Figure 5

Predicted developmental state is inversely correlated with ribosomal protein expression in cALL cells and is a major source of intra-individual transcriptional heterogeneity. (A) Left: UMAP representation of the maturation spectrum of healthy pediatric and adult B cells (CD34+ → B cells → CD20 + B cells) used for the B cell developmental state classifier. Right: UMAP representation of the maturation spectrum of healthy pediatric and adult T cells (CD34+ → immature hematopoietic → T cells) used for the T cell developmental state classifier. Cells were projected onto the loess fit of the spectrum and assigned a pseudotime value of 0 to 1 from the first stem-like cell to the last mature cell. (B) Observed vs predicted pseudotime of healthy B and T cells using a hundred 70/30 cross-validation splits; mean RMSE was computed over all splits. (C) Density of predicted developmental state pseudotime distributions of leukemia cells per sample and intra-individual transcriptional cluster. (D) Boxplot of ribosomal protein (RP) expression percentage in leukemia cells per sample and intra-individual transcriptional cluster. (E) Pseudotime vs ribosomal protein expression in samples showing strong (HHD.1) vs weak (ETV6.RUNX1.2) intra-individual pseudotime and ribosomal protein expression shifts. (F) Heatmap of normalized and scaled expression of ribosomal protein genes per cell sorted from high to low. An expression gradient correlated to cluster assignment can be observed in sample HHD.1 but not ETV6.RUNX1.2.