Trustworthy ML/AI for Aging Clocks: Preventing Systematic Prediction Bias in Biological Age Estimation
- 1University of Maryland, School of Medicine
Abstract
Machine learning (ML)- and artificial intelligence (AI)-based aging clocks are increasingly used to quantify physiological and molecular aging from omics and medical imaging data as distinct from chronological age. Here, we characterize a fundamental but underappreciated statistical limitation of commonly used ML/AI regression models for continuous outcomes: systematic prediction bias and its propagation to downstream association estimates. This issue becomes more challenging when the true outcome, biological age, is latent and therefore unobserved during ML/AI model training. We demonstrate that systematic prediction bias can distort and, in some cases, even reverse downstream association analyses that use aging clocks as ML/AI-predicted outcomes to assess their associations with exposures or clinical factors. For example, it can produce spurious associations suggesting that older predicted brain age is linked to better cognitive performance, or that older epigenetic age is associated with better kidney function. To address this problem, we introduce a principled and broadly applicable ML/AI regression framework based on constrained optimization, yielding better calibrated aging-clock estimates and valid downstream inference.
Related
- Towards Trustworthy Biological Alignment in TabPFN-Probed Pathology Foundation Models
- From objective discovery to prediction of global ocean eco-provinces: A pathway for trustworthy learning
- Biomaterials and Biological Therapies in Tympanic Membrane Regeneration: Systematic Review and Perspective
- Inferential boundaries of age prediction: why prediction does not establish biological age measurement
- MutCleaner: Cleaning and Standardizing Biological Mutation Datasets for Variant Effect Prediction
BibTeX
@misc{lee2026trustworthy,
title = {{Trustworthy ML/AI for Aging Clocks: Preventing Systematic Prediction Bias in Biological Age Estimation}},
author = {H. Lee and Z. Ye and Y. Yang and Y. Pan and B. Maron and Z. Wang and P. Kochunov and P. Thompson and L. E. Hong and T. MA and C. Chen and S. Chen},
howpublished = {bioRxiv},
year = {2026},
month = may,
doi = {10.64898/2026.05.27.728155},
url = {https://www.biorxiv.org/content/10.64898/2026.05.27.728155v2},
note = {CC BY},
}
RIS
TY - UNPB
TI - Trustworthy ML/AI for Aging Clocks: Preventing Systematic Prediction Bias in Biological Age Estimation
AU - H. Lee
AU - Z. Ye
AU - Y. Yang
AU - Y. Pan
AU - B. Maron
AU - Z. Wang
AU - P. Kochunov
AU - P. Thompson
AU - L. E. Hong
AU - T. MA
AU - C. Chen
AU - S. Chen
PY - 2026
DA - 2026/05/27
PB - bioRxiv
DO - 10.64898/2026.05.27.728155
UR - https://www.biorxiv.org/content/10.64898/2026.05.27.728155v2
AB - Machine learning (ML)- and artificial intelligence (AI)-based aging clocks are increasingly used to quantify physiological and molecular aging from omics and medical imaging data as distinct from chronological age. Here, we characterize a fundamental but underappreciated statistical limitation of commonly used ML/AI regression models for continuous outcomes: systematic prediction bias and its propagation to downstream association estimates. This issue becomes more challenging when the true outcome, biological age, is latent and therefore unobserved during ML/AI model training. We demonstrate that systematic prediction bias can distort and, in some cases, even reverse downstream association analyses that use aging clocks as ML/AI-predicted outcomes to assess their associations with exposures or clinical factors. For example, it can produce spurious associations suggesting that older predicted brain age is linked to better cognitive performance, or that older epigenetic age is associated with better kidney function. To address this problem, we introduce a principled and broadly applicable ML/AI regression framework based on constrained optimization, yielding better calibrated aging-clock estimates and valid downstream inference.
ER -