Fig. 2: Visualization of AML single cells by UMAP and projection onto a reference NBM knn force graph.

a A knn force graph was constructed from 48,656 normal bone marrow (NBM) cells (top left) representing eight samples. Cells from individual AML samples (indicated in blue) were projected onto the reference NBM knn force graph (indicated in gray). The number of cells projected onto a region is indicated by the size of each hexagon and the proportion of mutated cells in that hexagon is indicated by shade of blue. Hexagons with too few genotype reads are marked in brown (three or fewer reads). The genomic subtype of each sample is indicated in the top left corner. Sample identifier and number of cells are indicated below each plot. A selection of genes mutated in each sample is presented below each plot, with the variant allele frequency indicated by whole exome sequencing denoted in parentheses. Genes targeted by single cell mutation detection using an in-house PCR are indicated in bold. Genes determined to have heterozygous mutations are underlined; these were used for inferring the proportion of mutated cells. Genes with presumed passenger mutations are denoted in gray. b–d UMAP representation of 245,073 cells from 38 AML samples and 8 NBM samples. b Cells are colored based on graph-based clustering, which identified 58 distinct clusters of cells within the data. c Cells are colored by cell type based on classification using a reference dataset of NBM cells. d Cells are colored by AML subtype or NBM sample type. Notably, all AML samples were associated with a distinct cluster of cells outside the regions occupied by cells from NBM samples. The majority of cells in these AML-specific clusters were classified as hematopoietic stem cells (HSC) or lymphoid-primed multipotent progenitors (LMPP) based on the reference dataset. The cells in the AML-specific clusters were therefore reclassified as AML immature. GMP Granulocyte-monocyte progenitors. Source data are provided as a Source Data file.