Individual cancer cells exhibit functional heterogeneity of many cancer hallmarks including the capacity for sustaining long-term clonal maintenance, a stemness property involving self-renewal. Our studies established that only rare AML cells possessed such leukemia stem cell (LSC) properties and that AML is a cellular hierarchy. LSC were found to be highly relevant to human disease as gene signatures specific to LSC were much more predictive of patient response to therapy and overall survival compared to the bulk non-LSC AML cells. We have now provided insights into how LSC develop during leukemogenesis. Through sequencing of purified populations of normal blood cells, AML cells, and xenografts from paired diagnosis/relapse AML samples, we tracked the full arc of leukemia development in humans: from the cell of origin (an HSC) that acquires the first genetic mutation; to pre-leukemic clonal expansion of HSC; the generation of genetically diverse LSC; and finally to the cellular origin of relapse (rare LSC subclones) within the diagnosis sample. Similar studies have been undertaken in B-ALL where relapse-fated subclones were found within the diagnosis sample before exposure to chemotherapy. These clones possess distinct epigenetic and metabolic properties that underlie their partially resistance to chemotherapy and they possess stemness signatures that drive their ability to regenerate relapse disease. These studies resulted in identification of novel therapeutic targets of relapse-fated subclones to begin envisioning approaches to eradicate relapse before it develops. Finally, the identification of clonally expanded pre-leukemic HSPC in the diagnosis blood sample of many AML samples predicted that it should be possible to identify individuals who are at risk for progressing to AML long before AML develops. We genotyped individuals from the general population who were enrolled in the large cohort (EPIC) who eventually developed AML and compared them to the enrollment sample of others who never progressed to AML. We have found a clear signature that is able to predict with high accuracy those individuals who have age related clonal hematopoiesis who progress to AML, almost 10 years prior to AML development. This is distinct from individuals who have benign clonal hematopoiesis who never progress. These new findings offer the potential for future intervention to permit AML prevention trials.