Researchers develop ChromAgeNet, a deep-learning tool that identifies age-related changes in stem cell DNA organization, potentially opening new avenues for aging research and the discovery of rejuvenation therapies.

Scientists are exploring a new way to measure biological aging by analyzing microscopic images of hematopoietic stem cells, the specialized cells responsible for producing blood and maintaining the body’s blood-forming system.

The research introduces ChromAgeNet, an artificial intelligence-based model designed to detect age-related changes in the organization of chromatin, the complex of DNA and proteins that packages genetic material inside a cell’s nucleus. The findings demonstrate how advanced image analysis could help researchers identify cellular signatures of aging and evaluate potential treatments aimed at restoring youthful cell function.

Aging affects the hematopoietic system, contributing to declining function and reduced healthspan. Although changes in chromatin architecture are recognized as an important hallmark of aging, their potential as measurable indicators of biological age remains relatively underexplored.

Chromatin plays a crucial role in regulating gene activity. Depending on how DNA is packaged, whether tightly compacted or more loosely arranged, different genetic sequences may become accessible to the cellular machinery responsible for gene expression. These structural patterns can change in response to cellular conditions, and some changes are associated with aging.

To investigate whether these patterns could serve as indicators of age, the researchers developed ChromAgeNet, a convolutional neural network trained on three-dimensional microscope images of hematopoietic stem cell nuclei stained with DAPI, a fluorescent dye that binds to DNA.

The model was trained to distinguish between young and aged hematopoietic stem cells from mice. It achieved an area under the receiver operating characteristic curve (AUROC) of 0.77 ± 0.03, indicating a meaningful ability to differentiate between the two age groups. It also outperformed traditional machine-learning models that relied on manually selected chromatin features from the same dataset.

Identifying the Cellular Signatures of Aging

Beyond predicting age-associated differences, the researchers used explainable artificial intelligence techniques to investigate which nuclear features influenced the model’s predictions.

The analysis highlighted three important markers: chromatin entropy, peripheral heterochromatin, and chromatin condensates. Together, these features offer clues about how the spatial organization of genetic material changes as stem cells age.

The ability to identify these features is particularly valuable because it makes the model more interpretable. Rather than simply classifying cells as young or aged, ChromAgeNet can help researchers investigate the underlying structural patterns associated with the aging process.

A Potential Tool for Testing Rejuvenation Treatments

The researchers also explored whether ChromAgeNet could be used to screen aged hematopoietic stem cells treated with epigenetic drugs. These compounds influence the mechanisms that regulate gene activity without changing the underlying DNA sequence.

This proof-of-concept experiment examined the model’s potential to detect changes associated with cellular rejuvenation. Such an approach could eventually help researchers assess whether candidate treatments alter age-related chromatin patterns and prioritize promising compounds for further investigation.

However, identifying a more youthful chromatin pattern does not necessarily mean that a cell has regained its youthful function. Further studies will be needed to establish whether changes detected by the model correspond to meaningful improvements in stem cell performance, tissue maintenance, or overall health.

Advancing the Search for Biological Age Markers

ChromAgeNet illustrates how artificial intelligence can extract biologically relevant information from high-resolution cellular images, transforming complex patterns of DNA organization into measurable features associated with aging.

The framework could support future research into stem cell aging, provide a tool for investigating the relationship between chromatin structure and cellular function, and help accelerate the screening of potential rejuvenation therapies.

For now, the findings represent an early step toward image-based aging clocks. The model was developed using mouse stem cells, and further validation will be required before its accuracy, usefulness, and applicability to human aging can be established.

Nevertheless, the study highlights a promising research direction: the microscopic architecture of a cell’s DNA may contain measurable clues about its biological age, and artificial intelligence may help scientists uncover them.