Earlier detection of biomarkers responsible for cancer relapse facilitates more rational cancer treatment regimens to be designed. Herein, we develop a mass cytometry-based strategy for unbiased mining of cell subsets that potentially contribute to cancer recurrence through panoramic examination of the immunophenotypic features and multidrug resistance characteristics. The incorporation of metal tags enables multiplexed information of single cells to be interrogated based on metal fingerprint. Using acute lymphoblastic leukemia (B-ALL) as a showcase, we show overexpressed multidrug resistance biomarkers, i. e., BCRP, Bcl-2, MRP1, and P-gp in B-ALL cells compared with healthy control, and a positive correlation among different multidrug resistance biomarkers. Different cell subsets with multidrug resistance are well-defined, featured with CD34CD38CD10 and CD34CD38CD10. Importantly, we uncovered that CD34 expression level is positively correlated to multidrug resistance, indicative of a higher potential of immature cells to induce B-ALL relapse. In addition, the cell subsets positively expressing CD73 and CD304 (CD34CD10CD304; CD34CD38CD10CD73) also overexpress multidrug resistance biomarkers, suggesting that they may serve as additional new biomarkers for B-ALL stratification and prognosis. Our data provide the first evidence that highly expressed multidrug resistance biomarkers in certain cell subpopulations with specific immunophenotypes may potentially induce B-ALL recurrence. The incorporation of multidrug resistance features with cell phenotypes using mass cytometry proposed in this study provides a general strategy for risk assessment and the prediction of recurrence of different types of cancers.