Beyond point estimates: quantifying predictive uncertainty reveals hidden dimensions of biological age acceleration and improves risk interpretation

Chen Wang, Hanqing Wu, Shinichi Namba, Jun Young Park, Koichi Matsuda, Yukinori Okada, Zihuai He, Iuliana Ionita-Laza

1. 01

   Framework generates calibrated prediction intervals and individual probabilities of accelerated/decelerated aging

2. 02

   Applied to 3 composite and 11 organ-specific proteomic clocks in UK Biobank Pharma Proteomics Project

3. 03

   Many low-accuracy organ-specific clocks provided little evidence for a confident individual-level call despite large apparent age gaps

4. 04

   Prediction-interval width was independently associated with disease risk and mortality, especially for composite, brain and immune clocks

5. 05

   Replicated in Biobank Japan and an independent Stanford clinical cohort

Study Design

Preprint

Sample Size

Not reported

Study Duration

Not reported

Methodology

Conformal-style calibrated prediction interval framework applied to proteomic biological age clocks in UK Biobank Pharma Proteomics Project, with replication in Biobank Japan and a Stanford clinical cohort.

Limitations

PREPRINT - posted to bioRxiv 5 August 2026 and not peer-reviewed. Applies to proteomic clocks specifically; DNA methylation clocks were not evaluated. Should not be cited as established evidence, though the direction of the argument is consistent with peer-reviewed work on clock heterogeneity.