High frequencies of activated NK cells predict poor clinical prognosis in high-risk B-ALL. (A) Schematic diagram depicting the steps to estimate the relative proportions of activated and resting NK cells in patients with B-ALL from the COG P9906 clinical trial (GSE11877) using CIBERSORT. The LM22 signature matrix (GSE22886), which contains the transcriptomic profile of resting NK cells and cytokine (IL-2, IL-15)–activated NK cells, was used to deconvolute the bulk transcriptomic profiles of COG P9906 patients with B-ALL. After excluding 4 patients who lacked NK cells, the remaining patients were assigned to 2 cohorts: resting NK > activated NK (n = 104) and activated NK > resting NK (n = 99). Comparison of RFS probabilities of COG P9906 patients with B-ALL divided into resting NK > activated NK and activated NK > resting NK groups (B), resting NK > activated NK CNS⁺ and activated NK > resting NK CNS⁺ groups (C), resting NK > activated NK MRD⁺, activated NK > resting NK MRD⁺, resting NK > activated NK MRD⁻, and activated NK > resting NK MRD⁻ groups (D), 100% activated NK and 100% resting NK groups (E), 100% activated NK CNS⁺ and 100% resting NK CNS⁺ groups (F), and 100% activated NK MRD⁺, 100% resting NK MRD⁺, 100% activated NK MRD⁻, and 100% resting NK MRD⁻ groups (G). Survival curves were calculated using the Kaplan-Meier method. Exact P values were calculated using the log-rank test.