Extended Data Fig. 9: Correlation of HSPC states with acute leukemia scRNA-seq and assessment of age state prognostic independence in acute myeloid leukemia (AML).

(a) scRNA-seq datasets of human B lymphoblastic ALL⁶⁵ and AML²¹ were classified against the 22 HSPC states using singleCellNet with classification scores for each state indicated. Only cells annotated as expressing CD19 are included in the ALL analysis and only cells annotated as bearing AML-associated somatic mutations are included in the AML analysis. (b) Cox multivariate hazard ratio analysis comparing age composite score as an independent variable to either age at diagnosis, bone marrow blast percentage, FAB classification, ELN 2017 favorable risk designation, or mutational status of CEBPA, FLT3-ITD, IDH, NPM1, RAS, TP53, and DNMT3A in the BeatAML dataset⁶⁸. Hazard ratios and 95% confidence intervals for hazard ratios are indicated. (c) Comparison of mixed age scores for AMLs that were preceded by MDS or not preceded by MDS in the BeatAML dataset. The points representing individual AMLs are color-coded by whether they are in the high or low age score group. Significance was calculated using two-sided Wilcoxon rank-sum testing. (d) Overall survival probability in the TCGA dataset for high and low AML age score groups was compared with p-value shown. (e) Overall survival probability in the TCGA dataset for high and low TF score groups was compared with p-value shown. All survival curve p-values were calculated by log-rank test.