Li Zexi, Liu Jing, Wu Yurui
Department of Thoracic Surgery and Oncology, Children's Hospital Affiliated to Capital Institute Pediatrics, Beijing, China.
Transl Cancer Res. 2025 May 30;14(5):2677-2689. doi: 10.21037/tcr-24-2048. Epub 2025 May 26.
Neuroblastoma (NBL) is a common pediatric malignancy with diverse prognoses influenced by multiple factors. Accurate overall survival (OS) predictions are essential for guiding treatment. However, the contribution of specific cell types within the tumor microenvironment (TME), which significantly influence disease progression, is often overlooked. This study aimed to develop an NBL prognostic model that incorporates TME, genetic, and clinical factors to improve prediction accuracy and clinical relevance.
Data were collected from the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) database (n=106, test set) and the Gene Expression Omnibus (GEO) database (n=238, train set). Including clinical details such as MYCN amplification, International NBL Staging System (INSS) stage, age at diagnosis, and OS outcomes. Additionally, single-cell RNA sequencing (scRNA-seq) data from 16 NBL patients (160,910 cells) were included to improve model precision. Uniform manifold approximation and projection (UMAP) was utilized for cell clustering, while weighted gene co-expression network analysis (WGCNA) helped identify cell-type-specific modules. Prognostic genes were pinpointed using univariate and multivariate Cox regression analyses, which also served to refine the model by integrating essential clinical variables and molecular markers. The model's effectiveness was assessed through Kaplan-Meier survival curves, receiver operating characteristic (ROC) curves, and calibration plots. Additional evaluations included immune cell infiltration and drug sensitivity analysis.
MYCN amplification was present in 79.4% of patients in the train set and 79.2% of patients in the test set, and the majority of patients in both cohorts were classified as Stage 4. The median age at diagnosis was 399.5 days in the train set and 1,069 days in the test set. Key findings demonstrate that Schwann cell-specific genes (, , ) considerably affect survival outcomes in NBL patients. The initial model showed robust predictive accuracy in the train set with areas under the curve (AUCs) of 0.832 and acceptable performance in the test set with AUC of 0.777. A refined model, incorporating three genes, two clinical indicators (age and INSS stage), and MYCN amplification, exhibited enhanced accuracy with AUC of 0.857. Differences in immune cell expression between high-risk and low-risk groups were noted, alongside significant disparities in drug sensitivity, indicating lower half maximal inhibitory concentration (IC50) values for targeted therapies in the high-risk group.
This study developed a model for predicting OS in NBL by integrating Schwann cell-specific genes, clinical factors, and the TME. The model highlights the importance of specific cellular contributions to prognosis and provides a more personalized approach to NBL treatment, particularly for high-risk patients.
神经母细胞瘤（NBL）是一种常见的儿科恶性肿瘤，其预后受多种因素影响，具有多样性。准确的总生存期（OS）预测对于指导治疗至关重要。然而，肿瘤微环境（TME）中特定细胞类型对疾病进展有显著影响，但其作用常常被忽视。本研究旨在开发一种整合TME、基因和临床因素的NBL预后模型，以提高预测准确性和临床相关性。
数据收集自治疗应用研究以生成有效治疗方案（TARGET）数据库（n = 106，测试集）和基因表达综合数据库（GEO）（n = 238，训练集）。包括诸如MYCN扩增、国际神经母细胞瘤分期系统（INSS）分期、诊断年龄和OS结果等临床细节。此外，纳入了16例NBL患者（160,910个细胞）的单细胞RNA测序（scRNA-seq）数据，以提高模型精度。使用均匀流形近似和投影（UMAP）进行细胞聚类，而加权基因共表达网络分析（WGCNA）有助于识别细胞类型特异性模块。使用单变量和多变量Cox回归分析确定预后基因，这也通过整合基本临床变量和分子标记来优化模型。通过Kaplan-Meier生存曲线、受试者工作特征（ROC）曲线和校准图评估模型的有效性。其他评估包括免疫细胞浸润和药物敏感性分析。
训练集中79.4%的患者和测试集中79.2%的患者存在MYCN扩增，两个队列中的大多数患者被归类为4期。训练集的诊断中位年龄为399.5天，测试集为1,069天。主要发现表明，雪旺细胞特异性基因（ ， ， ）对NBL患者的生存结果有显著影响。初始模型在训练集中显示出强大的预测准确性，曲线下面积（AUC）为0.832，在测试集中表现可接受，AUC为0.777。一个改进的模型，纳入三个基因、两个临床指标（年龄和INSS分期）以及MYCN扩增，表现出更高准确性，AUC为0.857。注意到高风险组和低风险组之间免疫细胞表达的差异，以及药物敏感性的显著差异，表明高风险组中靶向治疗的半数最大抑制浓度（IC50）值较低。
本研究通过整合雪旺细胞特异性基因、临床因素和TME开发了一种预测NBL患者OS的模型。该模型突出了特定细胞对预后的重要性，并为NBL治疗提供了更个性化的方法，特别是对于高风险患者。
