Project Grant R21AG098680

The National Institute on Aging awarded the University of Maryland, Baltimore $427,625 under Project Grant R21AG098680 on August 15, 2026, to develop machine learning regression tools that remove systematic prediction bias in aging clock calculations.

Aging clocks estimate biological age from high-throughput multi-omics data and multimodal medical imaging, providing more accurate measures of age-related physiological functions than chronological age alone. Existing machine learning regression models used to derive aging clocks—including epigenetic clocks based on DNA methylation and brain age estimates from neuroimaging—produce systematically biased predictions that distort biological age estimates and compromise the accuracy of downstream clinical inferences, such as assessments of anti-aging treatment efficacy. The recipient will develop novel constrained optimization techniques for multiple regression model types including random forest, gradient boosting, and shrinkage regression, customized to account for each model's specific optimization strategy. The work includes comprehensive evaluation using simulated datasets under various conditions and multiple large-scale real-world datasets, culminating in delivery of an open-source software package with user tutorials to enable unbiased aging clock calculations across the research and clinical communities.

Performance occurs in Baltimore, Maryland, with a period of performance from August 15, 2026, through July 31, 2028. This award falls under the National Institute on Aging's Aging Research program (CFDA 93.866), which supports biomedical and behavioral research directed toward understanding aging and age-related diseases. The assistance type is a Project Grant.