A convolutional neural network trained on 3D microscopy images of DAPI-stained hematopoietic stem cell nuclei distinguished young from aged murine cells with an AUROC of 0.77, outperforming classical models built on manually defined chromatin features. Explainability analysis identified chromatin entropy, peripheral heterochromatin, and chromatin condensates as the structural features carrying the age signal. Because blood and immune cell production declines measurably with age, an image-based readout of stem cell chromatin state offers a quantitative endpoint for testing rejuvenation interventions.
Key Points
- Neural network separated young from aged stem cell nuclei, AUROC 0.77
- Chromatin entropy, peripheral heterochromatin, condensates drove age predictions
- Model detected chromatin rejuvenation in epigenetic drug-treated aged cells
Longevity Analysis
Hematopoietic stem cells supply the entire defense and oxygen-carrying compartment, and their functional drift with age propagates into immune competence, tissue repair capacity, and blood-borne inflammatory tone. Reading nuclear architecture directly converts a diffuse hallmark of aging into a measurable signal, which matters more than another correlative biomarker because it responds to intervention within the same assay. The practical value lies in throughput: a phenotypic endpoint that flags chromatin reorganization after treatment allows candidate rejuvenation compounds to be ranked before they reach functional or clinical testing.
Original published by Wiley Aging Cell, by Pablo Iáñez Picazo, Eva Mejía‐Ramírez, Dario Di Bari, Elena Vitali, Maria Carolina Florian, Paula Petrone .