Increasing evidence suggests that breast cancer is caused by cancer stem cells and the cure of breast cancer requires eradication of breast cancer stem cells. In this study, we established and characterized a sphere culture model derived from side population cells from the human breast cancer cell line MCF7. The sphere culture could be maintained long term and was enriched in cells expressing known breast cancer stem cell marker CD44+CD24 -. These sphere cells showed higher colony formation ability in vitro and higher tumorigenicity in vivo than MCF7 cells, suggesting the enrichment of breast cancer stem/progenitor cells. To identify compounds that preferentially inhibit the sphere cells, we performed a compound library screening. Two lead compounds, NSC24076 and NSC125034 and an analog of NSC125034, 8-quinolinol (8Q), were identified as having preferential activity against the sphere cells. 8Q showed some antitumor activity alone but had much better therapeutic effect and relapse prevention when combined with paclitaxel than either 8Q or paclitaxel alone in both MCF7 and MDA-MB-435 xenograft models. We propose that compounds selectively targeting cancer stem/progenitor cells when combined with standard chemotherapy drugs may produce an improved treatment of cancer without significant relapse.