Human haematopoietic stem and progenitor cells (HSPCs) exhibit heterogeneous lineage output, but the molecular programs underlying clonal fate remain poorly defined. To address this, we developed a human haematopoietic organoid supporting differentiation into 15 lineages and used it to track barcoded HSPC clones over time. By integrating single-cell transcriptomes, surface phenotypes, and clonal fate, we applied machine learning to identify clonal fate modules, gene and marker signatures predictive of lineage commitment. This approach uncovered hidden transcriptional and surface correlates of multipotency, including CD200, which marked a subset of HSCs with broad output capacity, which we leveraged to increase manufactured type 1 dendritic cell purity for immunotherapy applications. Our study provides a framework for decoding clonal fate decisions in human HSPCs and identifies molecular features that distinguish truly multipotent clones, advancing strategies for stem cell purification and therapeutic engineering
Tomei et al. (2025) studied this question.