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

Paper Details

Abstracts, authors, affiliations, and disclosures.

Monday, December 14, 2026
02:45 p.m. - 04:15 p.m.
MCCNO - 393-396
This session highlights the transition of AI from diagnostic innovation to clinical decision support and personalized patient management. Presentations demonstrate how AI can predict treatment-related toxicities and complications, identify cardiovascular and bleeding risks, enhance MRD assessment, automate diagnostic triage, and improve disease risk stratification. Together, these studies illustrate the emerging role of AI in anticipating clinical outcomes, guiding therapeutic decisions, and enabling more precise, individualized care in hematologic malignancies.
Technology and Procedures, Assays, Imaging, Molecular Testing, Pathology, Bioinformatics, Emerging Technologies, Treatment Considerations, Biological Therapies, (CAR)-T Cell Therapies, Adverse Events, Artificial Intelligence (AI), Computer Vision, Deep Learning, Machine Learning, Natural Language Processing, Diseases, Bleeding and Clotting, Thromboembolism, Thrombotic Disorders, Lymphoid Malignancies, Lymphoid Leukemias, Lymphomas, Plasma Cell Disorders, Measurable Residual Disease, ALL, Non-Hodgkin Lymphoma, B Cell Lymphoma, Aggressive Lymphoma, Research, Translational Research, Clinical Research, Real-World Evidence
PAPER

956 Ultra‑sensitive MRD assessment in DLBCL using phased variants in cfDNA: Automated workflow and first real‑world comparison with PET/CT

Johannes Duell, MD, Manuela Harloff, PhD2, Constance Baer, PhD3, Kathrin Hildebrand4, Stephan Hutter, PhD5, Wolfgang Kern, MD, PhD6, Florian Eisele7, Andreas Rosenwald, MD8, Prof. Dr. Max Topp, MD9, Torsten Haferlach, MD, PhD10, Manja Meggendorfer, PhD11
1 University Hospital of Wuerzburg, Wuerzburg, Germany
2 MLL Munich Leukemia Laboratory, Munich, Germany
3 MLL Munich Leukemia Laboratory, Munich, Germany
4 MLL Munich Leukemia Laboratory, Munich, Germany
5 MLL Munich Leukemia Laboratory, Munich, Germany
6 MLL Munich Leukemia Laboratory, Munich, Germany
7 University Hospital of Wuerzburg, Wuerzburg, Germany
8 University of Wuerzburg, Wuerzburg, Germany
9 University Hospital of Wuerzburg, Wuerzburg, Germany
10 MLL Munich Leukemia Laboratory, Munich, Germany
11 MLL Munich Leukemia Laboratory, Munich, Germany

Background
PET/CT is the current standard for staging and response assessment in diffuse large B-cell lymphoma (DLBCL), but has limited specificity, cannot assess measurable residual disease (MRD), and entails relevant radiation exposure. Circulating tumor DNA (ctDNA) provides a minimally invasive alternative for serial disease monitoring, and recent phased-variant–based assays have achieved detection limits down to tumor fractions of approximately 10?6.

Goal

To implement a minimally invasive, highly automated MRD approach with short turnaround times and high sensitivity.

Methods
We established a cfDNA-based MRD assay in DLBCL using a UMI-based, error-suppressed targeted sequencing panel (KAPA HyperCap DS NHL Panel; 383 genes, 341 kb) and a phased-variant (PV) approach, implemented on a fully automated liquid handling platform (AVENIO Edge, Roche, Basel, Switzerland) and NovaSeq X Plus sequencing (Illumina, San Diego, CA, USA). Baseline material consisted of plasma cfDNA, with FFPE or tissue punch tumor material and peripheral blood as germline control where available. Single-nucleotide variants (SNVs) were called at baseline at a variant allele frequency (VAF) ≥0.1% and UMI family size ≥3, followed by identification of doublet and triplet PVs within 100 bp and construction of patient-specific reporter variant lists. Follow-up samples were analyzed for known PVs without VAF cutoff and de novo PVs (VAF ≥0.1%) to assess MRD and clonal evolution, with MRD tumor fractions, limits of detection, and MRD p-values calculated using an in-house pipeline and AULE statistics. MRD results were compared with PET/CT at closely matched time points in a real-world cohort of first-line and relapsed patients treated with immunotherapies including CAR T-cell therapy and bispecific antibodies, with samples collected between March 2024 and May 2026.

Results
101 baseline cfDNA samples from a DLBCL cohort were analyzed; 74% (75/101) harbored PVs (median: 150 PVs per patient; range: 1–1,183), while 26% had no detectable PVs despite SNVs in a subset of genes. Among 50 patients with longitudinal cfDNA and PET/CT data (median follow-up of 9.1 months; range: 2.2–25.6), 13 (26%) developed clinically confirmed relapse; 11/13 showed clonal evolution with newly acquired PVs. In 8/13 cases where relapse was detected earlier by MRD than by PET/CT, MRD identified recurrence a median of 5.5 weeks earlier (range 1–17), although the representativeness of this finding is potentially limited by non-standardized data collection. Two patients never became MRD-negative despite PET/CT-defined remission, showing a tumor fraction of 0.0035% and 0.0151% at first complete remission. In a per-examination analysis of 140 paired assessments in 50 patients, ctDNA-based MRD and PET/CT showed a negative percent agreement of 77.6% (45/58), positive percent agreement of 59.8% (49/82), and overall agreement of 67.1%. Notably, MRD remained positive in a substantial fraction of PET/CT-negative time points and MRD stayed negative in cases of suspected false-positive PET/CT findings. The phased-variant assay demonstrated linear tumor fraction measurements and a limit of detection down to 1.8×10?6, with markedly reduced background error rates compared with CAPP-Seq-like single-SNV tracking (Cancer Personalized Profiling by deep Sequencing; VAF ≥0.1%). Molecular characterization of 45 cases by SNV analysis across 30 DLBCL-relevant genes revealed a median of three variants per sample (range: 1–7), VAFs spanning 0.1% to 50%, and at least one pathogenic or likely pathogenic variant in 91% of patients. Patients without PVs had fewer SNVs in all 383 panel genes investigated than patients with PVs at baseline (median: 7 vs. 203), suggesting that additional CAPP-Seq-style single-SNV tracking on cfDNA alone is unlikely to yield robust MRD markers, whereas sequencing of tumor DNA may add reporter PV candidates in this subgroup.

Conclusions
We implemented a highly automated cfDNA-based phased-variant assay for MRD monitoring in DLBCL that achieves detection limits down to 10?6 and integrates an in-house bioinformatics pipeline for PV tracking and clonal evolution analysis. Both wet-lab and bioinformatics automation reduced hands-on time to 10%. In a real-world cohort, ctDNA-MRD showed higher functional sensitivity than PET/CT and enabled earlier relapse detection, but its clinical utility and potential for MRD-guided intervention require prospective evaluation in controlled trials.
Manja Meggendorfer, PhD: MLL Munich Leukemia Laboratory - Current Employment