Abstract 3250: Tumor-independent monitoring of minimal residual disease in multiple myeloma using cfDNA fragmentomics

Abstract Introduction: Monitoring minimal residual disease (MRD) is critical in multiple myeloma (MM) to predict outcomes and guide therapy. Traditional bone marrow (BM) aspirates for MRD detection are invasive and limited by sample quality. We therefore explored cell-free DNA (cfDNA) fragmentation as a less invasive alternative for MRD detection. Methods: We performed 30-40X whole-genome sequencing (WGS) on peripheral blood cfDNA from 45 MM patients, collecting baseline (n = 45) and follow-up samples (n = 98) from eight Canadian sites (M4 and IMMAGINE studies) and one U.S. site (SPORE study), plus 25 healthy controls. Samples were collected post-induction therapy (n = 13), 100 days post-autologous stem cell transplantation (ACST, n = 36), and during maintenance therapy (n = 49). Multiparameter flow cytometry (MFC) MRD testing was conducted on 77 samples. We evaluated insert size metrics and MM-specific chromatin accessibility (from Ordoñez et al., 2020) using Griffin (described in Doebley et al., 2022). Results: Baseline samples had a higher proportion of short fragments (20-150 bp) than follow-up samples (p = 0.0037) and healthy controls (p = 0.019). A fragment score (FS) based on weighted fragment distribution (per Vessies et al., 2022) was highest at baseline (mean = -0.384), followed by MRD-positive (mean = -0.495) and MRD-negative (mean = -0.550) samples. MRD-negative FS was lower than baseline samples (p < 0.001) but higher than healthy controls (mean = -0.639; p = 0.016). Baseline samples had lower coverage at MM-specific chromatin regions than MRD timepoints (p < 0.05; baseline mean = 0.982; MRD-positive = 0.989; MRD-negative = 0.994; healthy = 0.993), suggesting higher transcriptional activity pre-treatment. MRD-negative samples had higher coverage than MRD-positive cases (p < 0.01). Using logistic regression, we assessed the predictive performance of FS and coverage at MM-specific sites. FS alone achieved an AUC of 0.632 (sensitivity [SN] 39.4%, specificity [SP] 92.1%, accuracy [AC] 67.6%). Coverage yielded an AUC of 0.681 (SN 75.8%, SP 60.5%, AC 67.6%). Combining FS and coverage at MM sites improved performance (AUC = 0.734; SN 48.5%, SP 92.1%, AC 71.8%). We next evaluated whether adjusting the BM tumor cell percentage cutoff by MFC could improve performance. At 0.017% (1 in 5, 789 cells) for MRD positivity, the combined model’s accuracy reached 83% with an AUC of 0.757 (SN 17%, SP 97%). Conclusions: cfDNA fragmentomic analysis achieved high specificity (92-97%) in detecting MRD, providing a less invasive alternative to facilitate serial monitoring. While sensitivity remains an area for improvement, the high specificity of cfDNA makes it valuable for confirming MRD negativity and reducing invasive procedures. Future efforts aim to enhance sensitivity by integrating additional fragmentomic features. Citation Format: Dor D. Abelman, Jenna Eagles, Aimee Wong, Saumil Shah, Stephanie Pedersen, Stephenie Prokopec, David S. Scott, Sarah Bridges, Darrell White, Irwindeep Sandhu, Kevin Song, Esteban Braggio, Alli Murugesan, Anthony Reiman, Suzanne Trudel, Trevor J. Pugh. Tumor-independent monitoring of minimal residual disease in multiple myeloma using cfDNA fragmentomics [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3250.