Hematopoietic stem cells, the rare self-renewing cells that continually replenish blood and immune populations, do not simply “wear out” with age. They undergo a complex biological transformation that reshapes how they read DNA, respond to stress, interact with their surroundings and decide which blood-cell lineages to produce. A new review by Park, Yanai, Ding and colleagues in Experimental & Molecular Medicine brings together decades of research on this process, focusing on how transcriptional changes and multi-omics technologies are revealing the molecular architecture of hematopoietic stem-cell aging—and how artificial intelligence could help turn that growing data mountain into clinically useful predictions.
Hematopoietic stem cells, or HSCs, reside primarily in the bone marrow, where they remain mostly quiescent until new blood cells are needed. Their defining abilities are self-renewal and multipotency: one HSC can generate daughter cells that preserve the stem-cell pool while also producing progenitors capable of becoming red blood cells, platelets, myeloid cells and lymphocytes. This system is remarkably durable, but its performance changes over time. Aging HSCs often become more abundant yet less efficient, showing impaired regenerative capacity, reduced lymphoid output and a tendency to favor myeloid production. That imbalance can weaken immune defense and contribute to anemia, inflammation and blood disorders.
The review emphasizes that aging is reflected in the transcriptional identity of HSCs—the pattern of genes that are active, repressed or poised for activation. Young and old HSCs may share the same genome, but they do not use it in the same way. Age-associated transcriptional programs can alter cell-cycle control, DNA repair, mitochondrial function, protein quality control and inflammatory signaling. Genes involved in maintaining quiescence may be disrupted, while pathways linked to stress responses and innate immune activation become more prominent. These changes are not uniform across every stem cell. Instead, aging creates a mosaic of HSC states, with some cells retaining youthful characteristics and others acquiring molecular features associated with dysfunction.
One of the most important shifts identified by modern research is the changing balance between lymphoid and myeloid differentiation. In younger organisms, HSCs generally support a broad and flexible output of blood and immune cells. With age, many HSCs become “myeloid-biased,” meaning their descendants are more likely to produce cells such as monocytes, macrophages and neutrophils than B or T lymphocytes. The consequences reach beyond a simple change in blood-cell proportions. Reduced lymphoid production can compromise adaptive immunity, while excessive or chronically activated myeloid output may reinforce systemic inflammation. Researchers are investigating whether this bias is caused by selective expansion of particular HSC clones, altered signaling from the bone-marrow environment, intrinsic epigenetic changes, or a combination of all three.
The bone-marrow niche is a central part of the story. HSCs do not age in isolation; they are influenced by neighboring stromal cells, endothelial cells, immune cells, extracellular matrix components and circulating factors. Aging can alter the physical and chemical properties of this niche, changing oxygen availability, inflammatory signaling, nutrient delivery and cell-to-cell communication. Molecules such as inflammatory cytokines can push HSCs out of quiescence or modify their differentiation choices. At the same time, age-related changes inside the HSC—including mitochondrial stress, accumulated DNA damage and declining proteostasis—can make the cell more vulnerable to signals that would have been manageable earlier in life. The result is a feedback loop in which intrinsic and environmental damage reinforce one another.
Multi-omics is allowing scientists to examine this process at several molecular layers simultaneously. Single-cell RNA sequencing measures gene-expression programs in individual HSCs rather than averaging signals across millions of cells. This is crucial because a population that appears uniform in bulk analysis may contain many distinct subgroups. Single-cell chromatin-accessibility methods, such as ATAC-seq, reveal which regulatory regions are open and potentially available for transcription. DNA-methylation profiling can identify epigenetic marks associated with biological age, while proteomics and metabolomics provide information about the proteins and chemical reactions that ultimately execute cellular functions. When these data are combined, researchers can connect regulatory changes at the genome level with altered metabolism, signaling and cell behavior.
The review also highlights the importance of clonal hematopoiesis, a phenomenon in which blood-cell production becomes disproportionately dominated by clones carrying acquired genetic or epigenetic alterations. Clonal hematopoiesis becomes more common with age and can arise when a stem cell gains a competitive advantage, often through mutations in genes involved in epigenetic regulation, DNA damage responses or cellular signaling. Many people with clonal hematopoiesis remain healthy, but the condition is associated with increased risks of blood cancers, cardiovascular disease and inflammatory disorders. Understanding how aging transcriptional programs interact with these mutations could help explain why some clones expand rapidly while others remain stable, and why only a fraction progress toward malignancy.
Artificial intelligence is emerging as a possible tool for integrating these complex signals. Conventional analysis often examines one data type at a time, but machine-learning models can identify patterns across gene expression, chromatin state, methylation, protein abundance, metabolism and clinical outcomes. In principle, such models could classify HSCs according to biological age, predict which cells are likely to become myeloid-biased, identify early signatures of clonal expansion or estimate how a patient might respond to transplantation and other treatments. Network-based approaches may also help reconstruct regulatory relationships, revealing how transcription factors, signaling pathways and epigenetic enzymes cooperate to maintain or disrupt stem-cell function.
Yet the authors present AI as a promising research partner rather than a replacement for biological validation. HSC datasets are often small, technically variable and generated using different experimental platforms. Samples may come from different tissues, ages, disease states or species, making direct comparison difficult. Algorithms can also reproduce biases present in their training data or detect correlations that do not represent causal mechanisms. A gene signature associated with aging, for example, may reflect inflammation, medication exposure or changes in the marrow niche rather than an irreversible alteration in the HSC itself. Reliable AI applications will therefore require standardized sampling, carefully annotated longitudinal datasets, interpretable models and experiments that test whether computational predictions hold true in living systems.
The review ultimately frames HSC aging as a dynamic and potentially modifiable process rather than an unavoidable cellular countdown. Future therapies could aim to restore healthier epigenetic regulation, improve mitochondrial and DNA-repair capacity, reshape inflammatory signaling or rejuvenate the bone-marrow niche. Such strategies would need to be precisely controlled: stimulating aged stem cells indiscriminately could increase exhaustion, abnormal clonal expansion or cancer risk. The most effective interventions may instead be tailored to the molecular state of an individual’s HSC population, using multi-omics measurements and AI-assisted analysis to distinguish reversible dysfunction from permanent damage. As these technologies mature, the study of blood formation is moving toward a more detailed view of aging—one in which the fate of the immune system may be read in the regulatory circuits of individual stem cells.
Subject of Research: Hematopoietic stem-cell aging, transcriptional regulation, multi-omics analysis and potential applications of artificial intelligence.
Article Title: Hematopoietic stem cell aging: a review of transcriptional and multi-omics insights and potential paths for AI integration
Article References: Park, B., Yanai, H., Ding, J. et al. Hematopoietic stem cell aging: a review of transcriptional and multi-omics insights and potential paths for AI integration. Exp Mol Med (2026). https://doi.org/10.1038/s12276-026-01805-0
Image Credits: AI Generated
DOI: https://doi.org/10.1038/s12276-026-01805-0
Keywords: Hematopoietic stem cells, stem-cell aging, transcriptional regulation, multi-omics, single-cell sequencing, epigenetics, clonal hematopoiesis, bone-marrow niche, artificial intelligence, inflammaging.
