基于生物信息学和机器学习识别与验证卵巢癌中糖酵解相关生物标志物

Identification and validation of glycolysis-related biomarkers in ovarian cancer based on bioinformatics and machine learning

目的基于生物信息学和机器学习,旨在从糖酵解信号通路相关基因中探索对卵巢癌(OC)具有诊断价值的潜在生物标志物、治疗靶点。方法OC数据集来自GEO数据库及TCGA数据库,从KEGG数据库以及文献收集、整理得到糖酵解相关基因。通过差异表达分析、蛋白间相互作用(PPI)网络构建和机器学习算法包括最小绝对收缩和选择算法逻辑回归、支持向量机递归特征消除和随机森林算法来识别生物标志物。开发受试者工作特征(ROC)曲线以评估诊断价值。为研究识别到的生物标志物在OC中的潜在作用机制,随后进行了富集分析、药物敏感性分析,并采用TIMER、EPIC、MCPCOUNTER三种算法对生物标志物进行了免疫浸润分析。结果共获得67个糖酵解相关基因,其中有20个基因在OC中差异表达。通过PPI网络筛选出10个枢纽基因,并通过机器学习识别到8个生物标志物,两者有6个交集基因。诊断价值评估表明8个生物标志物均具有较高的诊断价值(ROC曲线下面积为>0.7),此外,它们与肿瘤免疫细胞浸润、药物反应联系密切。结论多数糖酵解相关基因在OC患者中表达异常,为鉴别OC提供了诊断价值,其中GPI、ENO3性能最优。本研究可能为OC患者提供潜在的诊断生物标志物以及治疗靶点。

Objective To explore potential biomarkers and therapeutic targets with diagnostic value for ovarian cancer(OC)from genes related to glycolysis signaling pathways based on bioinformatics and machine learning.Methods The OC dataset was obtained from GEO database and TCGA database,and glycolysis-related genes were collected and collated from KEGG database and literature.Biomarkers were identified by differential expression analysis,protein-protein interaction(PPI)network construction,and machine learning algorithms including least absolute shrinkage and selection algorithm logistic regression,support vector machine-recursive feature elimination,and random forest algorithms.Receiver operating characteristic(ROC)curves were developed to assess diagnostic value.In order to study the potential mechanism of action of the identified biomarkers in OC,enrichment analysis and drug sensitivity analysis were conducted,and three algorithms including TIMER,EPIC,and MCPCOUNTER were used to perform immunoinfiltration analysis on the biomarkers.Results A total of 67 glycolysis-related genes were obtained,20 of which were differentially expressed in OC.Ten pivotal genes were screened by PPI network,and eight biomarkers were identified by machine learning,with six intersecting genes between them.Diagnostic value assessment showed that all the eight biomarkers had high diagnostic value(area under the ROC curve>0.7).In addition,they were closely associated with tumor immune cell infiltration and drug response.Conclusion Most glycolysis-related genes are aberrantly expressed in OC patients,providing diagnostic value for identifying OC,with GPI and ENO3 performing optimally.This study may provide a potential diagnostic biomarker as well as a therapeutic target for OC patients.