摘要
目的 使用无创的区域放射学相似性网络(R2SN)方法获取胶质瘤肿瘤组织信息,构建脑胶质瘤的预后预测模型.方法 使用R2SN结合基因相似性网络,计算胶质瘤病人与健康对照间的差异基因.采用使用基因表达信息估计恶性肿瘤中的间质和免疫细胞结合加权基因共表达网络分析,鉴定与肿瘤免疫微环境相关的基因.通过Lasso回归筛选关键基因并结合临床因素建立胶质瘤预后模型.结果 确定3个关键基因(PIWIL4、TGIF1和XKR8)建立风险标签,以风险标签结合临床因素构建的脑胶质瘤预后模型预测1、3、5年生存率的受试者工作特征曲线下面积为0.870、0.888和0.800,预测效果较好.结论 结合影像和基因信息可以无创、准确地预测脑胶质瘤的预后.
Abstract
Objective To establish a prognostic prediction model for glioma by obtaining the information on glioma tumor tissue through noninvasive regional radiomics similarity network(R2SN).Methods R2SN combined with gene similarity net-work was used to calculate the differentially expressed genes between glioma patients and healthy controls.The gene expression da-ta were usedto estimate stromal and immune cells in malignant tumor,and the weighted gene co-expression network analysis was also performed to identify genes associated with tumor immune microenvironment.ALasso regression analysis was used to identify the key genes,and a prognostic model was established for glioma with reference to clinical factors.Results Three key genes,i.e.,PIWIL4,TGIF1,and XKR8,were identified to establish risk labels.The prognostic model for glioma established based on risk labels andclinical factors had an area under the ROC curve of 0.870,0.888,and 0.800,respectively,in predicting 1,3,and 5 year survival rates,suggesting that the model had a good predictive effect.Conclusion The combination of imaging findings and genetic information can predict the prognosis of glioma in a noninvasive and accurate manner.
维普
万方数据