AI Identifies Regulatory Hub Controlling Blood Stem Cell Aging

Researchers at Tohoku University have leveraged artificial intelligence to identify a primary genetic regulator governing the aging of hematopoietic stem cells, providing a molecular framework for understanding age-related blood production decline. The findings, published in Science Advances, demonstrate how machine learning models can decode complex cellular transitions that traditionally resist systematic mapping. Hematopoietic stem cells in bone marrow sustain lifelong blood formation by balancing self-renewal with the differentiation of mature red blood cells, lymphocytes, and platelets. With advancing age, these cells undergo functional decline, characterized by increased stem cell abundance but reduced output of mature blood lineages. This imbalance contributes to anemia, immunosenescence, thrombosis, and increased susceptibility to blood cancers. Led by researcher Keiyo Takubo, the Tohoku team analyzed single-cell gene expression profiles from young and aged mouse hematopoietic stem cells. They deployed Geneformer, a deep learning architecture pretrained on approximately thirty million cells and fine-tuned on 160,000 young and aged blood stem and progenitor cells. The AI model screened 143 high-priority genetic candidates, ultimately pinpointing the transcription factor Pbx1 as a central regulatory hub. Experimental validation confirmed Pbx1 as a decisive factor in remodeling stem cell fate. In aged cells, Pbx1 activity correlates strongly with gene networks that maintain an immature state while biasing differentiation toward platelet production. Artificially elevating Pbx1 in young stem cells recapitulated key aging signatures, activating over seventy-three percent of genes typically upregulated during natural aging. In vivo transplantation assays further demonstrated that Pbx1-overexpressed cells suppress the erythropoiesis regulator Gata1, resulting in reduced red blood cell output and a relative surge in platelet generation. The study redefines aged hematopoietic stem cells not as functionally degraded, but as cells that have transitioned into a distinct, stable developmental state with inherent lineage biases. By integrating AI-driven prediction, large-scale single-cell screening, and multi-omics analysis, the research establishes a critical checkpoint in hematopoietic aging. Subsequent investigations will focus on verifying whether the Pbx1 mechanism operates identically in human systems and evaluating its potential as a therapeutic target for age-related hematological disorders, including chronic anemia, clonal hematopoiesis, and myeloid malignancies.