Linzhi Xie,1,* Meng Gao,2,* Shiming Tan,1 Yi Zhou,1 Jing Liu,1 Liwen Wang,1 Xin Li1 

1Department of Hematology, Third Xiangya Hospital, Central South University, Changsha, Hunan, People’s Republic of China; 2Department of Blood Transfusion, Third Xiangya Hospital, Central South University, Changsha, Hunan, People’s Republic of China

*These authors contributed equally to this work

Correspondence: Liwen Wang, Department of Hematology, Third Xiangya Hospital, Central South University, Changsha, Hunan, 410013, People’s Republic of China, Email wanglevin2021@outlook.com Xin Li, Department of Hematology, Third Xiangya Hospital, Central South University, Changsha, Hunan, 410013, People’s Republic of China, Email xy3lixin@outlook.com

Background: Multiple myeloma (MM) is a haematological malignancy, driven by complex interactions between tumor and immune cells. Nevertheless, the overall pattern of immune cells and MM pathogenesis within the bone marrow tumor microenvironment (BM-TME) remains underexplored.
Methods and Results: Firstly, we performed Mendelian Randomization analysis for 731 immunocyte phenotypes and MM, identifying 21 immune traits significantly associated with increased MM risk (OR> 1, PFDR< 0.05). Flow cytometry analysis confirmed that the MFI of CD14 (p< 0.01) and HLA-DR (p< 0.05) on CD14+ monocytes was significantly elevated in early-stage MM. Secondly, we analyzed monocytes gene characteristics in the MM BM-TME via scRNA-seq, identifying 1,447 differentially expressed genes (moDEGs) (p< 0.05). Subsequently, based on 482 prognostic moDEGs, we developed and validated an optimal model, termed the Monocyte-related Gene Prognostic Signature (MGPS), by integrating 101 predictive models generated from 10 machine learning algorithms across multiple transcriptome sequencing datasets. MGPS was found to be an independent prognostic factor for MM (HR 2.72, 95% CI: 1.84– 4.0, p< 0.001), and the MGPS-based nomogram exhibits robust and reliable predictive performances. Next, MM patients with the low MGPS score exhibiting significantly better overall survival (OS) than the high MGPS score (p< 0.0001). Finally, we evaluated the predictive value of MGPS for treatment response and explored its molecular mechanisms. Results indicated that low-risk patients are more likely to benefit from immunotherapy, while a high MGPS score reflects cellular functional impairment.
Conclusion: Our findings reveal a complex interplay between immune cells and MM. Through multi-omics analyses and machine learning algorithms, we established a robust monocyte-related prognostic signature. By identifying high-risk patients, MGPS may help refine treatment strategies, such as intensifying immunomodulatory therapies, potentially improving survival and immunotherapy outcomes for MM patients.

Keywords: immunophenotype, multiple myeloma, machine learning, Mendelian randomization, monocyte, multi-omics