Data Science and Statistics Seminar by Dr. Judy Wang (Rice University)

Title: Rank-based Fusion Learning of Biological Age from Multiple Epigenetic Clocks

Abstract: Biological age is intended to capture an individual’s aging status more directly than chronological age, but in practice different epigenetic clocks often give substantially different age estimates for the same subject. We propose a rank-based framework for combining multiple epigenetic clocks to assess relative biological aging and estimate biological age. For each subject, clock-specific conditional ranks are obtained by comparing epigenetic age estimates with those of age-matched peers. These ranks are fused into a consensus aging rank and mapped back to an age scale through a calibration distribution. The approach avoids restrictive structural assumptions on raw clock outputs and is estimated using nonparametric quantile regression. Our ongoing work examines calibration, identifiability, and weighting strategies for stable fusion across clocks. We present preliminary results from simulations and an analysis of motivating DNA methylation age data from the ELEMENT study.