加权基因共表达网络联合机器学习识别影响卵巢癌患者生存期的关键基因  

Weighted Gene Co-expression Network Combined with Machine Learning Identifies Key Genes Influencing Survival of Ovarian Cancer Patients

作　　者：张后娟 ZHANG Houjuan(Department of Obstetrics and Gynecology,Ningyang Hospital of Traditional Chinese Medicine,Ningyang County,Shandong Province 271400)

机构地区：[1]山东省宁阳县中医院妇产科,271400

出　　处：《医学理论与实践》2025年第4期550-554,共5页The Journal of Medical Theory and Practice

摘　　要：目的:利用加权基因共表达网络和机器学习方法,筛选和鉴定对卵巢癌患者生存率产生影响的关键基因,为卵巢癌的治疗提供新的思路和靶点。方法:从公共数据库下载GSE14407、GSE15578和GSE38666三个包含卵巢癌样本的表达矩阵,将前两个矩阵进行合并和去批次作为训练集用于后续分析,最后一个表达矩阵作为外部验证集。按照一定的标准筛选训练集中卵巢癌样本同正常样本之间的差异基因,对训练集进行加权基因共表达网络分析,获取相关性最强模块中的特征基因。使用3种机器学习对差异基因与特征基因的交集基因进行深度学习,并识别卵巢癌的关键基因。最后对关键基因的诊断效能和对患者生存期的影响进行验证。结果:共获得770个差异基因,191个特征基因,两者交集基因138个;通过机器学习发现AOX1、GPRASP1可能是对卵巢癌患者生存期产生影响的关键基因;受试者工作曲线及Kaplan-Meier分析发现AOX1、GPRASP1不但具有良好的诊断效能,同时确实对卵巢癌患者生存率有显著影响。结论:基于加权基因共表达网络和机器学习对公共数据集的分析发现,AOX1、GPRASP1是影响卵巢癌患者生存率的关键基因,可以为卵巢癌的治疗提供新的思路。Objective:Utilizing weighted gene co-expression network analysis and machine learning methods to screen and identify key genes affecting the survival rate of ovarian cancer patients,providing new insights and targets for the treatment of ovarian cancer.Methods:Gene expression matrices containing ovarian cancer samples were downloaded from public databases GSE14407,GSE15578,and GSE38666.The first two matrices were merged and batch effects were removed to form a training set for subsequent analysis,with the last matrix used as an external validation set.Differential genes between ovarian cancer samples and normal samples within the training set were selected according to certain criteria.Weighted gene co-expression network analysis was performed on the training set to identify feature genes in the modules with the strongest correlations.Three machine learning techniques were used for deep learning on the intersection genes between differential genes and feature genes,aiming to identify key genes of ovarian cancer.The diagnostic efficacy and impact on patient survival of the key genes were validated.Results:A total of 770 differential genes,191 feature genes,and 138 intersecting genes were obtained.Machine learning revealed that AOX1 and GPRASP1 may be key genes influencing the survival of ovarian cancer patients.Receiver operating characteristic curve analysis and Kaplan-Meier analysis showed that AOX1 and GPRASP1 not only have good diagnostic efficacy but also significantly impact the survival rate of ovarian cancer patients.Conclusion:Analysis of public datasets using weighted gene co-expression network and machine learning methods identified AOX1 and GPRASP1 as key genes influencing the survival rate of ovarian cancer patients,offering new avenues for the treatment of ovarian cancer.

关 键 词：加权共表达网络 机器学习 卵巢癌 关键基因 

分 类 号：R737.31[医药卫生—肿瘤]