Barcelona AI tool detects aging in blood stem cells, opening path to rejuvenation

A research team in Barcelona has built an artificial intelligence system that can read the age written into the three-dimensional architecture of blood stem cells—and potentially identify drugs that could reverse it. The tool, called ChromAgeNet, works by analyzing images of cell nuclei stained with a standard, inexpensive dye, then using neural networks to detect the subtle structural signatures that distinguish young cells from old ones. The work, published in the journal Aging Cell and led by Maria Carolina Florian at IDIBELL and Paula Petrone at BSC-CNS and ISGlobal, represents a bridge between computational power and the biological reality of aging: as the body grows older, its capacity to generate new blood cells deteriorates, weakening immunity and regenerative function. Understanding why requires looking inside the cell nucleus itself.

The researchers trained ChromAgeNet on three-dimensional images of mouse stem cell nuclei, teaching the algorithm to recognize patterns invisible to the human eye. On average, the system correctly classified cells as young or aged about 68 percent of the time using only the standard DAPI stain—a technique already available in most laboratories. What sets ChromAgeNet apart is that it does not simply make a prediction and hide its reasoning. The system is built as explainable AI, meaning researchers can investigate which features in the images actually drove each classification. When they did this, a clear pattern emerged: in young cells, chromatin—the packaged form of DNA—sits compact and organized at the edges of the nucleus. In aged cells, that same architecture becomes dispersed and loose. The difference is not dramatic enough for a human observer to reliably spot, but the algorithm catches it consistently.

To test whether this insight could lead somewhere practical, the team conducted a proof of concept. They took aged cells and exposed them to various epigenetic drugs—compounds designed to alter how DNA is packaged without changing the genetic code itself. Then they ran those treated cells through ChromAgeNet to see whether the drugs had shifted the chromatin organization toward a younger pattern. Some did. This suggests the tool could become a screening platform: researchers could test hundreds of candidate compounds to identify which ones actually induce changes in cell architecture consistent with rejuvenation, rather than relying on slower, more expensive traditional methods.

Florian explained to Euronews that the current version works as a proof of concept in mouse blood stem cells, but the team is now working to adapt it for human cells. "We are now working to optimise it and adapt it to human stem cells, with the hope of using it to evaluate new drugs that can act on nuclear architecture and be used to rejuvenate aged stem cells." The researchers are applying for additional European Union funding and actively seeking collaborations with other scientists across Europe who see practical applications for the work. Paula Petrone, the computational partner on the project, emphasized the efficiency of the approach: a deep-learning algorithm can distinguish young from aged stem cells by analyzing images stained with a low-cost substance, a significant advantage for labs with limited budgets.

The team has already released the tool's code and a public database of three-dimensional stem cell images to the international scientific community, accelerating the pace at which other researchers can build on the work. The underlying logic is straightforward: if blood stem cells decline with age, and if that decline is written into the physical organization of the nucleus, then identifying the molecular mechanisms behind that decline could open doors to therapies. Florian's laboratory has previously demonstrated that aged blood stem cells can be stimulated and rejuvenated in ways that improve not only immune function but also life expectancy and overall organism health. ChromAgeNet is positioned as a tool to understand those mechanisms more deeply and to identify new drugs that could trigger similar improvements.

Petrone framed the long-term vision clearly: "In future, this technique could be used to screen new drugs with rejuvenation potential. By treating aged cells with different candidate compounds, AI could use image analysis to identify those that induce changes consistent with a more youthful state." She also articulated what matters most: extending not just lifespan but healthspan—the years lived in good health. New drugs with rejuvenating potential could help restore immune function as people age, supporting what researchers call healthy aging. Yet Petrone was careful to temper expectations. "We must be very careful not to overestimate our results," she told Euronews. "These new technologies take years to be developed, validated and eventually reach the clinic. Our research is just one small contribution to the study of how we age and to the search for strategies that can support healthy ageing." The work is real and the potential is genuine, but the path from laboratory discovery to clinical application remains long and uncertain.