Abstract 4078: Single-cell sequencing of multiple myeloma patients with early death and long survival

Abstract Background: Multiple Myeloma (MM) is the second most common hematological malignancy in the world, characterized by diverse genomic landscape with clonal heterogeneity and complex interaction between malignant plasma cells and the immune microenvironment. Treatment advancement has improved MM survival remarkably over the past decade. However, despite being managed similarly, there are still patients who had early death while some having long survival. We aim to use single-cell sequencing to build a machine learning-based classifier that could predict patients’ clinical outcomes. Methods: Single-cell study evaluating the transcriptomic profile and protein expression, T-cell receptor and B-cell receptor sequencing were performed using droplet-based 10X Genomics 5’ version 2 platform with 60 DNA-barcoded antibodies. We evaluated a total of 26 frozen bone marrow samples from functional high-risk (FHR) group who had early death within 3 years of MM diagnosis and suboptimal response to induction therapy or early relapse within 12 months and long survival (LS) group who are alive for more than 7 years from MM diagnosis and still alive at the time or recruitment. This cohort also included longitudinal samples at presentation, post-treatment, and/or at relapse from 4 FHR and 3 LS patients. All patients were deemed fit to receive intensive treatment at diagnosis and received proteasome-inhibitor based induction therapy. Results: This is a preliminary result of a classifier from 11 patients. We trained the model on all cells from 3 newly diagnosed FHR samples and 5 newly diagnosed LS samples. We then tested our classifier on 3 samples that were unseen during training, which consisted of 2 relapsed FHR samples and 1 refractory LS samples. We showed in a histogram that the number of cells binned by the response predictor score correlated well with patients’ outcomes, in which samples with large numbers of poor scoring cells would respond poorly to treatment. We also evaluated the top 30 most important genes found by our classifier across all cell types in a heat map. For the next steps, we plan to include more samples to build our model. We also plan to run our classifier on selected cell types and identify key genes from these selected cell types, including plasma cells, T-cells, and monocytes. Longitudinal analysis of the samples from the same patient will also be performed to track clonal evolution with treatment and disease progression. Conclusion: We developed a machine learning-based classifier that allows us to identify MM patients with early death and long survival based on single-cell sequencing. Citation Format: Cinnie Y. Soekojo, Jonathan Adam Scolnick, Stacy Xu, Fangfang Song, Melissa Ooi, Sanjay de Mel, Wee Joo Chng. Single-cell sequencing of multiple myeloma patients with early death and long survival [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 4078.