A single-cell atlas and aging clock define biological age and risk-associated stem cell states in human hematopoiesis

Aging of hematopoietic stem and progenitor cells (HSPCs) impairs regenerative capacity and predisposes to hematological diseases. Here, we constructed a comprehensive single-cell transcriptomic atlas comprising 186,123 CD34⁺ HSPCs spanning early prenatal development (6 post-conception weeks) to late adulthood (74 years). We identified six molecular programs (MPs) that define distinct functional states within the HSC/multipotent progenitor (MPP) compartment. Among these, two MPs exhibited conserved age-associated dynamics across independent datasets, highlighting their central roles in stem cell aging. Leveraging these age-associated programs, we developed a machine learning-based stem cell aging clock from 84 donors to predict chronological age from HSC/MPP transcriptomes. Applying this aging clock to acute myeloid leukemia (AML) defines transcriptional age deviation (TAD), a measure of biological age divergence in leukemic HSC/MPPs. TAD captures disease-associated variation linked to genetic risk stratification and patient survival. Our study provides a high-resolution reference map of human HSPC aging, establishes a stem cell-specific aging clock and demonstrates its utility for uncovering aging-related dysregulation with prognostic relevance in hematological malignancies.