-Author name in bold denotes the presenting author
-Asterisk * with author name denotes a Non-ASH member
Clinically Relevant Abstract denotes an abstract that is clinically relevant.

PhD Trainee denotes that this is a recommended PHD Trainee Session.

Ticketed Session denotes that this is a ticketed session.

2231 Novel Liquid Biopsy Approach for Detection and Characterization of Disease Progression of Plasma Cell Disorders

Program: Oral and Poster Abstracts
Session: 803. Emerging Tools, Techniques, and Artificial Intelligence in Hematology: Poster I
Hematology Disease Topics & Pathways:
Bioinformatics, Computational biology, Technology and Procedures, Study Population, Human, Machine learning, Molecular testing, Omics technologies
Saturday, December 7, 2024, 5:30 PM-7:30 PM

Yuefan Huang, PhD1*, Riya Dilip Thakkar, MS1*, Katerina Kappes, BSc2*, Kristan Steffen, PhD1*, Oliver Venn, PhD1*, Rita Shaknovich, MD, PhD1, Qinwen Liu, PhD1* and Samir Parekh, MD2

1GRAIL, Inc., Menlo Park, CA
2Icahn School of Medicine at Mount Sinai, New York, NY

Introduction: Monoclonal Gammopathy of Undetermined Significance (MGUS) and Smoldering Multiple Myeloma (SMM) commonly affect the elderly and have prevalence of >3% in individuals aged ≥50 years. MGUS and SMM progress to Multiple Myeloma (MM) at a rate of about 1% and 10% per year, respectively. Current diagnostic approaches use invasive procedures, such as bone marrow biopsies and imaging modalities (PET/CT scans); moreover, current biomarkers do not accurately predict which patients would progress and who may benefit from earlier treatment. We demonstrate a liquid biopsy approach to non-invasively detect Plasma Cell Disorders (PCDs) using targeted methylation sequencing of cell-free DNA and reveal underlying methylation biology with diagnostic and prognostic significance.

Methods: The PCD classifier uses the same 2-stage architecture as a commercially available multi-cancer early detection test (Galleri®) and is set at 95% specificity. It first distinguishes PCD patients from non-cancer cases, then classifies cases into MGUS, SMM, or MM. The training data included 47 MGUS, 59 SMM, and 39 MM biobank samples, and 500 additional non-cancer samples from participants of the Circulating Cancer Genome Atlas Substudy 2 (CCGA2; NCT02889978). An independent holdout data set, consisting of 570 non-cancer and 32 MM samples from CCGA2, was used to evaluate the detection performance of the PCD classifier. In order to learn the biology of diseases and which genomic features differentiate among diagnostic entities, methylation features identified by classifier training were used to isolate regions relevant to subtype biology by pairwise comparison using log-odds ratios and Fisher's exact tests. Disease subtype relationships were explored by principal components analysis (PCA) and isometric feature mapping (ISOMAP). A progression trajectory was constructed using ISOMAP embeddings in PCA space. Methylation features significantly associated with disease progression were identified by a trajectory-based differential analysis using a negative binomial-generalized additive model.

Results: The PCD classifier (set at 95% specificity) showed high cross-validated sensitivities across all three subtypes: MGUS (93.6 [95% CI: 82.5-98.7]), SMM (91.5 [95% CI: 81.3-97.2]), and MM (100.0 [95% CI: 91.0-100.0]). Evaluation with the holdout set discriminated MM with high sensitivity (96.9 [95% CI: 83.8-99.9]) and specificity (93.2 [95% CI: 90.8-95.1]) as compared to non-cancer controls. Pairwise feature analysis highlighted the complexity of methylation changes across subtypes. Of the 1,423 significant features identified through pairwise comparisons from a total 1,993 raw features, 317 features showed progressively higher levels of methylation abnormality from MGUS to SMM and then to MM. The trajectory analysis highlighted the continuous spatial pattern of non-cancer, MGUS, SMM, to MM samples in the methylation feature space, indicating the gradual disease progression from precursors to malignancies. This finding was validated by the holdout MM and non-cancer samples. 1,046 abnormally methylated features significantly correlated with progression were identified by the trajectory-based differential analysis and found to be enriched in CpG islands and CpG shores. These abnormal features are differentially methylated at different stages of disease progression: early MGUS markers, SMM progression markers, and markers indicating entering the MM phase.

Conclusion: Our PCD classifier identifies DNA methylation signatures distinguishing precursor plasma cell neoplasms and improves our understanding of the disease progression continuum. It serves as a feasibility proof for developing a non-invasive, liquid biopsy approach with high sensitivity to detect early PCDs and potential to predict progression to a more aggressive clinical disease.

Disclosures: Huang: GRAIL, Inc.: Current Employment, Current equity holder in publicly-traded company. Dilip Thakkar: GRAIL, Inc.: Current Employment, Current equity holder in publicly-traded company. Kappes: GRAIL, Inc.: Research Funding. Steffen: GRAIL, Inc.: Current Employment, Current equity holder in publicly-traded company; Illumina, Inc.: Current equity holder in publicly-traded company. Venn: GRAIL, Inc.: Current Employment, Current equity holder in publicly-traded company. Shaknovich: GRAIL, Inc.: Current Employment, Current equity holder in publicly-traded company; Illumina, Inc: Current equity holder in publicly-traded company. Liu: GRAIL, Inc.: Current Employment, Current equity holder in publicly-traded company; Illumina, Inc: Current equity holder in publicly-traded company.

*signifies non-member of ASH