miRNA expression clusters in 111 BCP-ALL cases based on unsupervised consensus clustering and clinical impact of the MLC miRNA signature. (A) Heat map of miRNA expression by 111 primary samples from the TARGET cohort, along with the clinical information for each case. Two stable clusters were identified by consensus clustering of the 111 samples using 500 miRNAs (the distance method: Pearson) (supplemental Figure 2A). (B) Gene expression profiling of the 111 primary samples is shown as a 2-dimensional tSNE plot. Each dot represents 1 sample. The top 1000 most variable mRNA genes (based on the median absolute deviation) (top) and the top 500 most variable miRNAs (bottom) were selected and processed by the tSNE algorithm, with a perplexity score of 30. Genetic BCP-ALL subtypes are highlighted in different colors. Circles and triangles indicate cluster 1 and cluster 2, respectively. (C) MA plot for differential expression analysis between cluster 1 and cluster 2 generated by DESeq2: for each miRNA, the log2 (fold change) (log2[cluster 2/cluster 1]) is plotted (A, y-axis) against the log2 (average normalized expression) of the gene in the 2 clusters (M, x-axis). Significantly differentially expressed genes with an adjusted P value < .05 are shown in red. (D) Volcano plot showing differences in miRNAs expressed in cluster 2 samples (compared with cluster 1 samples) among the highly expressed miRNAs with a median of ≥25 reads per million. Significantly differentially expressed miRNAs showing a log2-fold change (cluster 2/cluster 1) > 1 or < −1, and an adjusted P value < .05, are shown in red. (E) Kaplan-Meier survival curves of EFS and OS for 111 BCP-ALL cases and 23 hyperdiploid BCP-ALL cases with and without the MLC signature. P values were calculated by the log-rank test. (F) Forest plot showing multivariate Cox regression analysis of the effect of different parameters on EFS. Squares represent the hazard ratio, and horizontal lines represent the CI. WBC at Dx, white blood cell count at diagnosis.