A predictive model based on immune related genes for diffuse large B cell lymphoma (DLBCL)

Introduction

Background

Diffuse large B cell lymphoma (DLBCL) is a lethal and aggressive lymphoid malignancy (1,2). Based on the Cell-of-origin (COO) classification of DLBCL, there are three subtypes of DLBCL: (I) germinal center B-like DLBCL (GCB); (II) activated B-like DLBCL (ABC); and (III) unclassified in accordance with the gene expression profile of DLBCL (3-5). The cases exhibiting rearrangement of BCL2 and mutations in chromatin modifier genes such as EZH2, CREBBP, and KMT2D were strongly concerned with a GCB transcriptional subtype; the cases enriched with frequent BCL6 chromosomal translocations and activating mutation of NOTCH2 mostly fall into the category of ABC or transcriptionally unclassified ones; the cases that are strongly enriched for mutations of SGK1, TET2, SOCS1, and genes related to the JAK/STAT and ERK pathways are almost all GCB in origin; the cases characterized by widespread aneuploidy and biallelic inactivation of TP53 can be either of ABC origin or of GCB origin; NOTCH1-mutated cases are strongly relevant with an ABC transcriptional subtype (4,5). Each genetic type has a different prognosis (6).

The tumor microenvironment (TME) plays a vital role in the progression of malignancies and therapeutic strategies have been developed subsequently (5-10). In past decades, the role of TME in DLBCL was unclear and relatively neglected but is gaining increased attention with the advance of techniques (1,9). It has been reported that the cellular components of TME would impact the prognosis and treatment outcomes in DLBCL (10-15). For instance, the high infiltration of T-cells is significantly favorable for prognosis while the infiltration of PD-1⁺ T cells is considered to be correlated with reduced progression-free survival (PFS) and overall survival (OS) (12,16,17). Not only the cellular components, the products of immune cells have been reported to be prognostic factors (18-24). Interleukin-6 (IL-6; T cell-derived) and IL-10 (derived by helper T-cell and recently discovered by tumoral B-cell) are negatively associated with the prognosis of patients (19-23). Furthermore, evidences suggest that genetic factors, arising principally in lymphoma cells themselves, are among the most crucial (24). On the genic level, the mutations in cells such as immune cells, stroma cells or lymphoma cells themselves in DLBCL could, by regulating the comportment of cells in TME, influence the prognosis (24,25). For instance, glycolysis genes in DLBCL cells were positively correlated with malignancy degree; M2 macrophage-related genes could affect the outcome of DLBCL patients; serum IL-10 levels are also regulated by gene JAK2 (20,24-28).

Rationale and knowledge gap

Although R-CHOP treatment (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) has been considered as standard therapy for DLBCL, efforts should be made to screen novel second-line chemotherapy for those relapsed or refractory DLBCL (29). It is important to precisely diagnose DLBCL at an early stage or predict the relative risks of patients to facilitate precise therapy, thus sensitive and specific biomarkers and models are in urgent call.

Objective

The phenomenon explained above sheds light on the early diagnosis of DLBCL. Since the cellular components and products can indicate the prognosis of DLBCL, we hypothesized that immune-related genes, from transcription level, would bring new insight into the establishment of a novel diagnosis model of DLBCL.

Here, with regard of potential prognostic value of immune-related genes in DLBCL, we established a novel diagnostic model based on immune-related genes to generate a risk score. In accordance with our hypothesis, the risk score predicted advanced clinical stages and a shorter OS. Further, based on the risk score, we screened out four second-line chemotherapies for patients as a rescue. The risk score also correlated with remodeled immune landscape, in which the high risk indicated a more infiltrated suppressive immune cells and less anti-tumor immune cells infiltration. To ensure the accuracy and diagnostic ability of the risk model, we validated the performance of this model in external datasets and conducted reverse transcription and quantitative real-time polymerase chain reaction (RT-qPCR) to validate the expression of genes involved in the model in seven DLBCL cell lines. In summary, our model can be utilized in clinical practice to screen potential patients, predict the prognosis of DLBCL and guide the treatment of patients. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-2043/rc).

Methods

Data acquisition

The accession numbers of messenger RNA (mRNA) expression profiles are GSE10846 (420 samples) and GSE4475 (121 samples). The expression matrix was normalized via “scale” function in R to carry the further experiments and analysis.

‘Immunity’ was set as the keyword for the search of immune-related genes. A total of 1,621 immune-related genes were obtained from GeneCards (https://www.genecards.org/). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Establishment and validation of the risk prediction model

A random 50% of GSE10846 as training set was used to perform least absolute shrinkage and selection operator (LASSO) regression via R ‘glmnet’ package.

After 10-round cross-validation, we acquired the lambda with the minimum partial likelihood deviance to identify genes with a significant contribution to the model. With 10-round cross-validation to establish the risk model, a formula for calculating risk scores of each patient was established after being weighted by each estimated regression coefficient. The formula is as below:

Risk score = IL7R * (−0.210173566566956) + TOX * (−0.197754399969957) + TNFRSF9 * (−0.180584906393706) + RGS13 * (−0.0857138439699949) + LYZ * (−0.0679823188860567) + TIMP1 * (−0.0540310270946308) + TMEM123 * (−0.0130266588651968) + RAC2 * 0.0508157887674277 + CCL3 * 0.124306307293556)

To validate the diagnostic ability of the risk model, we tested the performance of risk model in internal validation set (the residual 50% of GSE10846) and an additional validation dataset (GSE4475). The time-dependent receiver-operating characteristic (ROC) was conducted via R ’survivalROC’ package to assess the accuracy of model predictions in 1, 3, and 5 years.

Establishment of nomogram

The nomogram was established to predict survival probability of DLBCL patients. Risk score, age, gender and stage were used to construct the nomogram using R “rms” and “survival” packages. For nomogram test, proportional hazard assumption was conducted through Schoenfeld residual calculation.

Estimation of a fraction of infiltrated immune cells

CIBERSORT (30) was utilized to evaluate the fraction of infiltrated immune cells.

Estimation of chemosensitivity

The Genomics of Drug Sensitivity in Cancer (GDSC) (https://www.cancerrxgene.org/) and R package ‘pRRophetic’ were used to predict the chemotherapeutic sensitivity and screen the potential second-line chemotherapy for patients in high- and low-risk. Ridge regression was utilized to estimate half maximal inhibitory concentration (IC₅₀) of each treatment with a specific chemotherapy drug. The prediction accuracy was determined through 10-fold cross-validation in the GDSC training set.

Cell culture

DLBCL cell lines OCI-Ly-1, OCI-Ly-3, Karpas-422, DB, SUDHL-4, SUDHL-8, and SUDHL-10 were obtained from Shanghai Institute of Hematology (Ruijin Hospital affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, China) and Department of Biochemistry and Molecular Cell Biology, Shanghai Jiao Tong University School of Medicine, Shanghai, China. Briefly, cells were cultured in Gibco/BRL RPMI-1640 medium (Gibco, New York, NY, USA), added with 10% fetal bovine serum (FBS) (Gibco, New York, NY, USA), under a 5% CO₂ humidified atmosphere at 37 °C. OCI-Ly-1 and OCI-Ly-3 cell lines were respectively cultured in Iscove’s Modified Dulbecco’s Medium (IMDM) (Gibco, New York, NY, USA) supplemented with 10% FBS.

Reverse transcription and quantitative real-time polymerase chain reaction (RT-qPCR)

Total RNAs were extracted from different cell line samples with TRIzol reagent (Invitrogen, Carlsbad, CA, USA). Complementary DNA (cDNA) template was then synthesized using a PrimeScript RT Kit (Takara Bio Inc., Beijing, China). RT-qPCR was conducted using a SYBR qPCR Master Mix (vazyme, Q711-02) on a Roche LightCycler^® 480 machine. All primer sequences were listed in Table 1.

Competing endogenous RNA (ceRNA) network

Firstly, the R package ‘multiMiR’ was utilized to predict possible microRNAs (miRNAs) (31). We input all nine mRNAs involved in our model, and only selected the mi-RNAs with convincing evidences whose degrees are greater than one. In order to obtain a miRNA centric network, the predicted 18 miRNAs were then uploaded to ‘miRNet’ (https://www.mirnet.ca/miRNet/) (32-36), where we only kept miRNAs existing in bone marrow, and filtered out long non-coding RNAs (lncRNAs), circular RNAs (circRNAs), and pseudogenes with degree not greater than one. Moreover, only the shortest paths were considered. Thus, we attained the interaction map between seven mRNAs, six key miRNAs, 43 circRNAs and two lncRNAs. Finally, we loaded all nodes and edges into the software ‘Cytoscape’ to visualize the result on an optimal basis. The green hexagons represent mRNAs; the blue rectangles, miRNAs; the purple ellipses, circRNAs; the yellow ellipses, lncRNAs.

Statistic and bioinformatics analysis

Bioinformatics and statistical analyses were conducted with R software (ver. 4.0.5) and GraphPad Prism software (ver. 9.0.0.121) in this study. Data were displayed as the mean and standard error of measurement (SEM). Univariate COX regression was performed to screen key immune-related genes by ’survival’ R package. A Kaplan-Meier method and log-rank comparison were performed for survival rate assessment. Pearson’s correlations were calculated using ‘Cor.test’ functions in R. Statistical tests in this study were all two-sided and P<0.05 was considered statistically significant.

Results

Identification of 41 key immune-related genes in DLBCL

A total of 1,621 immune-related genes were obtained but whether all of these genes play an essential role in the DLBCL and are correlated with the progression of DLBCL has yet to be determined (table available at https://cdn.amegroups.cn/static/public/tcr-24-2043-1.xlsx). Hence, we conducted a univariate Cox regression of 1,621 obtained to screen out the key immune related genes which may contribute to the prognosis in 420 patients from GSE10846. As a result turned out, 41 genes with P<0.05 were selected as key immune-related genes in DLBCL (Table S1). Among these genes, 13 genes with a hazard ratio >1 were considered to be hazardous, whereas those with a hazard ratio <1 were considered to be protective.

Establishment of an immune-related genes-based risk model via LASSO regression

After we screened out 41 key immune-related genes in DLBCL, we conducted LASSO regression to establish a risk model to predict the risk of DLBCL. First, a randomly 50% of patients from GSE10846 were set as the training dataset for the LASSO regression model and the residual 50% of GSE10846 was set as the internal testing dataset. Ten-round cross-validation was then conducted to determine the lambda with the minimum partial likelihood deviance (Figure 1A). Further, the contribution of each gene was calculated, and nine genes (IL7R, TOX, TNFRSF9, RGS13, LYZ, TIMP1, TMEM123, RAC2, CCL3) with the most significant contribution to the model were determined (Figure 1B). The formula of the risk model was introduced in the Materials and Methods section.

Figure 1 Establishment of an immune related genes-based risk model via LASSO regression. (A) Minimum lambda was selected via cross validation. (B) Selection of genes with significant contribution to the model and determination of corresponding coefficients. (C-E) Kaplan-Meier survival analysis of high- and low-risk groups in training, testing, and validation datasets. LASSO, least absolute shrinkage and selection operator.

In accordance with the coefficient of each gene, those genes with a positive coefficient (RAC2, CCL3) were hazardous and the other seven genes with a negative coefficient were protective.

Kaplan-Meier survival analysis was conducted to validate the performance of risk model (Figure 1C-1E). Risk score corresponding to each patient was generated by the risk score and subsequently divided patients into the high- and low-risk groups based on the median risk score. The analysis was conducted in the training dataset (50% of GSE10846), internal testing set (residual 50% of GSE10846), and additional external validation datasets (GSE4475). The results indicated that in all three datasets, patients with high-risk scores had a significantly lower median survival time compared with those with low-risk scores.

Validation of diagnostic ability of the risk model

To evaluate the diagnostic ability and accuracy of the risk model, time-dependent ROC analysis of the risk model was conducted. The area under the curve (AUC) represents the accuracy of the anticipation. The higher the AUC, the higher the diagnostic ability is. The AUC of the risk model in all three datasets for 1, 3, 5 years was all higher than 0.7, which indicated that our model performed accurately and had a robust diagnostic ability (Figure 2A-2C).

Figure 2 Validation of diagnostic ability of the risk model. (A-C) ROC analysis of the risk model in training, testing, and validation datasets. Curves and AUC values for 1, 3, 5 years were represented in different color. AUC, area under the curve; FPR, false positive rate; ROC, receiver operating characteristic; TPR, true positive rate.

Construction of a nomogram based on risk score and other common clinical indices

We found that the distribution of risk scores in patients of different ages and stages showed great variance. Patients over 60 years old showed a higher risk score in comparison with those patients under 60 years old, suggesting an altered and senescent immune response to malignancy (Figure 3A). Further, the patients with advanced stage showed a significantly higher risk score, indicating a dampened anti-tumor immunity in patients at advanced stages (Figure 3B).

Figure 3 Construction of nomogram based on risk score and other common clinical index. (A,B) The risk scores distribution is statically different in patients with different age and stages. (C) A nomogram was established based on age, gender, stage and risk scores. In gender, M stands for male, and F stands for female. The total points was the sum of points for the respective index.

To evaluate patients with an overall view, we combined common clinical indices, like gender, age and stage, with the risk score to construct a nomogram for clinical practice, which enables physicians to evaluate the survival probability for 3 and 5 years (Figure 3C).

Potential chemotherapy predicted by the risk score

In the hope of screening novel potential chemotherapy, chemotherapeutic sensitivity was predicted for the high- and the low-risk groups (Figure 4A-4D). Bexarotene, Bleomycin and Cisplatin were estimated to be sensitive for patients with high risk (Figure 4A-4C). The results suggested that patients in low risk would be sensitive to metformin (Figure 4D). Although these drugs were not in common use for the treatment of DLBCL, our data indicated that these drugs could be a potential rescue.

Figure 4 Potential chemotherapy predicted by the risk score. (A-D) Sensitivity of chemotherapy in high- and low-risk groups were evaluated based on estimated IC50, whose average value is compared through Wilcoxon test. IC50, half maximal inhibitory concentration.

The risk score revealed a remodeled immune landscape of DLBCL

To further explore the relationship between risk score and immune landscape of DLBCL, we conducted CIBERSORT analysis to estimate the infiltrating fraction of immune cells in DLBCL. As the result of CIBERSORT, the infiltrating fraction of immune cells showed significant difference between patients with high risk and low risk (Figure 5A). The correlation test further revealed that the risk score is positively correlated with immunosuppressive cells like M2 macrophages and Tregs and negatively correlated with T cells gamma delta and M0 macrophages (Figure 5B). The anti-tumor immunity was dampened in both group while more suppressive immune cells, like M2 macrophages and fewer anti-tumor immune cells, like T cells gamma delta infiltrated in patients with high-risk score, indicating a further suppressed anti-tumor immunity.

Figure 5 The risk score revealed a remodeled immune landscape of DLBCL. (A) Distribution of infiltration fraction of immune cells in patients with high-risk score and low risk score. (B) Correlation between infiltrated immune cells and risk score. *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. DLBCL, diffuse large B cell lymphoma; NK cells, natural killer cells.

RT-qPCR confirmed the expression of nine involved genes in real world

Beyond previous validation of the performance of the model in silico, we assessed the expression of the nine involved genes in seven cell lines via RT-qPCR. The results showed that all nine genes were expressed in all seven cell lines, which confirmed the expression of the nine involved genes in the real world (Figure 6).

Figure 6 RT-qPCR confirmed the expression of 9 involved genes in real world. (A-I) RT-qPCR validation of each genes in vitro cell lines. RT-qPCR, quantitative real-time polymerase chain reaction.

The ceRNA network reveals inner modulation of gene expression

ceRNA network is a regulatory mechanism in cellular biology where different RNA molecules interact with each other to modulate gene expression. The establishment of this ceRNA network enhances our comprehension of gene regulatory networks and may either play its role in prognosis prediction based on modulation or provide insights into novel therapeutic strategies targeting these interactions (Figure 7).

Figure 7 The ceRNA network based on the model-related genes. The green hexagons represent mRNAs; the blue rectangles, miRNAs; the purple ellipses, circRNAs; the yellow ellipses, lncRNAs. (A) The interaction between miRNAs and mRNAs. (B) The interaction between miRNAs with circRNAs and lncRNAs. ceRNA, competing endogenous RNA; circRNA, circular RNA; lncRNA, long non-coding RNA; miRNA, microRNA.

Discussion

Key findings

Currently, there is a pressing need for highly specific and sensitive biomarkers to facilitate early diagnosis, to evaluate the prognosis of patients, and to predict the therapeutic response of DLBCL, so as to help develop more precise therapeutic regimens for DLBCL. To address this gap, multi-omics approach of genomics, transcriptomics, epigenetics, proteomics, metabonomics, radiomics, and the currently developing single-cell technologies has come under the spotlight as a comprehensive and advanced method to identify potential biomarkers (37). Transcriptomics studies mainly concentrate on mRNAs, as well as non-coding RNAs, including miRNAs and lncRNAs, which play vital regulatory roles (37,38).

Beyond the cellular and protein level, here, we established a novel diagnostic model from transcription level, which is based on immune-related mRNAs as biomarkers to generate a risk score for finding out the potential patients, predicting the prognosis of DLBCL and guiding the treatment of patients (Figure 8). Furthermore, to help understand the gene regulation and discover more biomarkers, we constructed a ceRNA network, which opens up to our research the dimension of ceRNAs, including miRNAs, lncRNAs, and circRNAs.

Figure 8 The workflow map. Created in BioRender. Ma, J. (2025) https://BioRender.com/m09c026. DLBCL, diffuse large B cell lymphoma; LASSO, least absolute shrinkage and selection operator.

Strengths and limitations of our research

LASSO regression is a well-established method and has been proven to be suitable and superior to other commonly used methods (e.g., subset selection and ridge regression) in previous studies by others (39-44). Given its stability, robust performance and interpretability, LASSO regression was chosen to carry the further experiments. In our study, we screened out nine immune-related genes (IL7R, TOX, RGS13, LYZ, TIMP1, TMEM123, RAC2, CCL3) to construct the risk model via LASSO regression. Among these nine genes, TOX, thymocyte selection-associated HMG BOX, is a transcription factor that has been reported to be involved in the differentiation of CD4⁺ CD8⁺ double-positive thymocytes and the exhaustion of CD8⁺ T cells (45-48). This gene has previously been considered a biomarker and immunotherapeutic target for hematological malignancies (45). Intriguingly, TOX is also the essential factor for the mature of natural killer cells (NK cells) which have potent anti-tumor function (49-52). In this study, our data supported that TOX may serve as a protective factor in DLBCL, suggesting a potential augmentation of anti-tumor immunity via promoting the mature of NK cells. Further, our data suggested that CCL3 is a hazardous factor in DLBCL, which is in accordance with a previous report (53).

After the establishment of the risk model, we validated the diagnostic ability of the risk model. The risk model exhibited a robust performance in the prediction. The validation in silico ensured the accuracy of prediction, and the results of RT-qPCR confirmed the expression of involved genes in the real world. Further, in order to comprehensively evaluate the survival probability and extend the usage of the model in clinical practice, we constructed a nomogram which combined the risk score with other common clinical indices (gender, age and stage). Such a nomogram would facilitate the diagnosis of DLBCL.

Nevertheless, in adopting the conclusions of this study, several limitations also need to be considered. First of all, though we implemented RT-qPCR to validate the expression of mRNAs involved in our model, more solid evidences in vitro or in vivo are required. Moreover, the relationship between the nine immune-related genes included in the gene prediction model and the biological mechanism of TME have not been thoroughly studied. Finally, our predictive model only focuses on gene expression at mRNA level, whose explanatory power is limited due to the complex structure of TME. To overcome the challenge, further analysis aiming at single-cell RNA transcriptomics data and spatial transcriptomics data is needed.

Explanations of findings

DLBCL is an aggressive malignancy of lymphoid origin with a lethal nature (1,2). The role of TME has been gradually recognized in recent years (7-9,11,17). Both the cellular components and products from immune cells in TME have been demonstrated to be of prognostic value (12,16,19,22). These findings shed light on the early diagnosis of DLBCL.

Implications and actions needed

Along with the prognostic value of risk score, the risk score also predicted four novel second-line chemotherapies (bexarotene, bleomycin, cisplatin and metformin). Among these four drugs, metformin is a classic treatment of type 2 diabetes mellitus but has recently been reported to serve as an anti-tumor therapy in DLBCL (54-56). Our data suggested that patients in low risk would be sensitive to metformin, indicating that use of metformin treatment at the early stage of DLBCL may facilitate improving the prognosis.

Regarding the reshape of TME in DLBCL, the risk score revealed an immunosuppressive shift. Although the anti-tumor immunity was dampened in DLBCL, patients with high-risk score had more suppressive immune cells infiltrated but less infiltration of anti-tumor immune cells, in comparison with those patients with low-risk scores. Such a finding explained the reason why patients with high risk would have an unfavorable prognosis.

Conclusions

We established an immune related genes-based risk model to predict the survival and progression of DLBCL. We conducted LASSO regression to select those genes with significant contribution to DLBCL and established a risk model to generate a risk score. Validation of this risk model in the internal test dataset and additional external validation datasets confirms the robust performance of this model. To guide the treatment, we also found that four novel second-line chemotherapies can be used to treat patients with different risk scores. Overall, this novel model can be utilized in clinical practices for guiding the treatment.

Acknowledgments

We would like to thank Mr. Junrui Ma from Shanghai Jiao Tong University School of Medicine Department of Medical Laboratory Science for creating the workflow map (Figure 8) by BioRender.com for us.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-2043/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-2043/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-2043/prf

Funding: This work was supported by National Natural Science Foundation of China (Nos. 81600151, 81302039 and 82060040).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-2043/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Takahara T, Nakamura S, Tsuzuki T, et al. The Immunology of DLBCL. Cancers (Basel) 2023;15:835. [Crossref] [PubMed]
2. Campo E, Jaffe ES, Cook JR, et al. The International Consensus Classification of Mature Lymphoid Neoplasms: a report from the Clinical Advisory Committee. Blood 2022;140:1229-53. [Crossref] [PubMed]
3. Alizadeh AA, Eisen MB, Davis RE, et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000;403:503-11. [Crossref] [PubMed]
4. Bento L, Glass B, Schmitz N. Large B-Cell Lymphoma. 2024 Apr 11. In: Sureda A, Corbacioglu S, Greco R, et al. editors. The EBMT Handbook: Hematopoietic Cell Transplantation and Cellular Therapies. 8th ed. Cham (CH): Springer; 2024:Chapter 86.
5. Alig SK, Chapuy B, Ennishi D, et al. Evolving molecular classification of aggressive B-cell lymphoma. Histopathology 2025;86:94-105. [Crossref] [PubMed]
6. Dabrowska-Iwanicka A, Nowakowski GS. DLBCL: who is high risk and how should treatment be optimized? Blood 2024;144:2573-82. [Crossref] [PubMed]
7. Jin MZ, Jin WL. The updated landscape of tumor microenvironment and drug repurposing. Signal Transduct Target Ther 2020;5:166. [Crossref] [PubMed]
8. Bejarano L, Jordāo MJC, Joyce JA. Therapeutic Targeting of the Tumor Microenvironment. Cancer Discov 2021;11:933-59. [Crossref] [PubMed]
9. Scott DW, Gascoyne RD. The tumour microenvironment in B cell lymphomas. Nat Rev Cancer 2014;14:517-34. [Crossref] [PubMed]
10. Wang L, Yuan W, Li L, et al. Immunogenomic-Based Analysis of Hierarchical Clustering of Diffuse Large Cell Lymphoma. J Immunol Res 2022;2022:9544827. [Crossref] [PubMed]
11. Coutinho R, Clear AJ, Mazzola E, et al. Revisiting the immune microenvironment of diffuse large B-cell lymphoma using a tissue microarray and immunohistochemistry: robust semi-automated analysis reveals CD3 and FoxP3 as potential predictors of response to R-CHOP. Haematologica 2015;100:363-9. [Crossref] [PubMed]
12. Monti S, Savage KJ, Kutok JL, et al. Molecular profiling of diffuse large B-cell lymphoma identifies robust subtypes including one characterized by host inflammatory response. Blood 2005;105:1851-61. [Crossref] [PubMed]
13. Chapuy B, Stewart C, Dunford AJ, et al. Molecular subtypes of diffuse large B cell lymphoma are associated with distinct pathogenic mechanisms and outcomes. Nat Med 2018;24:679-90. [Crossref] [PubMed]
14. Fangazio M, Ladewig E, Gomez K, et al. Genetic mechanisms of HLA-I loss and immune escape in diffuse large B cell lymphoma. Proc Natl Acad Sci U S A 2021;118:e2104504118. [Crossref] [PubMed]
15. Challa-Malladi M, Lieu YK, Califano O, et al. Combined genetic inactivation of β2-Microglobulin and CD58 reveals frequent escape from immune recognition in diffuse large B cell lymphoma. Cancer Cell 2011;20:728-40. [Crossref] [PubMed]
16. Li L, Sun R, Miao Y, et al. PD-1/PD-L1 expression and interaction by automated quantitative immunofluorescent analysis show adverse prognostic impact in patients with diffuse large B-cell lymphoma having T-cell infiltration: a study from the International DLBCL Consortium Program. Mod Pathol 2019;32:741-54. [Crossref] [PubMed]
17. Roider T, Seufert J, Uvarovskii A, et al. Dissecting intratumour heterogeneity of nodal B-cell lymphomas at the transcriptional, genetic and drug-response levels. Nat Cell Biol 2020;22:896-906. [Crossref] [PubMed]
18. de Jong D, Enblad G. Inflammatory cells and immune microenvironment in malignant lymphoma. J Intern Med 2008;264:528-36. [Crossref] [PubMed]
19. Blay JY, Burdin N, Rousset F, et al. Serum interleukin-10 in non-Hodgkin's lymphoma: a prognostic factor. Blood 1993;82:2169-74. [Crossref] [PubMed]
20. Gupta M, Han JJ, Stenson M, et al. Elevated serum IL-10 levels in diffuse large B-cell lymphoma: a mechanism of aberrant JAK2 activation. Blood 2012;119:2844-53. [Crossref] [PubMed]
21. Garcia-Lacarte M, Grijalba SC, Melchor Sánchez J, et al. IL-10 from tumoral B cells modulates the diffuse large B-cell lymphoma microenvironment and response to immunotherapy. Blood 2025;blood.2024025755.
22. Giachelia M, Voso MT, Tisi MC, et al. Interleukin-6 plasma levels are modulated by a polymorphism in the NF-κB1 gene and are associated with outcome following rituximab-combined chemotherapy in diffuse large B-cell non-Hodgkin lymphoma. Leuk Lymphoma 2012;53:411-6. [Crossref] [PubMed]
23. Rousset F, Garcia E, Defrance T, et al. Interleukin 10 is a potent growth and differentiation factor for activated human B lymphocytes. Proc Natl Acad Sci U S A 1992;89:1890-3. [Crossref] [PubMed]
24. Cerchietti L. Genetic mechanisms underlying tumor microenvironment composition and function in diffuse large B-cell lymphoma. Blood 2024;143:1101-11. [Crossref] [PubMed]
25. Steen CB, Luca BA, Esfahani MS, et al. The landscape of tumor cell states and ecosystems in diffuse large B cell lymphoma. Cancer Cell 2021;39:1422-1437.e10. [Crossref] [PubMed]
26. Dai L, Fan G, Xie T, et al. Single-cell and spatial transcriptomics reveal a high glycolysis B cell and tumor-associated macrophages cluster correlated with poor prognosis and exhausted immune microenvironment in diffuse large B-cell lymphoma. Biomark Res 2024;12:58. [Crossref] [PubMed]
27. Wang X, Liu H, Fei Y, et al. Metabolic pathway-based subtyping reveals distinct microenvironmental states associated with diffuse large B-cell lymphoma outcomes. Hematol Oncol 2024;42:e3279. [Crossref] [PubMed]
28. Guo B, Duan Y, Cen H. Prognostic model based on M2 macrophage-related signatures for predicting outcomes, enhancing risk stratification, and providing therapeutic insights in diffuse large B-cell lymphoma. Heliyon 2024;10:e41007. [Crossref] [PubMed]
29. Sehn LH, Salles G. Diffuse Large B-Cell Lymphoma. N Engl J Med 2021;384:842-58. [Crossref] [PubMed]
30. Newman AM, Liu CL, Green MR, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 2015;12:453-7. [Crossref] [PubMed]
31. Ru Y, Kechris KJ, Tabakoff B, et al. The multiMiR R package and database: integration of microRNA-target interactions along with their disease and drug associations. Nucleic Acids Res 2014;42:e133. [Crossref] [PubMed]
32. Chang L, Xia J. MicroRNA Regulatory Network Analysis Using miRNet 2.0. Methods Mol Biol 2023;2594:185-204. [Crossref] [PubMed]
33. Chang L, Zhou G, Soufan O, et al. miRNet 2.0: network-based visual analytics for miRNA functional analysis and systems biology. Nucleic Acids Res 2020;48:W244. [Crossref] [PubMed]
34. Fan Y, Habib M, Xia J. Xeno-miRNet: a comprehensive database and analytics platform to explore xeno-miRNAs and their potential targets. PeerJ 2018;6:e5650. [Crossref] [PubMed]
35. Fan Y, Xia J. miRNet-Functional Analysis and Visual Exploration of miRNA-Target Interactions in a Network Context. Methods Mol Biol 2018;1819:215-33. [Crossref] [PubMed]
36. Fan Y, Siklenka K, Arora SK, et al. miRNet - dissecting miRNA-target interactions and functional associations through network-based visual analysis. Nucleic Acids Res 2016;44:W135-41. [Crossref] [PubMed]
37. Liang XJ, Song XY, Wu JL, et al. Advances in Multi-Omics Study of Prognostic Biomarkers of Diffuse Large B-Cell Lymphoma. Int J Biol Sci 2022;18:1313-27. [Crossref] [PubMed]
38. Khanmohammadi S, Masrour M, Fallahtafti P, et al. MicroRNA as a Potential Diagnostic and Prognostic Biomarker in Diffuse Large B-Cell Lymphoma: A Systematic Review and Meta-Analysis. Cancer Rep (Hoboken) 2025;8:e70070. [Crossref] [PubMed]
39. Sun S, Guo W, Wang Z, et al. Development and validation of an immune-related prognostic signature in lung adenocarcinoma. Cancer Med 2020;9:5960-75. [Crossref] [PubMed]
40. Tang G, Qi L, Sun Z, et al. Evaluation and analysis of incidence and risk factors of lower extremity venous thrombosis after urologic surgeries: A prospective two-center cohort study using LASSO-logistic regression. Int J Surg 2021;89:105948. [Crossref] [PubMed]
41. Guo P, Zeng F, Hu X, et al. Improved Variable Selection Algorithm Using a LASSO-Type Penalty, with an Application to Assessing Hepatitis B Infection Relevant Factors in Community Residents. PLoS One 2015;10:e0134151. [Crossref] [PubMed]
42. Muthukrishnan R, Rohini R. LASSO: A feature selection technique in predictive modeling for machine learning. In: 2016 IEEE International Conference on Advances in Computer Applications (ICACA). Coimbatore, India: 2016;18-20. doi: 10.1109/ICACA.2016.7887916.10.1109/ICACA.2016.7887916
43. Wang Q, Qiao W, Zhang H, et al. Nomogram established on account of Lasso-Cox regression for predicting recurrence in patients with early-stage hepatocellular carcinoma. Front Immunol 2022;13:1019638. [Crossref] [PubMed]
44. Tibshirani R. Regression Shrinkage and Selection Via the Lasso. Journal of the Royal Statistical Society: Series B (Methodological) 1996;58:267-88. [Crossref]
45. Liang C, Huang S, Zhao Y, et al. TOX as a potential target for immunotherapy in lymphocytic malignancies. Biomark Res 2021;9:20. [Crossref] [PubMed]
46. Aliahmad P, Seksenyan A, Kaye J. The many roles of TOX in the immune system. Curr Opin Immunol 2012;24:173-7. [Crossref] [PubMed]
47. Cheng Y, Shao Z, Chen L, et al. Role, function and regulation of the thymocyte selection-associated high mobility group box protein in CD8(+) T cell exhaustion. Immunol Lett 2021;229:1-7. [Crossref] [PubMed]
48. Sekine T, Perez-Potti A, Nguyen S, et al. TOX is expressed by exhausted and polyfunctional human effector memory CD8(+) T cells. Sci Immunol 2020;5:eaba7918. [Crossref] [PubMed]
49. Aliahmad P, de la Torre B, Kaye J. Shared dependence on the DNA-binding factor TOX for the development of lymphoid tissue-inducer cell and NK cell lineages. Nat Immunol 2010;11:945-52. [Crossref] [PubMed]
50. Takeda K, Smyth MJ, Cretney E, et al. Critical role for tumor necrosis factor-related apoptosis-inducing ligand in immune surveillance against tumor development. J Exp Med 2002;195:161-9. [Crossref] [PubMed]
51. Orange JS. Natural killer cell deficiency. J Allergy Clin Immunol 2013;132:515-25. [Crossref] [PubMed]
52. Lorenzi L, Tabellini G, Vermi W, et al. Occurrence of nodular lymphocyte-predominant hodgkin lymphoma in hermansky-pudlak type 2 syndrome is associated to natural killer and natural killer T cell defects. PLoS One 2013;8:e80131. [Crossref] [PubMed]
53. Takahashi K, Sivina M, Hoellenriegel J, et al. CCL3 and CCL4 are biomarkers for B cell receptor pathway activation and prognostic serum markers in diffuse large B cell lymphoma. Br J Haematol 2015;171:726-35. [Crossref] [PubMed]
54. Ikhlas S, Ahmad M. Metformin: Insights into its anticancer potential with special reference to AMPK dependent and independent pathways. Life Sci 2017;185:53-62. [Crossref] [PubMed]
55. Wang Y, Maurer MJ, Larson MC, et al. Impact of metformin use on the outcomes of newly diagnosed diffuse large B-cell lymphoma and follicular lymphoma. Br J Haematol 2019;186:820-8. [Crossref] [PubMed]
56. Singh AR, Gu JJ, Zhang Q, et al. Metformin sensitizes therapeutic agents and improves outcome in pre-clinical and clinical diffuse large B-cell lymphoma. Cancer Metab 2020;8:10. [Crossref] [PubMed]