In human T-ALL, the expression level of NOTCH1 dimer target genes was inversely correlated with a proapoptosis gene signature and led to the identification of a subset of high-risk patients. (A) Unsupervised clustering of single-cell transcriptomes of primary leukemia cells derived from 41 different patients with T-ALL, including 23 in-house sequenced and without any expansion in immunocompromised mice and 18 patients with T-ALL from 3 publicly available data sets, namely GSE227122, GSE161901 and GSE132509. The tSNE plot shows the distribution of 20 distinct cell clusters that were identified using the Leiden algorithm based on the expression of highly variable genes from the single-cell RNA sequencing (scRNA-seq) data. (B) The tSNE plots showing the enrichment scores of the Notch1 dimer (NDIM) signature (left panel) and the proapoptosis (KEGG) gene set (right panel) in the 20 groups identified from the scRNA-seq data. (C) Comparison of the enrichment scores between the Notch1 dimer (NDIM) and the proapoptosis (KEGG) signatures in each cell cluster identified from the scRNA-seq data. ∗∗∗P < .001 (Student t test). (D) Heat map of the antibody-derived tag (Ab-seq) expression across cell clusters from the scRNA-seq data from primary samples that were sequenced in-house and derived from 23 different patients with T-ALL. The clusters are ordered by hierarchical clustering based on the Euclidean distance of the log-fold change values. (E) 2-Dimensional clustering of 262 patients with T-ALL from the COG TARGET study using the k-means algorithm. The top most variable genes were identified using the median absolute deviation and were used to generate the principal component analysis plot. In the 5 identified k-means clusters, each dot represents a patient. (F) Box plots showing the distribution of the Gene Set Variation Analysis (GSVA) score of the NOTCH1 dimer–associated (NDIM) signature and the normalized RNA expression levels of HES4 or BCL2L1 transcripts in the 5 identified k-means clusters of patients with T-ALL from the COG TARGET study. GSVA was used to compute the NDIM score based on significant genes from the RNA-seq data sets of human synthetic T-ALLs with a documented sequence-paired binding site in the promoter region. (G) Scatterplot of the correlation analysis between the HES4 and the NOTCH1 normalized RNA expression in the 5 identified k-means clusters of the 262 patients with T-ALL from the COG TARGET study. The Pearson correlation coefficients (R) and P values (p) are reported as indicated. (H) Box plots of the GSVA scores for the NDIM and proapoptosis (KEGG) signatures as indicated in the 5 identified k-means clusters of patients with T-ALL. ∗∗∗P < .001 (Student t test). (I) Box plots of the GSVA scores for the NDIM signature as indicated in the patients with T-ALL who were subdivided into CD2^lowCD5^low (n = 142), CD2^highCD5^low (n = 26), CD2^lowCD5^high (n = 66), and CD2^highCD5^high (n = 25) groups based on the transcriptional level of the CD2 and CD5 genes from the RNA-seq data of each patients with T-ALL from the COG TARGET study. (J) Kaplan-Meier curve showing the survival probability of patients with T-ALL from the COG TARGET study, which were subdivided into NDIM high (n = 178) and NDIM low (n = 81) subsets based on the expression of NOTCH1 dimer target genes. (K) Kaplan-Meier curve showing the survival probability of patients with T-ALL who were subdivided into the HES4^lowBCL2L1^low (n = 38), HES4^lowBCL2L1^high (n = 23), HES4^highBCL2L1^low (n = 107), and HES4^highBCL2L1^high (n = 91) groups based on the transcriptional level of the HES4 and BCL2L1 genes from RNA-seq data of each patient with T-ALL from the COG TARGET study. (L) Kaplan-Meier curve showing the survival probability of patients with T-ALL who were subdivided into the CD2^lowCD5^low (n = 142), CD2^highCD5^low (n = 26), CD2^lowCD5^high (n = 66), and CD2^highCD5^high (n = 25) groups based on the transcriptional level of the CD2 and CD5 genes from the RNA-seq data of each patient with T-ALL from the COG TARGET study. ns, not significant.