In previous work, single-cell analysis has been used to define the clonal architecture of acute lymphoblastic leukemia5 by multiplex fluorescence in situ hybridization and to define the tumor evolution of breast cancers6 using next-generation sequencing to quantify genomic copy number within individual nuclei. In both cases, the common progenitor clone of the diverse subclones was marked and identified by genome aberrations (chromosomal translocations and/or copy-number alterations). The outstanding question was whether somatic variants would reveal a similar clonal pattern at the sequence level. Understanding this issue in detail will have profound clinical implications. There are now several targeted therapies directed against cancer genes that are mutated (such as BRAF and EGFR) or amplified (such as HER2), which some have called ‘actionable mutations’. Tumor heterogeneity poses a considerable challenge to such therapies. For example, how will clinicians decide whether a targeted therapy is indicated if the mutation is present in only a minor proportion of cancer cells?