Fig. 3: Leukemia stem cell transcriptional signature for pediatric AML.

a The LSC17 gene signature was previously generated based on LASSO Cox regression on 47 genes enriched in LSC AML cell populations (LSC47). Analyzing LSC47 gene expression data within our cohort, AMLs cluster based on underlying fusion category. b The circos plot on the left indicates the previously described LSC17 gene set. Conversely, the circos plot on the right indicates the 17 most predictive genes within our training cohort using the same LASSO based Cox regression analysis. Subsequent risk stratification model building considers all 47 upregulated LSC genes (LSC47). c Kaplan–Meier estimates for the probability of EFS based on LSC17 versus LSC47 gene signatures and associated area under the curve receiver operating characteristic (AUC ROC) curve plotting true positive rates versus false positive rates as a function of LSC17 and LSC47 score thresholds. d Additionally, when AMLs are grouped based on underlying fusion, each class is associated with a distinct LSC gene set. e LSC47 variance and t-tests based on fusion category (n = 753 patients from the training cohort). Box plot data are presented as median values with hinges corresponding to the 25th or 75th percentiles and whiskers corresponding to the 10th or 90th percentiles (left panel). P-values were calculated based on two-sided t-tests (right panel). Source data are provided as a Source Data file. Kaplan–Meier estimates for the probability of EFS and AUC ROC curves among f KMT2A and g Other or No Fusion AML cohorts based on LSC17 versus LSC47. Survival differences were determined using the log-rank test (two-sided and without multiple-testing adjustments).