Abstract
The representation of driver mutations in preleukemic hematopoietic stem cells (pHSCs) provides a window into the somatic evolution that precedes acute myeloid leukemia (AML). Here, we isolate pHSCs from the bone marrow of 16 patients diagnosed with AML and perform single-cell DNA sequencing on thousands of cells to reconstruct phylogenetic trees of the major driver clones in each patient. We develop a computational framework that can infer levels of positive selection operating during preleukemic evolution from the statistical properties of these phylogenetic trees. Combining these data with 67 previously published phylogenetic trees, we find that the highly variable structures of preleukemic trees emerge naturally from a simple model of somatic evolution with pervasive positive selection typically in the range of 9%-24% per year. At these levels of positive selection, we show that the identification of early multiple-mutant clones could be used to identify individuals at risk of future AML.