Deep Learning Predicts Hematopoietic Stem Cell Aging From 3D Chromatin Images
Key claim: Deep learning can predict hematopoietic stem cell aging status from three-dimensional chromatin images.
Abstract
Aging Cell, Volume 25, Issue 10, October 2026.
Why this matters
Chromatin reorganization is a core feature of hallmark ‘epigenetic alterations,’ but it has mostly been read out indirectly through methylation clocks or bulk ATAC. A deep-learning classifier that recovers HSC aging status from 3D nuclear images turns nuclear architecture into a directly measurable, single-cell aging readout — relevant because HSC aging sits upstream of immunosenescence, clonal hematopoiesis, and myeloid-biased inflammaging. The work remains murine and classification-level, so its utility as a rejuvenation screen is proof-of-concept, not a validated human biomarker.