Establishment of BCP-ALL RT signatures. (A) Patient BCP-ALL stratification using RT profiling data. RT signatures are first identified by unsupervised k-means clustering of all 50-kb genomic segments that change between early and late replication from RT log₂ ratio of +0.5 or more to/from –0.5 or less, respectively.⁷  These stringent criteria identify features that replicate significantly differently in specific clusters of patients (P < 2 × 10⁻¹⁶ using Student t tests). Branches of the dendrogram were constructed on the basis of correlation values (distance = 1 minus the correlation value). (B) Exemplary RT profiles of RT signatures. The top panel show an example of an RT feature (containing the ROR1 gene) of TCF3-PBX1–positive (red) vs TCF3-PBX1-negative (gray) BCP-ALL cells. (C) Hierarchical clustering of BCP-ALL samples for RT signature 1. (D) Reverse transcription polymerase chain reaction (RT-PCR) to determine the TCF3-PBX1 fusion messenger RNA (mRNA) in patient samples classified as NOS (PALCRG and 10-795). Specific primers for the TCF3-PBX1 fusion mRNA and for ABL as a control were used. Two TCF3-PBX1–positive cell lines were used as positive controls (RCH-ACV and Kasumi2). Normal B-lymphoid cells (GM12878) were used as a negative control. (E) Heatmap of expression patterns for genes within the RT signatures 1. Chr, chromosome; H, hyperdiploid; N, NOS; T, T-cell acute lymphoblastic leukemia.