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

Paper Details

Abstracts, authors, affiliations, and disclosures.

Sunday, December 13, 2026
04:30 p.m. - 06:00 p.m.
MCCNO - 260-262
Technology and Procedures, Molecular Testing, Omics Technologies, Bioinformatics, Computational Biology, Treatment Considerations, Combination Therapy, Drug Development, Biological Therapies, Antibody Therapy, Immunotherapy, Artificial Intelligence (AI), Deep Learning, Machine Learning, Biological Processes, Genomics, Immune Mechanism, Metabolism, Molecular Biology, Multi-Systemic Interactions, Diseases, Lymphoid Malignancies, Lymphoid Leukemias, Lymphomas, Plasma Cell Disorders, Myeloid Malignancies, Acute Myeloid Malignancies, ALL, CLL, Hodgkin Lymphoma, Non-Hodgkin Lymphoma, B Cell Lymphoma, T Cell Lymphoma, Indolent Lymphoma, Aggressive Lymphoma, Research, Fundamental Science, AML, Translational Research, Clinical Research, Clinical Trials, Study Population, Animal Model
PAPER

657 A machine learning resource for lineage-specific cancer drug discovery reveals unique metabolic liabilities of B-cell malignancies

Mark E Robinson, PhD, Shalin Kothari, MD2, Qin Li, PhD3, Kadriye Nehir Cosgun, PhD4, Zhangliang Cheng, BSc5, Anna Mathew Lawrence6, Stephen Martindale7, Jodi Forward, MD8, Ruifeng Sun, BS9, Kohei Kume, PhD10, Daisuke Nakada, PhD11, Min Wu, PhD12, Dennise A De Jesús-Díaz, PhD13, Gregory Crimmins, PhD14, Jihoon Kim, PhD15, Felix Dietlein, MD, PhD16, Markus Müschen, MD-PhD17
1 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States
2 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States
3 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States
4 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States
5 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States, Department of Immunobiology, New Haven, Connecticut, United States
6 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States
7 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States
8 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States, Department of Pediatrics, Section of Hematology-Oncology, New Haven, Connecticut, United States
9 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States, Department of Immunobiology, New Haven, Connecticut, United States
10 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States
11 Molecular and Human Genetics Baylor College of Medicine Houston, Houston, Texas, United States
12 Remedy Plan Therapeutics, Gaithersburg, Maryland, United States
13 Remedy Plan Therapeutics, Gaithersburg, Maryland, United States
14 Remedy Plan Therapeutics, Gaithersburg, Maryland, United States
15 Department of Biomedical Informatics and Data Science, New Haven, Connecticut, United States
16 Computational Health Informatics Program, Boston Children’s Hospital, Boston, Massachusetts, United States
17 Center of Molecular and Cellular Oncology, New Haven, Connecticut, United States

Background and Significance. Across many cancer types, therapeutic vulnerabilities arise from lineage-specific developmental and transcriptional constraints. To systematically identify lineage-selective drug targets across 11 cancer cell lineages, we developed Delineate (Lymphoblasts.org/Delineate/), an interactive machine-learning resource integrating gene expression, protein abundance, metabolite levels, genetic dependencies, and compound-sensitivity profiles. Its network comprises 103,273 interactions involving 884 compounds, 509 drug targets, 7,534 genetic dependencies, 226 metabolites, and 772 pathways across 1,861 cancer cell lines. Because profound B-cell depletion is compatible with survival, multiple therapies have empirically targeted B cells as the root cause of B-cell malignancies and autoimmune diseases. As proof of concept, we applied Delineate to B-cell malignancies and validated previously unrecognized metabolic dependencies.

Results: Alongside established B-cell vulnerabilities, Delineate identified nuclear NAD+ salvage and its core enzymes, nicotinamide phosphoribosyltransferase (NAMPT) and nuclear nicotinamide nucleotide adenylyltransferase 1 (NMNAT1), as the highest-ranking B-cell dependencies. B-lymphoid cells were significantly more dependent on NAMPT and NMNAT1 than ten non-B-cell lineages, whereas the cytosolic and mitochondrial isoforms NMNAT2 and NMNAT3 were neither expressed nor essential. B-cell receptor engagement increased NAMPT protein abundance approximately 45-fold. Conditional deletion of Nampt or Nmnat1 in BCR::ABL1- or NRASG12D-transformed murine B-cell acute lymphoblastic leukemia (B-ALL) induced G0/G1 arrest, abrogated colony formation, and triggered apoptosis. Only wild-type or nucleus-targeted NAMPT rescued Nampt deletion, establishing a requirement for enzymatically active nuclear NAD+ salvage. B-cell-specific Nampt deletion in mice caused near-complete loss of B-cell development beyond the pre-B-cell receptor checkpoint. Acute deletion also depleted germinal-center B cells but spared existing plasma cells and antigen-specific antibody titers.

Although most lineages can produce NAD+ through parallel salvage and de novo biosynthetic routes, stable-isotope tracing revealed that B-lymphoid cells exclusively use salvage. B-ALL and lymphoma cells incorporated 2H4-nicotinamide into NAD+, NADH, and NADP but showed no detectable incorporation of 13C11-tryptophan through de novo synthesis, even without extracellular nicotinamide. In contrast, myeloid leukemia cells used both routes. B cells also suppressed nicotinamide N-methyltransferase and 1-methylnicotinamide production, preventing diversion of nicotinamide from NAD+ salvage.

This metabolic bottleneck was encoded by the B-cell identity program. C/EBPα-mediated myeloid reprogramming induced de novo pathway enzymes and mitigated the consequences of Nampt or Nmnat1 deletion, while dominant-negative PAX5 derepressed IDO1 and KYNU. Developmental and epigenomic analyses revealed two tiers of pathway repression: KYNU was silenced in pre-B cells and B-ALL but reactivated in mature B cells through a BCL6-associated program, whereas IDO1 remained repressed until plasmablast and plasma-cell differentiation, when it was induced by POU2F2. Metabolite-rescue experiments confirmed KYNU as the critical bottleneck in B-ALL and IDO1 as the remaining bottleneck in mature B-cell malignancies.

Accordingly, pharmacologic NAMPT inhibition or degradation selectively killed B-ALL, mantle cell lymphoma, Burkitt lymphoma, and germinal-center B-cell-like diffuse large B-cell lymphoma, while largely sparing plasmablast-derived activated B-cell-like diffuse large B-cell lymphoma and plasma-cell-derived multiple myeloma. NAMPT inhibitors were similarly active against treatment-naïve and relapsed, chemotherapy-refractory B-ALL patient-derived xenografts. A chemogenomic screen identified synthetic lethality between NAMPT inhibition and folate-mediated one-carbon metabolism, validated by synergy with methotrexate in refractory B-ALL and mantle cell lymphoma models. The hyperbolic NAMPT inhibitor RPT1F reduced disease burden and prolonged survival in both xenograft models.

Conclusion. Our study establishes Delineate as a machine-learning resource for systematic discovery of lineage-selective cancer dependencies and identifies nuclear NAD+ salvage as an actionable B-cell vulnerability rooted in the B-cell identity program and reversed during plasma-cell differentiation.

Mark E Robinson, PhD: No relevant conflicts of interest to declare.