Diffuse large B-cell lymphoma risk prognosis model based on lactic acid metabolism related genes

Abstract

The invention belongs to the technical field of prognosis models, and particularly relates to a diffuse large B-cell lymphoma risk prognosis model based on lactic acid metabolism related genes. The model construction comprises the following steps: S1, data acquisition and processing; s2, constructing a prognosis model; s3, verifying and evaluating the prognosis model; and S4, application of a prognosis model. GSE10846 and GSE87371 data sets are used for screening lactic acid metabolism related genes (LMRGs), a prognosis risk model is constructed and verified, and association of the prognosis risk model with clinical pathological characteristics, immune characteristics and treatment response is disclosed.

Description

Technical Field

[0001] The present invention belongs to the technical field of prognostic models, and particularly relates to a risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes. Background Art Diffuse large B-cell lymphoma (DLBCL), as the most common subtype of non-Hodgkin lymphoma, poses a clinical challenge of recurrence and refractoriness in 30-40% of patients. The existing International Prognostic Index (IPI) has limited predictive efficacy due to tumor heterogeneity, and further research is needed to optimize the stratification model with novel biomarkers to guide adverse outcomes.

[0003] Aerobic glycolysis, as a central metabolic hallmark of cancer cells, can generate an important metabolic fuel, lactate. Extracellular lactate participates in the tricarboxylic acid cycle to produce energy through the MCT1 transporter protein, and forms an acid concentration gradient in the tumor microenvironment (TME), inhibiting the immune activity of CD8+ T cells and promoting tumor growth. Therefore, treatment strategies targeting lactate and its related metabolic pathways are expected to slow down tumor progression by restoring immune efficacy and reducing the nutrient utilization rate of tumor cells. However, there is currently no evidence indicating the correlation between the gene set related to lactate metabolism and the prognosis of DLBCL. Summary of the Invention The present invention uses the GSE10846 and GSE87371 data sets to screen lactate metabolism-related genes (LMRGs), constructs and validates a prognostic risk model, and reveals its associations with clinicopathological features, immune features, and treatment responses.

[0005] To achieve the above technical objectives, the technical solution adopted by the present invention is: a risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes, and the method for establishing the model includes the following steps: S1: Data collection and processing: Log in to the website of the GEO database to obtain the gene expression profile data and clinical data of DLBCL patients in the GSE10846 and GSE87371 data sets, and perform standardization processing on the data; S2: Prognostic model construction: Based on the obtained data, perform univariate Cox regression analysis to screen LMRGs related to prognosis, and further use LASSO-Cox regression to construct a risk prognostic model containing 5 lactate metabolism-related genes (LMRGs; HIF-1A, MYC, NDUFB11, PITRM1, SDHA); the nucleotide sequences of the above 5 genes can all be obtained from the website (gene ID: 3091 (HIF-1A); 4609 (MYC); 54539 (NDUFB11); 10531 (PITRM1); 6389 (SDHA)); S3: Prognostic model validation and evaluation: Evaluate the predictive efficacy of the model through the training group (GSE10846; internal validation) and the validation group (GSE87371; external validation), and use the Kaplan-Meier curve, time-dependent ROC curve, and nomogram to verify the accuracy of the model efficacy; S4: Application of the prognostic model: Reveal the immune characteristics of the prognostic model by integrating the analysis results of CIBERSORT, ssGSEA, and ESTIMATE, and reveal the correlation between the prognostic model and treatment response through drug sensitivity and TIDE analysis.

[0006] Furthermore, the specific steps of S1 include: (1) Obtain the gene expression profiles and clinical data of diffuse large B-cell lymphoma (DLBCL) patients from GEO (GSE10846 and GSE87371); (2) After excluding samples with missing or unknown clinical survival information, 412 DLBCL patients in GSE10846 were set as the training group, and 221 patients in GSE87371 were used as the validation group.

[0007] Furthermore, the specific steps of S2 include: (1) Through literature retrieval and the Molecular Signatures Database (MSigDB, https: / / www.gsea-msigdb.org / gsea / msigdb), finally integrate 309 LMRGs. (2) Initially screen 98 prognosis-related LMRGs through univariate Cox regression analysis (threshold P < 0.05). (3) Further screen 5 key prognosis LMRGs (HIF-1A, MYC, NDUFB11, PITRM1, SDHA) by LASSO-COX analysis. (4) Multiply the RNA expression levels of the 5 LMRGs by the multivariate Cox regression coefficients respectively to construct a risk score formula (hazard score = (-0.6362086) × HIF-1A + 0.3394503 × MYC + (-0.3490697) × NDUFB11 + (-0.5430206) × PITRM1 + 0.6505132 × SDHA). Divide the patients into high-risk and low-risk groups according to the median risk score of the training group, and use the same threshold for the validation group.

[0008] Furthermore, the specific steps of S3 include: (1) Kaplan-Meier curves and log-rank tests were used to compare survival differences between high- and low-risk groups; (2) the area under the curve (AUC) was calculated using time-dependent ROC curves (timeROC R package) to evaluate the prognostic model efficacy; (3) in the training set, eight variables including age, gender, Eastern Cooperative Oncology Group performance status (ECOG PS) score, Ann Arbor stage, IPI score, number of extranodal involved lymph nodes, and prognostic risk score were integrated to construct nomograms for predicting 1-year, 3-year, and 5-year survival rates, and the calibration curves were used to compare the predicted survival rates with the actual survival rates to evaluate the model accuracy.

[0009] Furthermore, the specific steps of S4 include: (1) the CIBERSORT algorithm was used to evaluate the infiltration levels of 22 immune cells, ssGSEA was used to quantify the abundances of 28 immune cells, and the ESTIMATE method was used to calculate tumor purity, stromal / immune cell content, and immune microenvironment score; (2) the OncoPredict R package was used to predict the clinical drug responses of DLBCL patients. Using the GDSC V2 dataset (including gene expression matrix and IC50 data, downloaded from https: / / osf.io / c6tfx / ) as the training set and GSE10846 as the test set, the drug sensitivity (IC50) was predicted using the calcPhenotype function, and the TIDE algorithm was used to evaluate the potential response to immunotherapy.

[0010] The beneficial effects of the present invention are as follows: The prognostic risk model constructed based on lactate metabolism-related genes (LMRGs) screened from the GSE10846 and GSE87371 datasets of the present invention is not only significantly correlated with the clinicopathological characteristics of patients, but also exhibits excellent prognostic prediction efficacy. At the same time, through the risk scoring system, the model can indirectly reflect the heterogeneous characteristics of the tumor immune microenvironment of patients and effectively predict their therapeutic responses to specific chemotherapeutic drugs, thereby providing a scientific basis for formulating individualized treatment strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 is a flowchart of the steps of a method for establishing a risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes according to an embodiment of the present invention; Figure 2 is a schematic diagram of the process of a method for establishing a risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes according to an embodiment of the present invention; Figure 3 is a diagram of a risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes constructed in the training set In the figure: a Construct the LASSO regression model, b Show the radar chart of the multivariate regression coefficients of 5 LMRGs, c - e Gene expression heatmap, risk score distribution, and survival status of patients in the training group for 5 LMRGs, f Perform principal component analysis on DLBCL patients in the training group, g Survival analysis between the high - risk group and the low - risk group, h ROC curve to evaluate the predictive performance of the model at 1 year, 3 years, and 5 years, i and j Analyze the overall survival (OS) between the high - risk group and the low - risk group in DLBCL patients with ABC type (i) and GCB type (j), k Sankey diagram showing the association between the lactate metabolism - related gene signature clustering, risk signature, cell of origin (COO) status, and patient survival status, l Pie charts of the percentage of survival status / risk score / COO status in 2 lactate metabolism - related gene signature clusters; Figure 4 Figure for constructing a risk prognosis model of diffuse large B - cell lymphoma based on lactate metabolism - related genes in the validation group In the figure: a - c Gene expression heatmap, risk score distribution, and survival status in the validation group for 5 LMRGs, d Principal component analysis of patients in the validation group, e Log - rank curve between the high - risk group and the low - risk group, f ROC curve to evaluate the predictive performance of the model at 1 year, 3 years, and 5 years; Figure 5 Figure for the correlation analysis between the risk prognosis model based on LMRGs and clinicopathological features In the figure: a, b Correlation between the risk score and clinicopathological features (age and ECOG PS) in the training group, c - e Correlation between the risk score and clinicopathological features (age, stage, and IPI) in the validation group, f Perform univariate Cox regression analysis on age, gender, ECOG PS, stage, number of extranodal involved lymph nodes, lactate dehydrogenase (LDH) level, IPI, and risk score, g Perform multivariate Cox regression analysis on age, gender, ECOG PS, stage, number of extranodal involved lymph nodes, LDH level, IPI, and risk score, h Construct a nomogram of the risk score of clinicopathological features and the prognosis model, i Calibration curves of the nomogram at 1 year, 3 years, and 5 years; Figure 6 Figure for the immune microenvironment status between the low - risk group and the high - risk group in the training group a Survival curves between high-risk and low-risk groups in DLBCL patients treated with R-CHOP (rituximab combined with cyclophosphamide, doxorubicin, vincristine, and prednisone); b Evaluation of immune cell proportions using the CIBERSORT algorithm; c-e Analysis of immune characteristics of high-risk and low-risk groups using the CIBERSORT method (c), ssGSEA method (d), and ESTIMATE method (e). Figure 7 Differential analysis chart of the drug treatment response between the high-risk and low-risk groups a Analysis of the overall survival of high-risk and low-risk groups in DLBCL patients treated with CHOP; b-d TIDE analysis; e-o Drug sensitivity analysis of high-risk and low-risk groups. Detailed implementation manner

[0012] A risk prognosis model for diffuse large B-cell lymphoma based on lactate metabolism-related genes, and a method for establishing the model, comprising the following steps: S1: Data collection and processing: Log in to the website of the GEO database (https: / / www.ncbi.nlm.nih.gov / geo / ) to obtain the gene expression profile data and clinical data of DLBCL patients in the GSE10846 and GSE87371 datasets, and perform standardization processing on the data. S2: Prognosis model construction: Based on the obtained data, perform univariate Cox regression analysis to screen LMRGs related to prognosis, and further construct a risk prognosis model containing 5 lactate metabolism-related genes (LMRGs; HIF-1A, MYC, NDUFB11, PITRM1, SDHA) using LASSO-Cox regression; the nucleotide sequences of the above 5 genes can be obtained from the website https: / / www.ncbi.nlm.nih.gov / gene / (gene IDs: 3091 (HIF-1A); 4609 (MYC); 54539 (NDUFB11); 10531 (PITRM1); 6389 (SDHA)). S3: Prognosis model validation and evaluation: Evaluate the predictive efficacy of the model through the training group (GSE10846; internal validation) and the validation group (GSE87371; external validation), and use the Kaplan-Meier curve, time-dependent ROC curve, and nomogram to verify the accuracy of the model efficacy. S4: Application of the prognosis model: Reveal the immune characteristics of the prognosis model by integrating the analysis results of CIBERSORT, ssGSEA, and ESTIMATE, and reveal the correlation between the prognosis model and treatment response through drug sensitivity and TIDE analysis.

[0013] Furthermore, the specific steps of S1 include: (1) Obtaining the gene expression profiles and clinical data of diffuse large B-cell lymphoma (DLBCL) patients (GSE10846 and GSE87371) from GEO; (2) After excluding samples with missing or unknown clinical survival information, 412 DLBCL patients in GSE10846 were set as the training group, and 221 patients in GSE87371 were used as the validation group.

[0014] Furthermore, the specific steps of S2 include: (1) Through literature retrieval and the Molecular Signatures Database (MSigDB, https: / / www.gsea-msigdb.org / gsea / msigdb), 309 LMRGs were finally integrated; (2) 98 prognosis-related LMRGs (threshold P < 0.05) were preliminarily screened by univariate Cox regression analysis; (3) LASSO-COX analysis further screened 5 key prognosis LMRGs (HIF-1A, MYC, NDUFB11, PITRM1, SDHA); (4) The RNA expression levels of the 5 LMRGs were multiplied by the regression coefficients in the multivariate Cox to construct a risk score formula (hazard score = (-0.6362086) × HIF-1A + 0.3394503 × MYC + (-0.3490697) × NDUFB11 + (-0.5430206) × PITRM1 + 0.6505132 × SDHA). According to the median risk score of the training group, patients were divided into high-risk and low-risk groups, and the same threshold was used for the validation group.

[0015] Furthermore, the specific steps of S3 include: (1) Kaplan-Meier curves and log-rank tests were used to compare the survival differences between the high-risk and low-risk groups; (2) The time-dependent ROC curve (timeROC R package) was used to calculate the area under the curve (AUC) to evaluate the efficacy of the prognostic model; (3) In the training set, 8 variables including age, gender, Eastern Cooperative Oncology Group performance status (ECOG PS) score, Ann Arbor stage, IPI score, number of extranodal involved lymph nodes, and prognostic risk score were integrated to construct nomograms for predicting 1-year, 3-year, and 5-year survival rates, and the calibration curve was used to compare the predicted survival rate with the actual survival rate to evaluate the accuracy of the model.

[0016] Furthermore, the specific steps of S4 include: (1) Evaluating the infiltration levels of 22 immune cells using the CIBERSORT algorithm, quantifying the abundances of 28 immune cells by ssGSEA, and calculating tumor purity, stromal / immune cell content, and immune microenvironment scores using the ESTIMATE method; (2) Predicting the clinical drug responses of DLBCL patients using the OncoPredict R package. Taking the GDSC V2 dataset (including gene expression matrix and IC50 data, downloaded from https: / / osf.io / c6tfx / ) as the training set and GSE10846 as the test set, predicting drug sensitivity (IC50) through the calcPhenotype function, and evaluating the potential response to immunotherapy using the TIDE algorithm.

[0017] Figure 3 To construct a DLBCL risk prognosis model based on lactate metabolism-related genes in the training set a Construct a LASSO regression model. b Show a radar chart of the multivariate regression coefficients of 5 LMRGs. c - e Gene expression heatmaps, risk score distributions, and survival status of patients in the training group for 5 LMRGs. f Perform principal component analysis on DLBCL patients in the training group. g Survival analysis between the high-risk group and the low-risk group. h ROC curve to evaluate the prediction performance of the model at 1 year, 3 years, and 5 years. i and j Analyze the overall survival (OS) between the high-risk group and the low-risk group for ABC-type (i) and GCB-type (j) DLBCL patients. k A Sankey diagram showing the association between lactate metabolism-related gene signatures clustering, risk signatures, cell origin (COO) status, and patient survival status. l Pie charts of the percentages of survival status / risk score / COO status in 2 lactate metabolism-related gene signatures clusters. Result analysis: The 5-LMRGs risk model constructed in this study showed excellent prediction efficacy after multi-dimensional validation: ① The death risk of patients in the high-risk group was significantly increased, and the AUC values of 1 / 3 / 5-year survival prediction reached 0.722 - 0.690; ② Subgroup analysis results showed that regardless of the ABC / GCB subtype, the high-risk group had a worse prognostic outcome; ③ The Sankey diagram and pie charts further revealed the characteristics of this model: High-risk patients (mainly ABC subtypes in lactate metabolism-related gene signature cluster 1) had mostly poor prognostic outcomes. In contrast, low-risk patients (mainly GCB subtypes in lactate metabolism-related gene signature cluster 2) showed better survival outcomes.

[0018] Figure 4 To construct a diffuse large B-cell lymphoma risk prognosis model based on lactate metabolism-related genes in the validation group Heatmap of gene expression, risk score distribution, and survival status in the validation group for 5 LMRGs from a-c. d Principal component analysis of patients in the validation group. e Log-rank curves between the high-risk and low-risk groups, f ROC curves to evaluate the predictive performance of the model at 1 year, 3 years, and 5 years. Results analysis: Survival analysis showed that patients in the high-risk group had a worse prognosis in the validation group. In addition, this risk prognostic model had good predictive efficacy, and the AUC values for 1-year, 3-year, and 5-year survival predictions were 0.650, 0.705, and 0.732, respectively.

[0019] Figure 5 Correlation analysis diagram of the risk prognostic model based on LMRGs and clinicopathological features a, b Correlation between risk score and clinical features (age and ECOG PS) in the training group. c-e Correlation between risk score and clinical features (age, stage, and IPI) in the validation group. f Univariate Cox regression analysis of age, gender, ECOG PS, stage, number of extranodal involved lymph nodes, lactate dehydrogenase (LDH) level, IPI, and risk score. g Multivariate Cox regression analysis of age, gender, ECOG PS, stage, number of extranodal involved lymph nodes, LDH level, IPI, and risk score. h Construction of a nomogram of risk scores for clinical features and the prognostic model. i Calibration curves of the nomogram at 1 year, 3 years, and 5 years. Results analysis: We found that the risk score was correlated with higher age, higher ECOG PS, advanced disease stage, and higher IPI. The calibration curve results showed that the nomogram based on clinical features and prognostic risk scores had good consistency in predicting DLBCL outcomes and actual survival outcomes. These results indicate the strong feasibility and practicality of this prognostic model in clinical applications.

[0020] Figure 6 Immune microenvironment status map between the low-risk and high-risk groups in the training group Survival curves between the high-risk group and the low-risk group in DLBCL patients receiving R-CHOP (rituximab combined with cyclophosphamide, doxorubicin, vincristine, and prednisone) treatment. b Evaluation of immune cell proportions using the CIBERSORT algorithm. c-e Analysis of immune characteristics of the high-risk group and the low-risk group using the CIBERSORT method (c), the ssGSEA method (d), and the ESTIMATE method (e); Results analysis: In DLBCL patients receiving R-CHOP, the prognosis of patients in the high-risk group was worse than that of patients in the low-risk group. The results of the analysis of immune microenvironment characteristics showed that there were significant differences in the immune cell composition patterns between the high-risk group and the low-risk group. Therefore, the risk score of the prognostic model based on the present invention is beneficial for judging the possible tumor immune microenvironment status of patients.

[0021] Figure 7 Differential analysis chart of the high-risk group and the low-risk group in the drug treatment response a Analysis of the overall survival of the high-risk group and the low-risk group in DLBCL patients receiving CHOP treatment. b-d TIDE analysis. e-o Drug sensitivity analysis of the high-risk group and the low-risk group. Results analysis: In DLBCL patients receiving CHOP, the prognosis of patients in the high-risk group was worse. TIDE analysis suggested that high-risk group patients with higher TIDE scores might be more likely to benefit from immunotherapy than low-risk group patients. The results of the drug sensitivity experiment showed that cisplatin, cytarabine, bortezomib, ruxolitinib, temozolomide, vinblastine, and vinorelbine in the high-risk group (with lower IC50 values) showed more significant drug sensitivity than in the low-risk group. Therefore, the risk prognostic model based on LMRGs can guide clinical treatment selection.

Claims

1. A risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes, characterized in that: The method for establishing the model comprises the following steps: S1: Data collection and processing: Log in to the GEO database website to obtain the gene expression profile data and clinical data of DLBCL patients in the GSE10846 and GSE87371 datasets, and standardize the data; S2: Prognostic model construction: Based on the acquired data, univariate Cox regression analysis was performed to screen LMRGs related to prognosis, and LASSO-Cox regression was further used to construct a risk prognostic model containing 5 lactate metabolism-related genes (LMRGs; HIF-1A, MYC, NDUFB11, PITRM1, SDHA); the nucleotide sequences of the above 5 genes can be obtained from the website (gene ID: 3091 (HIF-1A); 4609 (MYC); 54539 (NDUFB11); 10531 (PITRM1); 6389 (SDHA)); S3: Validation and evaluation of the prognostic model: The predictive performance of the model was evaluated by the training group (GSE10846; internal validation) and the validation group (GSE87371; external validation), and the accuracy of the model performance was verified by using Kaplan-Meier curve, time-dependent ROC curve and nomogram; S4: Application of prognostic model: The immune characteristics of the prognostic model were revealed by integrating the results of CIBERSORT, ssGSEA and ESTIMATE analysis, and the drug sensitivity and TIDE analysis revealed the treatment response correlation of the prognostic model.

2. The risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes according to claim 1, characterized in that: The specific steps of S1 include: (1) obtaining the gene expression profiles and clinical data of diffuse large B-cell lymphoma (DLBCL) patients from GEO (GSE10846 and GSE87371); (2) after removing samples with missing or unknown clinical survival information, 412 DLBCL patients from GSE10846 were set as the training group, and 221 patients from GSE87371 were used as the validation group.

3. The risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes according to claim 1, characterized in that: The specific steps of S2 include: (1) Through literature retrieval and molecular signature database (MSigDB, https: / / www.gsea-msigdb.org / gsea / msigdb), 309 LMRGs were finally integrated; (2) 98 prognostic-related LMRGs were preliminarily screened by univariate Cox regression analysis (threshold P < 0.05); (3) LASSO-COX analysis further screened 5 key prognostic LMRGs (HIF-1A, MYC, NDUFB11, PITRM1, SDHA); (4) The RNA expression levels of the 5 LMRGs were multiplied by the multivariate Cox regression coefficient to construct the risk score formula (risk score = (-0.6362086) × HIF-1A +0.3394503 × MYC + (-0.3490697) × NDUFB11 + (-0.5430206) × PITRM1 + 0.6505132 × SDHA), the patients were divided into high-risk group and low-risk group according to the median risk score of the training group, and the same threshold was used in the validation group.

4. The risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes according to claim 1, characterized in that: The specific steps of S3 include: (1) Kaplan-Meier curve and log-rank test to compare the survival difference between high-risk and low-risk groups; (2) time-dependent ROC curve (timeROC R package) to calculate the area under the curve (AUC) to evaluate the effectiveness of the prognostic model; (3) in the training set, eight variables including age, gender, Eastern Cooperative Oncology Group performance status (ECOG PS) score, Ann Arbor stage, IPI score, number of extranodal involved lymph nodes and prognostic risk score were integrated to construct a nomogram for predicting 1-year, 3-year and 5-year survival rates, and the predicted survival rate was compared with the actual survival rate through the calibration curve to evaluate the accuracy of the model.

5. The risk prognostic model for diffuse large B-cell lymphoma based on lactate metabolism-related genes according to claim 1, characterized in that: The specific steps of S4 include: (1) using the CIBERSORT algorithm to evaluate the infiltration level of 22 immune cells, ssGSEA to quantify the abundance of 28 immune cells, and the ESTIMATE method to calculate tumor purity, matrix / immune cell content and immune microenvironment score; (2) using the OncoPredict R package to predict the clinical drug response of DLBCL patients, using the GDSC V2 dataset (including gene expression matrix and IC50 data, downloaded from https: / / osf.io / c6tfx / ) as the training set and GSE10846 as the test set, predicting drug sensitivity (IC50) through the calcPhenotype function, and evaluating the potential response to immunotherapy through the TIDE algorithm.

Images