follow‑up cohort (n = 7,469) with 625 deaths and a median follow‑up of 3.95 years, using Cox proportional hazards models adjusted for chronological age, sex, comorbidities, body‑mass index, mean arterial pressure and eGFR.
Each standard deviation increment in k‑UPBioAgeAcc was associated with increased risk of:
These associations were estimated after adjustment for multiple covariates and suggest that accelerated urinary peptidomic age relates to both mortality and several age‑related cardiovascular and renal outcomes.
The authors tested for effect modification by sex and baseline comorbidities. The association between k‑UPBioAgeAcc and outcomes did not differ by sex (P for interaction = 0.33). None of four tested comorbidity interactions remained significant after correction for multiple testing (adjusted P = 0.65–0.72). However, the clock did not show an association with outcomes in participants with severe kidney impairment (eGFR 15–29 mL/min/1.73 m2; n = 433, 84 deaths) or in those with macroalbuminuria (n = 92, 34 deaths), indicating that the predictive value may be limited in advanced kidney disease or heavy albuminuria.
The study demonstrates that multiple urinary peptides are significantly associated with chronological ageing and that a conditioned urinary peptidomic clock can predict all‑cause mortality and incident age‑related diseases. Because urine composition is affected by kidney filtration and albuminuria, the authors emphasise the need to correct urinary ageing signals for kidney function; they implemented FAC to achieve this.
The authors propose that the corrected urinary peptide‑based ageing clock may have utility for monitoring biological ageing or guiding personalised interventions, given its non‑invasive sampling and potential responsiveness to interventions. Key limitations reported include the retrospective multi‑cohort design and proprietary status of individual‑level peptidomic data; the manuscript notes that individual data are available only under data‑use agreements. Also, the conditioned clock did not predict outcomes among participants with very low eGFR or macroalbuminuria, so applicability in advanced kidney disease is uncertain.
Anonymised summary statistics and the coefficients, model weights and parent‑protein annotations for the 2,631 peptides retained in the model are provided as supplementary material. Individual‑level urinary peptidomic data are proprietary to Mosaiques Diagnostics GmbH and available subject to data‑use agreements; requests can be directed to the corresponding author.
Competing interests disclosed include that H.M. is founder and co‑owner of Mosaiques Diagnostics GmbH, and several co‑authors are employees of that company. Other co‑authors declared no competing interests. Ethical review for the present anonymised analysis was waived by the ethics committee of Hannover Medical School because all datasets were previously published and fully anonymised.