Abstract:Background: Previous researches have mainly focused on the discussion of whether cancer stem cells exist in diffuse large B-cell lymphoma (DLBCL). However, some subgroups with dismal prognosis that exhibit stem cell-like characteristics have been overlooked. Methods: Using an innovative one-class logistic regression (OCLR) machine learning algorithm, we calculated stemness indices and systematically assessed their ability to reflect the oncogenic dedifferentiation-like characteristics of DLBCL. Next, we identified signatures associated with oncogenic dedifferentiation-like features and prognosis using LASSO and SVM-RFE algorithms and developed a proportional hazards model (Riskscore) for DLBCL through Cox regression. Then we stratified the risk of DLBCL (n = 2133) and evaluated the prognostic value of the risk model across known clinical and molecular subgroups. We further compared the characteristics of high- and low-risk DLBCL with Burkitt lymphoma. Finally, we investigated the mechanisms of poor prognosis in high-risk DLBCL using transcriptomics, genomics (n = 576) and single-cell RNA sequencing (n = 19) data, as well as cell experiments and internal validation cohorts. Results: In this study, we identified and validated a DLBCL subgroup (25.6% of DLBCL) with stem cell-like characteristics and dismal prognosis. This high-risk group was defined as polyamine metabolism-cold immune tumor with upregulated polyamine metabolic key enzyme (ODC1) and desert-like immune infiltration, and had a poor prognosis with lower 3-year OS rate (54.3% vs. 83.6%, p < 0.0001) and PFS rate (42.8% vs. 75.2%, p < 0.0001) compared to the low-risk group. We found that some patients with MYC rearrangement, double-hit, double-expresser, or complete remission may have either favorable or poor prognosis, which can be accurately identified by our risk stratification model. Additionally, the high-risk group exhibited malignant proliferative phenotypes similar to Burkitt lymphoma. Genomic analysis revealed widespread copy number losses in the chemokine and interferon coding regions 8p23.1 and 9p21.3 in the high-risk group. Bulk and single-cell transcriptome analysis indicated that the upregulation of ODC1 might mediate the cold immune microenvironment of DLBCL, and knocking down ODC1 effectively inhibited DLBCL cell proliferation. Promisingly, our model effectively identified patients who were insensitive to immunotherapy (CAR-T and immune checkpoint antibodies). The research was funded by: This work was supported by grants from the National Natural Science Foundation of China (grant no. 81873450, 82170181), Beijing Hospitals Authority Youth Programme (code: QML20200201), and Beijing Natural Science Foundation (No. 7222027) to Liang Wang. Keywords: Bioinformatics, Computational and Systems Biology, Microenvironment, Risk Models No conflicts of interests pertinent to the abstract.

Latent class analysis‐derived classification improves the cancer‐specific death stratification of lymphomas: A large retrospective cohort study

Latent class analysis-derived classification improves the cancer-specific death stratification of molecular subtyping in colorectal cancer

Machine Learning‐Based Stem Cell‐Like Phenotype Identification and Novel Risk Stratification in Diffuse Large B‐Cell Lymphoma: Multi‐Omics Data from Multicenter Studies

Construction and analysis of a lysosome-dependent cell death score-based prediction model for non-small cell lung cancer

Causes of death and effect of non-cancer-specific death on rates of overall survival in adult classic Hodgkin lymphoma: a populated-based competing risk analysis

Integrative multi‐omic and machine learning approach for prognostic stratification and therapeutic targeting in lung squamous cell carcinoma

Longitudinal detection of subcategorized CD44v6 + CTCs and circulating tumor endothelial cells (CTECs) enables novel clinical stratification and improves prognostic prediction of small cell lung cancer: A prospective, multi-center study

Prognostic analysis of lung squamous cell carcinoma patients with second primary malignancies: a SEER database study

Single-cell and whole-transcriptome sequencing of lymph node metastasis-related gene signature: A large-scale pan-cancer cohort, machine learning and experimental study

Prognostic Factors and Survival in Primary Central Nervous System Lymphoma: A Population-Based Study

A prognostic model for overall survival of patients with early-stage non-small cell lung cancer: a multicentre, retrospective study

The clinical significance of a new classification algorithm in Chinese DLBCL cases

Epidemiologic Characteristics, Prognostic Factors, and Treatment Outcomes in Primary Central Nervous System Lymphoma: A SEER-Based Study

Distinct biological subtypes and patterns of genome evolution in lymphoma revealed by circulating tumor DNA

Identification and Validation of a Prognostic Prediction Model in Diffuse Large B-Cell Lymphoma

Subclassification of lung adenocarcinoma through comprehensive multi-omics data to benefit survival outcomes

44P A Large Population-Based Study on Large Cell Neuroendocrine Lung Cancer: A SEER Database Analysis

Prognostic factors and surgical approaches in the analysis of primary central nervous system diffuse large B-cell lymphoma: a large population-based cohort study and external validation

Comprehensive Analysis of Clinicalf Eatures, Treatment Options, Overall Survival, and Prognostic Factors in Lymphoma Cell Leukemia Patients: A Retrospective Study.

A Gene-Expression-based Signature Predicts Survival in Adults with T-cell Lymphoblastic Lymphoma: a Multicenter Study

Enhanced survival with lymphadenectomy in early-stage metachronous second primary lung cancer: A retrospective analysis