Autologous bone marrow transplantation (ABMT) is widely used as treatment for malignant disease. Although the major cause of treatment failure is relapse, it is unknown if this arises entirely because of residual disease in the patient or whether contaminating cells in the rescuing marrow contribute. Attempts to purge marrow of its putative residual malignant cells may delay hematopoietic reconstitution and are of uncertain efficacy. We now describe how retrovirus-mediated gene transfer may be used to elucidate the source of relapse after ABMT for acute myeloid leukemia and to evaluate the efficacy of purging. Clonogenic myeloid leukemic blast cells in patient marrow can be transduced with the NeoR gene-containing helper-free retrovirus, LNL6, with an efficacy of 0% to 23.5% (mean, 10.5%). Transduced colonies grow in selective media and the presence of the marker gene can be confirmed in individual malignant colonies by polymerase chain reaction. If such malignant cells remain in harvested “remission” marrow, they will therefore be marked after exposure to LNL6. Detection of the marker gene in the malignant cells present at any later relapse would be firm evidence that residual disease contributed to disease recurrence, and would permit rapid subsequent evaluation of purging techniques. The technique also marks normal marrow progenitors from patients with acute myeloblastic leukemia. These colony-forming cells can be detected in long-term marrow cultures at a frequency of 1% to 18% for up to 10 weeks after exposure to the vector. Animal models and analysis of probability tables both suggest that these levels of marking in vitro are sufficient to provide information about the mechanisms of relapse and the biology of marrow regeneration in vivo. These preclinical data form part of the basis for current clinical studies of gene transfer into marrow before ABMT.