基于GEO数据库构建神经母细胞瘤缺氧相关的预后模型

Construction of Hypoxia-Related Prognostic Signature in Neuroblastoma Based on GEO Database

目的:基于基因表达综合(Gene Expression Omnibus, GEO)数据库分析缺氧相关基因与神经母细胞瘤(NB)临床特征及免疫环境的关系,构建缺氧相关预后模型。方法:从GEO数据库下载数据集GSE62564,从MSigDB数据库下载缺氧相关基因集。通过共识聚类算法将NB患者分为高低缺氧组,通过Kaplan-Meier曲线、卡方检验、“ESTIMATE”算法等比较两组间临床、免疫环境差别。通过LASSO回归建立缺氧相关预后模型,使用受试者工作特征(ROC)曲线下面积(AUC)评价模型的预测性能,并对模型进行独立预后分析。结果:根据缺氧相关基因将NB样本分为两组,两组患者的生存、年龄、肿瘤分期及MYCN状态均存在显著差异。两组样本的免疫浸润情况存在显著差异,B细胞、树突状细胞、NK细胞及T细胞等多种免疫细胞含量在两组间存在差异。通过LASSO回归得到了6个缺氧相关基因(CSRP2, DTNA, SAP30, NCAN, WSB1, NAGK)构成的预后模型。通过模型的风险评分中位数将患者分为高、低风险组,Kaplan-Meier曲线显示两组间预后存在显著差异。模型在1,3,5年时的AUC值分别为0.882,0.916及0.914,说明其具有良好的预测价值。多因素COX回归分析表明风险评分可作为NB的独立预后因素。结论:缺氧影响NB的免疫环境及患者预后,由缺氧相关基因构成的预后模型能较好评估NB患者预后,并为寻找新的治疗靶点提供帮助。

Aims: Analyzing relationship between hypoxia-related genes and clinical features and immune en-vironment of neuroblastoma (NB) based on GEO (Gene Expression Omnibus) database to construct hypoxia-related prognostic signature. Methods: GSE62564 was downloaded from GEO database and hypoxia-related gene set was obtained from MSigDB database. Patients in GSE62564 were divided into high/low hypoxia subgroups via Consensus Clustering analysis. Differences in immune and clinical features between two groups were identified by Kaplan-Meier curve, Chi- square test and ESTIMATE algorithm. Hypoxia-related prognostic signature was constructed by Lasso-Cox regres-sion. AUC value of the signature was calculated for the evaluation of the prognostic model. Results: Clinical features including prognosis, age, tumor stage and MYCN status were significantly different in two subgroups. Immune environment and distribution of immune cell including B cells, T cells, NK cells and dendritic cells were also different between groups. A six-gene hypoxia-related prog-nostic signature (CSRP2, DTNA, SAP30, NCAN, WSB1, NAGK) was constructed using Lasso-Cox re-gression, and patients were divided into high-/low-risk group by median of risk score. Kaplan-Meier curve revealed the significant difference between prognosis of the two groups. The AUCs for the 1-, 3-, 5-year OS predictions for the signature were 0.882, 0.916 and 0.914 which revealed the prog-nostic value of the signature. Risk score was identified as an independent prognostic factor by the multivariate Cox regression analysis. Conclusion: Hypoxia has an effect on immune environment and prognosis of NB patients. A hypoxia-related prognostic signature was constructed, which will contribute to predicting prognosis of NB and finding new therapy target.