Multi-Omics Integration Is Reshaping How We Understand and Treat AML

Acute myeloid leukemia (AML) is a highly aggressive blood cancer driven by clonal proliferation of undifferentiated myeloid precursors. Despite advances in genomic profiling and targeted therapies, most patients still relapse or develop treatment resistance. The core problem is molecular heterogeneity — patients with similar genetic mutations can have dramatically different outcomes, and single-omics approaches consistently fail to predict therapeutic responses accurately. This review argues that only by combining multiple biological data layers simultaneously can researchers and clinicians capture the true complexity of AML.

At the genomic level, whole genome and whole exome sequencing have identified a growing catalog of driver mutations beyond the classic FLT3-ITD (found in 25–30% of cases) and NPM1 (approximately 30% of patients). A pivotal 2025 study by Khouri et al. analyzing cohesin complex mutations (STAG2, RAD21, SMC1A, SMC3) found these independently predicted adverse outcomes — median overall survival of 8.2 versus 18.6 months (p < 0.001) — and synergized with FLT3-ITD to accelerate leukemogenesis in vivo. Integrated ATAC-seq and transcriptomic data revealed dysregulation of HOXA/B stemness pathways in cohesin-mutant AML, a finding only accessible through multi-omics.

Transcriptomics, particularly single-cell RNA sequencing, has been transformative. Zeng et al. profiled 40,000 cells from 16 AML patients and mapped mutation-specific transcriptional states across leukemic stem cells, progenitors, and blasts. Key findings: FLT3-ITD mutations drive oxidative phosphorylation (OXPHOS) hyperactivity in monocyte-like blasts conferring venetoclax resistance; NPM1-mutant cells exhibit disrupted HOX expression promoting stemness; and co-occurring DNMT3A/RUNX1 mutations activate NF-κB/TNF inflammatory pathways in leukemic stem cells. The LSC17 17-gene signature, validated across multiple cohorts, stratifies patients into high- and low-risk groups independently of mutational status.

Proteomics has added another critical dimension. Reville et al. (2025) derived an 8-protein inflammatory risk score (IL-6, IL-8, TNF-α, CXCL10, OPG, MCP-1, Galectin-9, FLT3LG) from serum proteomics of 452 AML patients. This score outperforms ELN 2022 genetic risk criteria, stratifying patients with a hazard ratio of 3.2 (p < 0.001), detecting early relapse with AUC = 0.89, and predicting overall survival more accurately than mutation-based classifiers by capturing microenvironment-driven resistance. Mechanistically, IL-6 activates JAK/STAT signaling in leukemic stem cells while Galectin-9 promotes T-cell exhaustion in the bone marrow niche.

Epigenomic findings are equally compelling. IDH1/2-mutated AML — characterized by a hypermethylator phenotype — shows heightened sensitivity to hypomethylating agents like azacitidine and decitabine. Menin inhibitors disrupting the KMT2A-menin interaction have shown durable remissions in phase II trials for KMT2A-rearranged AML. Liquid biopsy platforms integrating ctDNA genomics and methylation signatures now achieve 90% concordance with bone marrow biopsies for minimal residual disease monitoring — a potential game-changer for treatment surveillance. Challenges including data harmonization, prohibitive costs, and the need for prospective validation of multi-omics-derived targets remain, but the clinical translation is accelerating rapidly.