Development of ChromAgeNet AI for stem cell aging

Researchers in Barcelona, including teams from the Bellvitge Biomedical Research Institute (IDIBELL) and the Barcelona Supercomputing Center, have developed ChromAgeNet, a deep learning framework designed to detect aging in hematopoietic stem cells (HSCs).

Published in the journal ‘Aging Cell’, the technology utilizes convolutional neural networks to analyze three-dimensional microscopy images of the nucleus. Rather than relying on molecular measurements like DNA methylation, the tool identifies changes in the physical organization of the DNA-protein complex, such as chromatin entropy and peripheral heterochromatin.

Initial reports indicated the AI achieved a 77% accuracy rate in distinguishing between young and aged cells. Subsequent details specify that the tool utilizes Explainable AI (XAI) to provide transparency in its decision-making processes. Using standard, low-cost DAPI staining, the model achieved an average correct classification rate of approximately 68% when observing how chromatin organization shifts from a compact, organized state in younger cells to a more scattered state in older cells. The researchers have demonstrated the tool's utility by using it to evaluate how certain epigenetic drugs can induce chromatin changes that mimic patterns seen in younger cells.