预测肾透明细胞癌病理分级的影像基因组学研究

Radiogenomics Study on Predicting Pathological Grading of Renal Clear Cell Carcinoma

目的 探究CT影像组学特征与mRNA的关联性,说明CT影像组学模型预测肾透明细胞癌(ccRCC)病理分级的潜在基因组学机制。方法 利用影像组学分析队列筛选影像组学特征并构建影像组学模型预测肾癌病理分级。利用基因组学分析队列识别病理分级相关基因模块,筛选前30个关键节点(Hub)基因并使用Hub基因构建基因模型预测病理分级,将结果映射至影像基因组学拓展队列并将基因与影像组学特征关联。结果 筛选出4个与病理分级相关的影像组学特征,其构建的支持向量机模型在测试集中曲线下面积为0.938。WGCNA分析得到5个基因模块与病理分级相关,影像基因组学分析发现5个基因模块中有271个基因相关,332个基因与组学特征不相关,前30个Hub基因有8个与影像组学特征相关,22个Hub基因不相关,两者所富集的通路不一致。从30个Hub基因筛选得到的5个Hub基因组成的逻辑回归基因模型在测试集中预测效能为0.736,用于构建模型的5个基因中有2个基因与影像组学特征相关,3个基因与影像组学特征不相关。结论 CT影像组学模型较基于mRNA构建的基因模型的预测效能高。CT影像组学特征与病理分级相关的mRNA之间的关联不具有普遍性。

Objective To develop the models based on radiomics and genomics respectively for predicting the histopathologic nuclear grade in patients with localized clear cell renal cell carcinoma(ccRCC) and to explore the correlation between radiomics features and RNA sequencing data for demonstrating the driving mechanism underlying that macro-radiomics models can predict the microscopic pathological changes.Methods In this multi-institutional retrospective study,a CT radiomics model was developed for the nuclear grade prediction.Based on genomics analysis cohort,nuclear grade-associated gene modules were identified,and gene model was constructed based on top 30 Hub mRNA to predict nuclear grade.Using radiogenomics development cohort,biological pathways were enriched from Hub genes and radiogenomics map was created.Results The 4-features-based SVM model predicted nuclear grade with an AUC of 0.938 in validation set.A 5-gene-based model predicted nuclear grade with an AUC of 0.736 in genomics analysis cohort.A total of five gene modules was identified to be associated with nuclear grade.Radiomics features were only associated significantly with 271 genes of 603 genes in 5 gene modules and 8 genes of top 30 Hub genes.Difference existed in the enrichment pathway between associated and unassociated with radiomics features,which were associated with 2 genes of five-gene signatures in mRNA model.Conclusion The CT radiomics models had higher predictive performance than mRNA models.The association between radiomics features and mRNA which related to nuclear grade is not universal.