Aging reshapes nearly every system in the body, and one of its quieter effects shows up in the blood. As we get older, the hematopoietic system — the organs and tissues responsible for producing blood cells — becomes less able to sustain adequate blood cell production. For researchers studying how blood stem cells age, and how their function might be preserved or restored, that decline is not just a curiosity. It is a measurement problem.
A team led by Dr. Maria Carolina Florian, a researcher in the Regenerative Medicine program at the Bellvitge Biomedical Research Institute (IDIBELL) and an ICREA research professor, and Dr. Paula Petrone of the Barcelona Supercomputing Center — Centro Nacional de Supercomputación (BSC-CNS) and the Barcelona Institute for Global Health (ISGlobal), has introduced a tool designed to help solve it. Named ChromAgeNet, the system uses artificial intelligence to detect aging-associated patterns in microscopy images of hematopoietic stem cells by analyzing the three-dimensional organization of chromatin. The findings are published in the journal Aging Cell, and the work formed a central part of the doctoral thesis of Pablo Iañez, a researcher at ISGlobal.
What Chromatin Says About a Cell
Chromatin is the material inside the cell nucleus — made up principally of DNA and proteins — that packages DNA and helps regulate which genes are active. That packaging is not merely structural. Because it influences gene activity, chromatin organization helps determine a cell's identity and function, and it shifts as cells grow older.
ChromAgeNet is built to exploit that relationship directly. Instead of relying on a handful of molecular markers, the tool reads the three-dimensional arrangement of chromatin inside nuclear images and answers a deceptively simple question: does this nucleus look young or aged?
How the Model Was Trained
To build the system, the researchers worked with three-dimensional images of mouse hematopoietic stem cell nuclei that had been stained with DAPI, a simple and widely used technique for visualizing DNA. A convolutional neural network — a class of AI model designed specifically to analyze images — was then trained to separate young cells from aged ones on the basis of appearance alone.
- Imaging input: 3D microscopy images of mouse hematopoietic stem cell nuclei stained with DAPI.
- Model type: a convolutional neural network suited to image analysis.
- Task: classify each cell as young or aged from nuclear appearance.
- Focus: the three-dimensional organization of chromatin rather than predefined markers.
This approach is notable because it asks the model to learn whatever distinguishing features exist in the images, rather than requiring researchers to specify them in advance.
Performance: 77% Accuracy and a Benchmark Beaten
According to the researchers, ChromAgeNet achieved a 77% probability of correctly classifying cells as young or aged based on the appearance of the nucleus. That result outperformed a machine learning model built on chromatin features the team had previously defined by hand.
The comparison matters. It suggests that a trained neural network can pick up on patterns in nuclear architecture that conventional, feature-based approaches miss — information that is present in the images but not captured by the specific measurements researchers had chosen to look at.
Detecting What the Eye Cannot See
The study's framing centers on a familiar problem in microscopy: human observers can describe a nucleus, but they cannot reliably rank subtle, three-dimensional differences in chromatin packing as a consistent indicator of age. ChromAgeNet is an attempt to turn those subtle differences into a signal that can be computed repeatedly and at scale.
If such a signal holds up, it could support the study of blood stem cell aging in several ways. It offers a potential readout for experiments that test whether a treatment preserves or restores stem cell function. It could also help researchers sort cells that appear similar under a microscope but behave differently in practice.
A Cross-Disciplinary Effort
The project brought together expertise in stem cell biology, aging, image analysis and artificial intelligence — a combination reflected in the institutions involved. The work was led by researchers affiliated with IDIBELL's Regenerative Medicine program, BSC-CNS, and ISGlobal, a center supported by the "la Caixa" Foundation. The study was peer-reviewed, and the imaging used in the published figures was credited to Dr. Eva Mejía of IDIBELL.
What the Results Do and Do Not Establish
The findings rest on mouse hematopoietic stem cells and DAPI-stained nuclear images, and the reported accuracy of 77% means the model is informative rather than infallible — roughly one in four classifications would be expected to go the other way. The published summary also does not describe clinical validation or use in human patients.
Those boundaries are worth keeping in view. ChromAgeNet is best understood as a research instrument: a way to extract an aging-related signal from images that are already standard in stem cell laboratories, using a staining technique that is inexpensive and routine.
Why It Matters
The broader stakes lie in the hematopoietic system's gradual decline with age. Because that system supplies blood cells, its deterioration touches immunity, oxygen transport and clotting, among other functions. Tools that make the aging of blood stem cells easier to observe could help researchers ask sharper questions about what changes, when, and whether any intervention alters the trajectory.
ChromAgeNet's contribution is methodological as much as biological. It shows that chromatin architecture, captured in ordinary microscopy, carries enough age-related information for a neural network to detect — and that this approach can outperform a model built on features scientists had defined themselves. Whether the same signal can be extended to other cell types, other imaging setups or human samples remains an open question the study does not settle.
The Takeaway
- An AI tool called ChromAgeNet identifies aging-associated patterns in hematopoietic stem cell images by analyzing 3D chromatin organization.
- The model classified mouse blood stem cells as young or aged with a 77% probability of being correct.
- It outperformed a machine learning model based on chromatin features researchers had defined in advance.
- The work was led by researchers at IDIBELL, BSC-CNS and ISGlobal and published in Aging Cell.
- The approach relies on DAPI staining and 3D nuclear imaging, techniques already common in stem cell research.
This article is based on reporting by Medical Xpress. Read the original article.
Originally published on medicalxpress.com
