Construction Of A Predictive Risk Score Model Based On Autophagy-Related Genes In Diffuse Large B-Cell Lymphoma

Objective Diffuse large B-cell lymphoma(DLBCL)is considered to be one of the usual subtypes of non-Hodgkin’s lymphoma(NHL),of which it is also a highly aggressive and heterogeneous malignancy of the hematological system.With the current standard treatment of rituximab immunochemotherapy,30-40% of patients still have difficulty in recovering or relapsing.Therefore,early identification of high-risk patients for stratified diagnosis and treatment is crucial.Autophagy is involved in various biological processes and exerts an influential effect in tumor formation and evolution.Nevertheless,the prognostic value of autophagy in DLBCL is uncertain.Methods We obtained two DLBCL gene sets and patient clinical characteristics data from TCGA and GEO databases(training set: GEO,n=412;validation set: TCGA-NCICCR,n=234).Overall,531 autophagy-related genes(ARGs)were available from the Human Autophagy Database,the Genome Enrichment Analysis website,and the literature.Subsequently,37 ARGs with a significant prognostic association were obtained using weighted gene co-expression network analysis(WGCNA),with which 37 genes were subjected to GO and KEGG enrichment analysis to characterize their biopathways and features.Then,using univariate and multivariate Cox regression and Lasso regression,three pivotal genes were identified for the construction of predictive prognostic models,and patients in the two datasets were grouped into high and low risk on the basis of median risk scores.This was followed by Kaplan-Meier survival analysis to assess the difference in predicted survival between the two groups,and thereafter,subject receiver operating characteristic(ROC)curves were generated and the area under the curve(AUC)values were considered an indicator of model strength.Independent prognostic analysis of clinical characteristics and risk scores of DLBCL was performed using univariate and multivariate Cox regression analysis,the inclusion of independent risk elements in nomograms,and assessment of their predictive efficacy.Finally,the degree of immune infiltration and the variability of immune checkpoints between high-and low-risk groups were compared employing single-sample gene set enrichment analysis(ss GSEA).Results Three signature ARGs,including BNIP3 L,TOMM40,and S100A9,were screened and used to construct a predictive prognostic risk score model.Patients with GSE10846 and TCGA-NCICCR were sorted into two groups: high-risk and low-risk.Of these,in the low risk group,the survival rate was remarkably high compared to the high risk cohort.The area under the ROC curve at 1,3 and 5 years was 0.642,0.692 and 0.690,respectively.In the training set,with relatively favorable predictive accuracy.The results of univariate and multivariate Cox regression analysis demonstrated that the value of risk score could be considered as an independent prognostic indicator.In addition,immune cell infiltration and immune checkpoint expression were significantly different between the two groups.Conclusion Based on three ARGs(BNIP3L,TOMM40 and S100A9),we developed a DLBCL risk score prediction model to provide some reference for patient prognosis assessment and clinical individualized treatment.