Fig. 1: Integrated scRNA-seq analysis of healthy and RR-AML bone marrow samples.

A Schematic overview and the number of samples and cells in this study. The cellular heterogeneity of RR-AML patients was deconstructed by integrating scRNA-seq data of seven RR-AML patients (in-house) and 20 healthy controls (GSE120221). Chemotherapy-induced reprogramming was investigated by analyzing longitudinal scRNA-seq data of 15 samples from six patients (in-house dataset and the public GSE116256 data). Results derived from scRNA-seq data were further validated by bulk transcriptomic data and in vitro and in vivo experiments. B UMAP visualization of cell type identification based on SingleR. C Heatmap showing the correlation of 98 clusters. Four major cell types were identified by unsupervised clustering, including HSC/Progenitors (brown), myeloid cells (green), erythrocytes (pink), and lymphocytes (blue). D UMAP visualization of 118,565 single cells that passed quality controls, colored by seven RR_AML patients and twenty HCs. E Heatmap showing the density ratio of the UMAP projections of RR-AML and healthy bone marrow cells. The UMAP visualization is split into 400 × 400 bins. F Boxplot showing the proportion of cell types in RR-AML patients and HCs. Wilcoxon rank-sum test was used to measure the differences between the two groups. G UMAP visualization of the identified leukemia-like cells. Cells from HCs were shown in gray, normal-like cells from AML patients were shown in sky blue, and leukemia-like cells from AML patients were shown in red. H Scatterplot showing the correlation between the proportions of leukemia-like cells predicted by this study (x-axis) and the classifier of Galen et al. (y-axis).