PMCID
string
Title
string
Sentences
string
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Consequently, DLBCL patients were classified into two subgroups, C1 (35 patients) and C2 (60 patients).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Further comparison of PCD-related DEGs revealed that 306 genes, including CTSB, CEBPB, CTSA, SIPPA, FTH1, TLR2, PLA2G7, GLA, ACP2, and SERPINA1, were significantly upregulated in C1 (P < 0.05), while 148 genes, such as PDK1, RPL8, HAX1, OGG1, HINT1, RPS7, NPRL2, and RPS3, were significantly upregulated in C2 (P < 0.05).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Principal component analysis (PCA) further validated that the expression of PCD-related DEGs effectively distinguished the two subgroups (Fig. 1D) (Supplementary File 3, S3).Fig.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
1Consensus Clustering Analysis of PCD-Related DEGs in DLBCL Patients.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A Cumulative Distribution Function (CDF) curve showing the evaluation of different k-values in consensus clustering analysis.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The slope of the CDF curve reaches its maximum at k = 2, indicating the optimal number of clusters.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
B Tracking plot displaying the stability of cluster assignments across different values of k. The consistent colors within each row at k = 2 suggest a stable clustering solution.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
C Heatmap illustrating the clear division of DLBCL samples into two subgroups (C1 and C2) based on PCD-related DEGs.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The distinct clustering patterns further confirm the effectiveness of the consensus clustering.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
D Principal Component Analysis (PCA) plot demonstrating the separation of the two subgroups, C1 and C2, based on the expression profiles of PCD-related DEGs.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
This separation confirms the validity of the clustering approach Consensus Clustering Analysis of PCD-Related DEGs in DLBCL Patients.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A Cumulative Distribution Function (CDF) curve showing the evaluation of different k-values in consensus clustering analysis.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The slope of the CDF curve reaches its maximum at k = 2, indicating the optimal number of clusters.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
B Tracking plot displaying the stability of cluster assignments across different values of k. The consistent colors within each row at k = 2 suggest a stable clustering solution.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
C Heatmap illustrating the clear division of DLBCL samples into two subgroups (C1 and C2) based on PCD-related DEGs.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The distinct clustering patterns further confirm the effectiveness of the consensus clustering.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
D Principal Component Analysis (PCA) plot demonstrating the separation of the two subgroups, C1 and C2, based on the expression profiles of PCD-related DEGs.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
This separation confirms the validity of the clustering approach The immune infiltration analysis revealed significant differences in the proportions of CD4+ T cells, regulatory T cells, naive B cells, memory B cells, activated NK cells, and M0 macrophages between the C1 and C2 subgroups (P < 0.05, Fig. 2A).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
GSVA enrichment analysis showed that upregulated pathways in the C2 subgroup were related to DNA repair, cell cycle, and energy metabolism, suggesting that this subgroup may have higher proliferative capacity and metabolic activity.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Conversely, downregulated pathways in the C2 subgroup were associated with immune response and lysosomal function, indicating the potential presence of immune evasion mechanisms and reduced intracellular degradation capacity (Fig. 2B).Fig.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
2PCD-related immune infiltration and GSVA enrichment analysis between DLBCL subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A Immune infiltration analysis comparing the proportions of various immune cell types between the C1 and C2 subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Significant differences were observed in the proportions of CD4+ T cells, regulatory T cells, naive B cells, memory B cells, activated NK cells, and M0 macrophages between the two subgroups (P < 0.05).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
B GSVA enrichment analysis illustrating the upregulated and downregulated pathways in the C2 subgroup compared to the C1 subgroup.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Upregulated pathways in the C2 subgroup include those related to DNA repair, cell cycle, and energy metabolism, indicating higher proliferative capacity and metabolic activity.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Downregulated pathways are associated with immune response and lysosomal function, suggesting potential immune evasion mechanisms and reduced intracellular degradation capacity in the C2 subgroup PCD-related immune infiltration and GSVA enrichment analysis between DLBCL subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A Immune infiltration analysis comparing the proportions of various immune cell types between the C1 and C2 subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Significant differences were observed in the proportions of CD4+ T cells, regulatory T cells, naive B cells, memory B cells, activated NK cells, and M0 macrophages between the two subgroups (P < 0.05).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
B GSVA enrichment analysis illustrating the upregulated and downregulated pathways in the C2 subgroup compared to the C1 subgroup.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Upregulated pathways in the C2 subgroup include those related to DNA repair, cell cycle, and energy metabolism, indicating higher proliferative capacity and metabolic activity.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Downregulated pathways are associated with immune response and lysosomal function, suggesting potential immune evasion mechanisms and reduced intracellular degradation capacity in the C2 subgroup WGCNA analysis was performed on the transcriptome expression data between DLBCL and healthy controls, as well as between the C1 and C2 subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The modules with the smallest Padjust values were selected (Padjust = 4e−124 for the DLBCL vs. healthy controls module; Padjust = 1e−10 for the C1 vs. C2 subgroup module) (Fig. 3A and 3B).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Module genes were filtered using the criteria "geneSigFilter = 0.5" and "moduleSigFilter = 0.8," resulting in 1266 and 26 module genes, respectively.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
By intersecting these results, seven characteristic genes of DLBCL were identified (Fig. 3C).Fig.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
3WGCNA Analysis of DLBCL and Healthy Controls, and Between C1 and C2 Subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A Module-trait relationships identified through WGCNA analysis between DLBCL patients and healthy controls.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The heatmap shows the correlation between module eigengenes (ME) and traits (control vs. treatment).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The module with the smallest Padjust value (Padjust = 4e−124) was selected for further analysis.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
B Module-trait relationships identified through WGCNA analysis between the C1 and C2 subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The heatmap displays the correlation between module eigengenes (ME) and traits (C1 vs. C2).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The module with the smallest Padjust value (Padjust = 1e−10) was selected for further analysis.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
C Venn diagram showing the intersection of module genes identified in WGCNA analyses between DLBCL and healthy controls, and between the C1 and C2 subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A total of seven characteristic genes of DLBCL were identified by intersecting the results from both analyses WGCNA Analysis of DLBCL and Healthy Controls, and Between C1 and C2 Subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A Module-trait relationships identified through WGCNA analysis between DLBCL patients and healthy controls.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The heatmap shows the correlation between module eigengenes (ME) and traits (control vs. treatment).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The module with the smallest Padjust value (Padjust = 4e−124) was selected for further analysis.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
B Module-trait relationships identified through WGCNA analysis between the C1 and C2 subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The heatmap displays the correlation between module eigengenes (ME) and traits (C1 vs. C2).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The module with the smallest Padjust value (Padjust = 1e−10) was selected for further analysis.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
C Venn diagram showing the intersection of module genes identified in WGCNA analyses between DLBCL and healthy controls, and between the C1 and C2 subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A total of seven characteristic genes of DLBCL were identified by intersecting the results from both analyses We integrated 12 machine learning algorithms to predict the expression patterns of programmed cell death (PCD)-related genes in DLBCL.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
To enhance model performance, we combined these algorithms into 91 different models.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
F-score evaluations showed that most combination models exhibited stable, high predictive performance on both training and test sets.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
AUC analysis revealed that most models achieved AUC values between 0.7 and 0.9, indicating strong classification ability.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Ensemble methods like XGBoost and GBM showed the highest AUC values, demonstrating superior performance in classifying DLBCL and PCD-related genes.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Feature importance analysis identified genes such as CTSB, DPYD, STOM, GBP1, SERPING1, C3AR1, and SCARB2 as critical for PCD-related expression patterns in DLBCL patients (Fig. 4).Fig.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
4Construction and performance evaluation of machine learning models for PCD-related gene expression in DLBCL.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A AUC (Area Under the Curve) values for various machine learning models applied to the training set (DatasetA) and two test sets (DatasetB and DatasetC).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The color gradient indicates the AUC values, with higher values suggesting better classification performance.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The models are listed on the y-axis, and the color-coding on the x-axis shows the respective datasets used for evaluation.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
B F-score evaluations for the same machine learning models across the training and test sets.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The F-score provides a balanced assessment by combining precision and recall.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The color gradient indicates F-score values, with higher values suggesting better overall performance.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The datasets used for evaluation are color-coded on the x-axis.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
C Feature importance rankings across various machine learning models.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The bar plot shows the average rank of each gene, with the color gradient indicating the number of models in which the gene was considered important.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Genes such as CTSB, DPYD, STOM, GBP1, SERPING1, C3AR1, and SCARB2 consistently ranked high, indicating their crucial roles in PCD-related gene expression patterns of DLBCL patients Construction and performance evaluation of machine learning models for PCD-related gene expression in DLBCL.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
A AUC (Area Under the Curve) values for various machine learning models applied to the training set (DatasetA) and two test sets (DatasetB and DatasetC).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The color gradient indicates the AUC values, with higher values suggesting better classification performance.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The models are listed on the y-axis, and the color-coding on the x-axis shows the respective datasets used for evaluation.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
B F-score evaluations for the same machine learning models across the training and test sets.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The F-score provides a balanced assessment by combining precision and recall.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The color gradient indicates F-score values, with higher values suggesting better overall performance.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The datasets used for evaluation are color-coded on the x-axis.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
C Feature importance rankings across various machine learning models.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The bar plot shows the average rank of each gene, with the color gradient indicating the number of models in which the gene was considered important.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Genes such as CTSB, DPYD, STOM, GBP1, SERPING1, C3AR1, and SCARB2 consistently ranked high, indicating their crucial roles in PCD-related gene expression patterns of DLBCL patients By comparing the differential gene expression between a DLBCL cell line (VAL) and a normal B lymphocyte cell line (IM-9), we identified a set of programmed cell death (PCD)-related genes and DLBCL-specific genes that were significantly upregulated or downregulated.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Specifically, 863 PCD-related genes were found to be differentially expressed (P < 0.05, Supplementary File 4, S4).
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Notably, the DLBCL-specific genes CTSB, DPYD, and SCARB2 were significantly upregulated, while STOM and GBP1 were significantly downregulated (Fig. 5).Fig.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
5Gene expression levels of selected PCD-related genes in DLBCL (VAL) and normal B lymphocyte (IM-9) Cell Lines Gene expression levels of selected PCD-related genes in DLBCL (VAL) and normal B lymphocyte (IM-9) Cell Lines Diffuse large B-cell lymphoma (DLBCL) is one of the most common hematological malignancies worldwide.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Despite advancements in immunotherapy, the prognosis for DLBCL patients remains poor due to the disease's high heterogeneity and complexity .
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Consequently, there is an urgent need to identify novel biomarkers and explore potential DLBCL subtypes to enhance risk assessment.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Programmed cell death (PCD) is critical for normal development and the maintenance of homeostasis in organisms .
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Increasing evidence suggests that PCD is closely linked to the initiation and metastasis of malignant tumors, and dysregulation of PCD pathways may offer valuable insights into prognosis and subtype classification in DLBCL [21–23].
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
This study employed a multi-layered and multi-method analytical approach to comprehensively explore the expression patterns of PCD-related genes in DLBCL patients and their potential roles in clinical prognosis.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Using consensus clustering analysis, we stratified DLBCL patients into two distinct subgroups (C1 and C2) and further analyzed the differences between these subgroups in the immune microenvironment and key biological pathways through immune infiltration and GSVA enrichment analyses.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Immune infiltration analysis revealed that the higher proportions of M0 macrophages, regulatory T cells, and memory B cells in the C2 subgroup suggest a more complex immune evasion mechanism, while the higher infiltration of CD4+ T cells and activated NK cells in the C1 subgroup indicates a stronger anti-tumor immune response.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
These findings highlight how the C2 subgroup may regulate the immune microenvironment to facilitate tumor progression and drug resistance.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
GSVA enrichment analysis further revealed upregulation of DNA repair, cell cycle, and energy metabolism pathways in the C2 subgroup, which are typically associated with rapid tumor cell proliferation and enhanced anti-apoptotic capacity.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Additionally, the downregulation of immune-related pathways, such as Toll-like receptor signaling, in the C2 subgroup suggests possible immune evasion, reinforcing the high-risk characteristics of this subgroup.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
These results align with the malignant features of the C2 subgroup, indicating the need for more aggressive clinical intervention strategies.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
WGCNA analysis identified key gene modules within the DLBCL subgroups, playing a central role in the molecular mechanisms of DLBCL.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
By screening differential gene modules between DLBCL and healthy controls, as well as between the C1 and C2 subgroups, we identified a set of characteristic genes that may be crucial in the development and progression of DLBCL.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
These genes exhibited significant differences between subgroups, providing valuable insights for further clinical research.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The construction and evaluation of machine learning models further deepened our understanding of the molecular characteristics of DLBCL subgroups.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
By integrating 12 machine learning algorithms, we developed high-performance predictive models for DLBCL.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Ensemble learning methods such as XGBoost and GBM demonstrated superior AUC values, underscoring their efficacy in handling complex gene expression data.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
Feature importance analysis revealed that genes such as SCARB2, GBP1, and STOM consistently ranked highly across multiple models, indicating their critical roles in the PCD-related mechanisms of DLBCL.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
To validate the expression characteristics of PCD-related genes and the seven DLBCL-specific genes in patient samples, we conducted transcriptome sequencing analysis on the DLBCL cell line VAL and the normal human B lymphocyte cell line IM-9.
PMC12003219
Predictive biomarkers and molecular subtypes in DLBCL: insights from PCD gene expression and machine learning
The results showed that the significant differential expression of PCD-related genes in the VAL cell line closely matched the molecular characteristics of the C2 subgroup, further supporting their potential as high-risk biomarkers.