International Conference on Bioinformatics and Biomedical Technology · May 22, 2026

Estimating Immune Age from B-Cell Receptor Repertoire Using Machine Learning

By Liu T, Yang F

Quantifying immunosenescence is pivotal for geriatric health management yet remains challenged by the low resolution of traditional macroscopic biomarkers. This study proposes a novel computational framework that integrates multi-dimensional bulk B-cell receptor (BCR) repertoire profiling with automated machine learning (AutoML) to decipher immune aging trajectories. Utilizing 326 samples from healthy and chronic disease cohorts, we extracted 53 distinct features, including somatic hypermutation (SHM) rates and isotype usage, to train a predictive model via rigorous 5-fold nested cross-validation. The resulting ensemble Random Forest model demonstrated robust performance in healthy individuals, achieving a Mean Absolute Error (MAE) of 12.0 years. Notably, validation in an independent HIV cohort revealed a counter-intuitive "negative aging" phenomenon, where the predicted immune age was significantly lower than chronological age (delta Age less than 0). Mechanistic analysis elucidated that this "pseudo-juvenile" profile is driven by specific pathological remodeling: HIV-induced germinal center disruption leads to artificially suppressed SHM rates, while memory B-cell depletion triggers compensatory emergency lymphopoiesis, elevating naive IgM proportions. These findings demonstrate that our framework not only accurately models natural aging but also sensitively characterizes pathological deviations, offering a high-resolution, interpretable tool for immune health monitoring.

doi.org/10.1145/3822521.3822540Open accessBiomarkers

AgingResearch.News