Human PBMC atlas - HSCs

Cell annotations were generated through iterative hierarchical clustering. Broad lineages were first identified and then recursively subdivided into finer sublineages and sub-sublineages (e.g., T → CD4 T → Effector CD4 T). Clusters were iteratively re-clustered at multiple resolutions, split or merged based on biological evidence, and refined through repeated quality-control steps. Final annotations were determined by consensus of an international group using canonical marker genes, one-vs-all and all-vs-all differential expression analyses, and cluster relationships within the hierarchical framework. Low-quality cells, doublets, and contaminating populations were identified and removed throughout the annotation process; following each round of removal, datasets were reprocessed and re-clustered to improve cluster resolution and biological consistency before subsequent annotation refinement.