基于WGCNA和SVM-RFE算法挖掘肺腺癌诊断和预后基因标志物

Identification of Gene Signatures Associated With Lung Adenocarcinoma Diagnosis and Prognosis Based on WGCNA and SVM-RFE Algorithm

目的肺癌是世界上最常见的癌症之一,在众多肺癌患者中,肺腺癌(lung adenocarcinoma,LUAD)的死亡率最高。基因表达谱的变化与肿瘤的发生和发展过程有关,通过识别与LUAD患者相关的诊断和预后基因标志物,可以为肺腺癌的预防和治疗提供理论依据。方法本研究以肿瘤基因组图谱(The Cancer Gene Atlas,TCGA)数据库为基础,采用加权基因共表达网络分析(weighted gene co-expression network analysis,WGCNA)、差异基因分析、cox回归分析、蛋白质互作网络(protein-protein interaction,PPI)分析等方法筛选与LUAD形成过程高度相关的hub基因。将TCGA和基因型组织表达(GTEx genotype tissue expression,GTEx)数据库中的RNA数据合并划分为训练集和内部验证集,利用基于支持向量机的递归特征消除算法(support vector machine recursive feature elimination feature,SVM-RFE)构建诊断模型并进行验证。GSE32863和GSE31210数据集分别用于验证诊断模型的准确性和基因标志物的预后价值。结果SVM-RFE算法得到的5个基因标志物(anln、cenpa、plk1、tpx2、cdca3)模型在LUAD患者分类中具有显著的诊断能力。功能富集分析表明,这5个基因与肿瘤发生发展的生物学过程密切相关。此外,这5个基因高表达的LUAD患者的预后表现不良,死亡率显著高于低表达的患者。结论我们的研究为LUAD的诊断和预后提供了具有5个基因特征的模型,这对于开发用于精确治疗的新靶点具有重要意义。

Objective Lung cancer is one of the most common cancers in the world. Lung adenocarcinoma(LUAD) has the highest annual mortality rate among lung cancer patients. It has been reported that changes in gene spectrum were associated with the process of tumorigenesis and its development. The purpose of this study is to identify the gene signatures associated with LUAD and to further analyze their prognostic significance. Methods Weighted gene co-expression network analysis(WGCNA), differential gene analysis, cox regression analysis, and protein-protein interaction(PPI) network analysis were used to screen the hub genes highly related to LUAD based on The Cancer Genome Atlas(TCGA) database. The RNA-seq data sets from TCGA and GTEx(Genotype Tissue Expression) database were combined and divided into a training set and a validation set, which were used to construct the diagnostic model by support vector machine recursive feature elimination feature(SVM-RFE) algorithm. GSE32863 and GSE31210 were used to verify the diagnostic accuracy of the model and the prognostic value of our obtained gene signatures,respectively. Results The results demonstrated that the model of 5 gene signatures(anln, cenpa, plk1, tpx2, cdca3) obtained by the SVM-RFE algorithm had an outstanding performance in the classification of LUAD patients. Functional enrichment analysis showed that these 5 gene signatures were highly related to the biological process of tumor initiation and progression. What’ s more, LUAD patients with high expression of these 5 genes also exerted a poor outcome in survival status. Conclusion Therefore, we could conclude that our study obtained useful models with 5 gene signatures for the diagnosis and prognosis of LUAD, which were essential for the development of novel targets applied in precision therapy.