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

Paper Details

Abstracts, authors, affiliations, and disclosures.

Saturday, December 12, 2026
05:30 p.m. - 07:30 p.m.
MCCNO - Hall F
Technology and Procedures, Assays, Molecular Testing, Omics Technologies, Pathology, Bioinformatics, Computational Biology, Emerging Technologies, Profiling, Treatment Considerations, Combination Therapy, Supportive Care, Biological Therapies, Antibody Therapy, Natural Killer (NK) Cell Therapies, Transplantation, (CAR)-T Cell Therapies, Immunotherapy, Non-Biological Therapies, Chemotherapy, Artificial Intelligence (AI), Machine Learning, Biological Processes, Apoptosis, Pathogenesis, Genomics, Immune Mechanism, Metabolism, Molecular Biology, Multi-Systemic Interactions, Clinical Practice (Health Services and Quality), Clinical Guidelines, Practice Models, Systems-Based Hematology, Diseases, Bone Marrow Failure Syndromes, Acquired Marrow Failure Syndromes , Genetic Disorders, Myeloid Malignancies, Acute Myeloid Malignancies, Chronic Myeloid Malignancies, Measurable Residual Disease, Research, Fundamental Science, AML, Translational Research, MDS, MPN, Clinical Research, Patient-Reported Outcomes, Real-World Evidence, Registries, Study Population, Human, Adult, Elderly, Pediatric, Young Adult
PAPER

1675 Lncsigaml: An AI-derived long non-coding RNA signature enables transcriptome-based prognostic stratification in acute myeloid leukemia across 2,681 patients

Lan TN Trinh, MD, Thuy An Nguyen, PhD2, Danh-Tai Hoang, PhD3, Bon Q. Trinh, PhD4
1 University of Virginia, Charlottesville, Virginia, United States
2 University of Virginia, Charlottesville, Virginia, United States
3 Cedars-Sinai, Los Angeles, California, United States
4 University of Virginia, Charlottesville, Virginia, United States

Acute myeloid leukemia (AML) is a biologically heterogeneous hematologic malignancy characterized by abnormal proliferation and impaired differentiation of myeloid blasts, diverse genetic alterations, variable treatment responses, and markedly different clinical outcomes. Current risk stratification systems based on cytogenetic and genetic abnormalities do not fully capture prognostic heterogeneity, particularly among intermediate-risk patients, and lack readily translatable clinical tools. Long non-coding RNAs (lncRNAs) have emerged as key regulators of AML biology, contributing to leukemogenesis, impaired differentiation, apoptosis resistance, and treatment response. Their association with patient survival suggests that lncRNAs may provide prognostic information beyond established genetic markers.

In this study, we developed and validated a bone marrow-derived lncRNA prognostic framework for acute myeloid leukemia using transcriptomic profiles from three independent AML cohorts (TARGET-AML, Beat AML, and TCGA-LAML) comprising 2,681 pediatric and adult patients. Among ~ 10,000 lncRNAs, a total of ten lncRNAs were selected, including MEF2C-AS1, LINC00477, PROSER2-AS1, LINC01122, KCNJ2-AS1, LINC01108, AP001042.1, SOCS2-AS1, ST3GAL6-AS1, and LINC00355. Survival random forest achieved the best performance and significantly stratified patients into high- and low-risk groups (cross-validated HR = 2.49, 95% CI: 2.09–2.98; log-rank P < 0.001) in TARGET-AML bone marrow samples. The lncSigAML signature robustly demonstrates significant prognostic effects across all cross-cohort and cross-tissue datasets separately. The pooled hazard ratio between low- and high-risk groups across the datasets was 2.88 (95%CI = 1.79-4.63; P < 0.001). The lncSigAML score remained independently associated with overall survival after adjustment for age, sex, and race across all external test datasets individually (pooled HR = 2.63; 95%CI = 1.62-4.24; P < 0.001). Furthermore, lncSigAML retained its independent prognostic value after adjustment for established genetic alterations and cytogenetic risks.

We next performed five-fold cross-validation to identify the optimal machine learning algorithm for developing lncSigAML, which was subsequently validated across four independent datasets. Multivariable Cox proportional hazards regression demonstrated that lncSigAML provided prognostic value beyond established clinical variables, cytogenetic subtypes, and genetic risk classifications. Kaplan–Meier analyses showed that the pooled 1-, 3-, and 5-year overall survival rates for the high-risk group across external validation cohorts were 52.1% (95% CI, 22.0%–82.2%), 28.1% (95% CI, 0.0%–57.3%), and 23.9% (95% CI, 0.0%–52.2%), respectively. In contrast, the corresponding pooled survival rates for the low-risk group were 80.5% (95% CI, 62.9%–98.0%), 63.4% (95% CI, 31.9%–94.9%), and 55.3% (95% CI, 11.1%–99.6%), respectively.

Pooled pathway analyses further revealed that the lncSigAML high-risk group was enriched for STK33-upregulated genes associated with enhanced cell survival and proliferation (normalized enrichment score [NES] = 1.74; adjusted P < 0.001), the IL2-downregulated pathway indicative of impaired immune surveillance (NES = 1.75; adjusted P < 0.001), and the PTEN-downregulated pathway, a hallmark of aggressive AML biology (NES = 1.58; adjusted P < 0.001). To facilitate clinical and research applications, we implemented the final lncSigAML model as an interactive, publicly accessible web application using the R Shiny framework.

In summary, we developed lncSigAML, a robust machine learning–based prognostic model that accurately predicts survival in AML. By integrating pediatric and adult AML cohorts, we identified a panel of lncRNAs with consistent prognostic value and demonstrated that lncSigAML is generalizable across independent datasets, resilient to substantial feature-level missingness, and provides prognostic information beyond established clinical variables, cytogenetic risk, and genetic subtypes. To facilitate clinical translation, we implemented lncSigAML as an interactive, publicly accessible web application that generates individualized prognostic estimates. Future prospective validation and functional characterization of the identified lncRNAs will further support the integration of lncSigAML into precision medicine strategies for patients with AML.

Bon Q. Trinh, PhD: No relevant conflicts of interest to declare.