2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Sunday, December 13, 2026
06:00 p.m. - 08:00 p.m.
MCCNO - Hall F
Technology and Procedures, Assays, Molecular Testing, Omics Technologies, Bioinformatics, Computational Biology, Emerging Technologies, Profiling, Treatment Considerations, Combination Therapy, Drug Development, Therapy Sequence, Biological Therapies, Antibody Therapy, Transplantation, Checkpoint Inhibitor, Immunotherapy, Non-Biological Therapies, Chemotherapy, Adverse Events, Artificial Intelligence (AI), Deep Learning, Machine Learning, Biological Processes, Apoptosis, Pathogenesis, Genomics, Hematopoiesis, Immune Mechanism, Metabolism, Molecular Biology, Multi-Systemic Interactions, Clinical Practice (Health Services and Quality), Practice Models, Value, Diseases, Myeloid Malignancies, Acute Myeloid Malignancies, Chronic Myeloid Malignancies, Measurable Residual Disease, Research, Fundamental Science, AML, APL, Translational Research, MDS, MPN, CMML, Clinical Research, Epidemiology, Health Outcomes Research, Health Disparities Research, Real-World Evidence, Survivorship, Study Population, Human, Adult
PAPER

3443 Biologically informed multi-omics integration improves prognostic stratification in acute myeloid leukemia

Dona Hasini Vathsalya Gammune, PhD, Doan Bui2, Tongjun Gu, PhD3
1 Versiti Blood Research Institute, Milwaukee, Wisconsin, United States
2 Versiti Blood Research Institute, Milwaukee, Wisconsin, United States
3 Versiti Blood Research Institute, Milwaukee, Wisconsin, United States

Acute myeloid leukemia (AML) is a highly heterogeneous hematologic malignancy with substantial variability in treatment response and clinical outcomes. The European LeukemiaNet (ELN)-2022 classification remains the standard for risk stratification and guides therapeutic decision-making; however, it relies predominantly on cytogenetic abnormalities and recurrent genetic mutations. Despite its clinical utility, considerable heterogeneity persists within each ELN risk category, particularly among intermediate-risk patients, where individuals with similar cytogenetic and molecular profiles often experience markedly different clinical outcomes. These observations suggest that important biological determinants of disease progression remain incompletely captured by current risk classification, highlighting the need for integrative approaches that characterize disease biology across multiple molecular layers.

Gene expression and microRNA (miRNA) expression provide complementary insights into AML biology beyond genomic alterations alone. While gene expression reflects the functional transcriptional state of leukemic cells, miRNAs regulate post-transcriptional gene expression and play critical roles in hematopoietic differentiation, leukemogenesis, treatment response, and disease progression. Because these molecular layers are biologically interconnected, integrating both has the potential to more comprehensively characterize disease biology than either modality alone. To leverage this complementary information, we developed SurvMOCA-GNN, a biologically informed multi-omics deep learning framework that integrates gene expression, miRNA expression, biological interaction networks, cross-omics regulatory relationships, and clinical variables, including age, sex, and ELN-2022 risk categories, to generate unified patient representations that capture complementary molecular information across multiple regulatory layers. The model was developed using matched RNA sequencing, miRNA sequencing, and clinical information from 162 patients with AML in TCGA-LAML cohort and validated in an independent GAML cohort.

Integrated multi-omics analysis identified two prognostically distinct patient groups with significantly different overall survival (log-rank P = 2.67 × 10⁻²; HR = 2.4; 95% CI, 1.5–4.0), demonstrating that integrating complementary molecular information captures clinically relevant disease heterogeneity associated with patient outcomes. When projected onto the ELN-2022 classification, the integrated molecular profiles further resolved patients within each risk category into groups with distinct survival outcomes, with the greatest separation observed among intermediate-risk patients. These findings indicate that patients with similar cytogenetic and genomic risk can harbor fundamentally different underlying disease biology that is not fully reflected by current ELN-2022 classification. Importantly, analyses based on either gene expression or miRNA expression alone failed to fully resolve this prognostic heterogeneity, demonstrating that each molecular layer captures distinct but complementary biological information. By integrating transcriptional activity with post-transcriptional regulatory mechanisms, the proposed approach provided a more comprehensive characterization of AML biology and improved identification of patients with distinct clinical outcomes. Validation in an independent AML cohort demonstrated that the trained models maintained high predictive performance and generalized successfully across heterogeneous patient populations, supporting the robustness and potential clinical utility of this biologically informed multi-omics deep learning approach.

These findings demonstrate that prognostically important biological variation in AML arises from coordinated dysregulation across multiple molecular layers and cannot be fully captured by conventional cytogenetic, genomic, or single-omics analyses alone. Biologically informed integration of complementary molecular and clinical information complements the current ELN-2022 classification by improving prognostic stratification and identifying patients with distinct clinical outcomes within existing risk groups. This approach has the potential to refine risk assessment and support precision medicine in AML.

Acknowledgment:

Supported, in part, by IRG #22-151-37-IRG from the American Cancer Society and by the MCW Cancer Center, and Versiti Blood Research Institute Foundation.

Dona Hasini Vathsalya Gammune, PhD: No relevant conflicts of interest to declare.