Fig. 2: Mutational patterns across leukemia samples.

a, Copy-number state in the PAX5-ZCCHC7 locus, revealing a single loss in the initial ALL and a subsequent second hit in the relapse samples with nearby, but unique break points resulting in a constitutive loss of PAX5 and ZCCHC7. Both deletions were flanked by recombination signal sequence (RSS) motifs, indicating that these deletions were likely RAG-mediated. b, Exposures of the three identified signatures in this patient, with signature A corresponding to COSMIC reference SBS1, signature B to SBS5 and signature C to SBS87. The latter is associated with thiopurine exposure. Only branches with 100 SNVs or more are displayed. c, Schematic of treatment courses of the patient, along with thiopurine-related mutagenesis. As the progression root and the ALL relapse root (its descendant) harbor thiopurine-induced mutations, the split between the ALL and AML lineages must have happened during the treatment for the initial ALL. d, Estimated clone size of ancestral AML lineage across samples, as estimated from interrogating all mutant sites in diploid regions (n = 2,318) in the AML root branch (Methods). The red line indicates the background error rate estimated from 32 unmatched blood samples. Error bars denote the 95% CI around maximum likelihood binomial estimate of clone size. ***P < 0.001, one-sided binomial test with null hypothesis that number of variant counts is drawn from background error distribution. The exact P value for the ALL R2 sample is 1.5 × 10⁻¹⁵¹ and 0 for all AML-related samples. e, Genome-wide copy-number profile of the two AML samples, obtained 3 months apart, showing that the later AML acquired extensive aneuploidy in a short time, likely facilitated by the biallelic loss of TP53.