Fig. 2.

Hierarchical clustering and survival curves of the 27 B-ALL patients based on expression of top selected genes in responding vs. no responding analysis. a Top 99 genes providing the biggest expression differences between good and poor response. p < 0.05 and fold change >2 were used to cluster the 27 B-ALL patients. b Distribution of 20 selected genes in B-ALL from list of 99 genes more differentially expressed between responders and no-responders to induction treatment. Genes with p < 0.03, fold change >3 were selected to cluster the 27 B-ALL bone marrow samples. Clustering analysis shows that set of 20 genes can distinguishes the same 3 groups identified with our list of 99 genes candidate predictors of response to therapy (blue, red and green bars). Kaplan Meier curves for EFS (c) and OS (d) in good and poor prognostic groups according to gene expression profile. Twenty-seven patients were assigned to either predicted good prognosis group (PGP) or predicted poor prognosis group (PPP) based on expression of 99 differentially expressed genes between responders and non-responders to induction therapy