A cross platform m7G risk score based on five gene pairs stratifies prognosis and stemness in acute myeloid leukemia

AML exhibits marked heterogeneity in genetics, stemness, and treatment response. We profiled m7G regulator expression across multi-cohort AML transcriptomes and identified two m7G-related phenotypes with distinct survival and enrichment of adverse cytogenetics and primitive hierarchy. We then developed a cross-platform five-gene-pair m7G risk score using LASSO-Cox modeling. The score robustly stratified overall survival in TCGA and external cohorts (BEAT AML and GSE37642) and remained an independent prognostic factor. High risk was associated with increased leukemic stemness, including enrichment of HSC/MPP-like states in single-cell AML and higher scores in LSC+ fractions. The score also discriminated induction response (AUC 0.838) and increased in paired diagnosis-relapse samples, supporting relevance to chemoresistance. Drug sensitivity integration further nominated candidate compounds with preferential activity in high-risk AML. These findings connect m7G-related transcriptional programs to AML risk and provide a practical tool for refined prognostication and therapeutic prioritization.