Endometrial polyps (EPs) are among the commonest gynaecological lesions, yet the cellular origins and molecular logic of their excessive growth remain undefined.
Here we chart a single-cell and spatially resolved transcriptomic atlas of EPs and matched healthy controlled endometrium. We identify two clonally expanded epithelial subpopulations—IGF1⁺ and MCC⁺—that display stem-like self-renewal and are both necessary and sufficient for polyp initiation in endometrial organoids. Ligand–receptor mapping uncovers an autocrine IGF1 circuit that sustains hyperproliferation; chemical or genetic blockade of this axis reverses the phenotype in patient-derived xenografts and organoid platforms. Integrating longitudinal clinical data with multi-omics signatures, we develop a machine-learning classifier that predicts effective drugs for individual application.
These findings establish EPs as a cell-selective, targetable disorder of endometrial homeostasis and provide an immediate route to precision interception of recurrence and oncogenic transformation.