An immune-metabolic transcriptomic balance stratifies acute myeloid leukemia and reveals distinct immunological and...

Acute myeloid leukemia (AML) remains genetically and clinically heterogeneous, and current cytogenetic risk categories do not fully capture biologic diversity or therapeutic vulnerability. We investigated whether a simple THBS1/VEGFA transcriptional ratio could define molecularly distinct AML subsets with prognostic and therapeutic relevance. The ratio was calculated as log₂(TPM_THBS1 + 1) - log₂(TPM_VEGFA + 1) and evaluated in the TCGA-LAML cohort (n = 132), with validation in BeatAML and outcome-independent external validation in GSE37642. Associations with overall survival, cytogenetic risk, leukemia-associated transcriptional programs, and microenvironmental features were assessed using differential expression, pathway enrichment, protein-protein interaction, Gene Ontology, and immune/stromal inference analyses, together with ex vivo drug-sensitivity profiling in the validation cohort. The THBS1/VEGFA ratio stratified overall survival more strongly than THBS1 or VEGFA alone and improved prognostic resolution within cytogenetic risk groups. HIGH-TRatio AML was characterized by an immune-associated molecular state marked by inflammatory signaling, myeloid-skewed immunomodulatory features, and extracellular matrix/microenvironmental remodeling, whereas LOW-TRatio AML showed a metabolically stressed program enriched for MYC targets, unfolded protein response, DNA repair, and proteostasis-related pathways. These transcriptional states were reproducible across cohorts and supported by compact, directionally consistent gene sets. Exploratory drug-sensitivity analysis suggested trends toward broader sensitivity among LOW-TRatio samples and relative resistance to several targeted agents among HIGH-TRatio samples, including VEGFR/multikinase inhibitors. These findings identify the THBS1/VEGFA ratio as a biologically interpretable transcriptomic marker that refines AML stratification beyond cytogenetics and captures clinically relevant molecular heterogeneity.