Supplementary Figure 3: MetaCell analysis of tier 3 cells.

a-c, Selection of 376 marker genes for tier 3 clustering (labeled in red, Methods). Plots depict gene specifity (the fraction of its total expression that is concentrated in 20% of the cells) against its total expression (a), gene correlation to the total UMI count of cells against total gene expression (c), and gene variance to mean ratio against its mean value (calculated on a UMI table down-sampled to 750 UMI/cell). d, Co-clustering of tier 3 (Lin- c-Kit + ) cells. Values indicate how likely two cells are to belong to the same meta-cell in each bootstrap iteration. Color bars depict lineage annotation as in Fig. 2a–b. e, Meta-cell distribution across 105 tier 3 amplification batches; meta-cells are annotated as in Fig. 2a. Color bar represents individual mice and FACS sorting sessions. f, Expression of 15 lineage specific marker genes used for meta-cell annotation across 138 fine grained meta-cells of tier 3 cells (Fig. 2a–b). g, Digital expression of lineage markers across tier 3 cells. h, Expression of key lymphoid and B cell genes in tier 3 meta-cells. Axes represent geometric mean divided by the median value across all meta-cells. Meta-cells are annotated by lineage marker as in Fig. 2a–b. i, In silico gating schemes for seven classical hematopoietic progenitor populations. Single cells were assigned to gates after sorting by their recorded FACS indices. j, Localization of conventional hematopoietic progenitor FACS-based populations when projected onto all tier 3 connected components.