基于机器学习识别膀胱癌双硫死亡关键特征基因  

Key Characteristic Genes Related to Disulfide Death in Bladder Cancer Based on Machine Learning

作　　者：吴江 杨柳 李岗[2] WU Jiang;YANG Liu;LI Gang(Changzhi Medical College)

机构地区：[1]长治医学院,山西长治046000 [2]长治医学院附属和济医院泌尿外科

出　　处：《长治医学院学报》2025年第1期6-12,共7页Journal of Changzhi Medical College

摘　　要：目的:利用生物信息学和机器学习方法筛选膀胱癌双硫死亡关键差异基因,并探讨其生物学功能。方法:从GEO数据库中获得膀胱癌数据集GSE13507,筛选膀胱癌癌组织与癌旁组织双硫死亡差异表达基因,对筛选差异表达基因进行功能富集分析;通过LASSO回归分析及支持向量机(SVM)算法寻找特征基因,将差异表达基因与特征基因取交集得到关键差异表达基因;对关键基因进行受试者工作特征曲线分析(ROC)及免疫浸润分析。结果:共得到7个双硫死亡差异表达基因,GO功能主要富集在细胞形态发生的调节、肌动蛋白细胞骨架组织的调节、基于肌动蛋白丝的过程调节等方面,KEGG主要富集在粘着斑、肌动蛋白细胞骨架的调节、粘附连接、紧密连接等通路;通过LASSO和SVM共筛选到3个双硫死亡差异关键基因ACTN4、FLNA、IQGAP1,其中ACTN4基因在正常组织和肿瘤组织表达无明显差异(P>0.05),FLNA、IQGAP1基因在正常组织中高表达,在肿瘤组织中低表达,3个特征基因ROC的AUC值均>0.8,具有良好的诊断效能;免疫浸润分析显示3个基因与CD8^(+)T细胞等免疫细胞浸润呈显著正相关。结论:通过生物信息学方法联合机器学习筛选出膀胱癌双硫死亡关键差异基因ACTN4、FLNA、IQGAP1,这些基因在膀胱癌的发生发展中发挥重要作用,因而有望成为疾病诊断的潜在分子生物标志物。Objective:To screen the disulfide-related key genes in bladder cancer and explore their biological functions using bioinformatics.Methods:The dataset GSE13507 of bladder cancer was obtained from the GEO database.The"limma"package in R software was used to analyze and screen the differentially expressed genes related to disulfide death in bladder cancer tissues and adjacent tissues.The functional enrichment analysis of the differentially expressed genes was performed.Feature genes were identified by the LASSO regression analysis and SVM algorithm The key differentially expressed genes were obtained by the intersection of differentially expressed genes and characteristic genes.Receiver operating characteristic curve(ROC)analysis and immune infiltration analysis were performed for key characteristic genes.Results:A total of seven differentially expressed disulfide death genes were obtained from the GSE13507 dataset.GO functions were mainly concentrated on the regulation of cell morphogenesis,the regulation of actin cytoskeleton tissue,and the regulation of actin filament-based processes.KEGG enrichment was mainly concentrated on the pathways that were adhesion plaques,the regulation of actin cytoskeleton,adhesion bonding,and tight bonding.Three key genes related to disulfide death,ACTN4,FLNA,and IQGAP1,were screened by LASSO and SVM.There was no significant difference in the expression of the ACTN4 gene between normal tissues and tumor tissues(P>0.05),while FLNA and IQGAP1 genes were highly expressed in normal tissues and low expressed in tumor tissues.The AUC values of the three characteristic genes of ACTN4,FLNA,and IQGAP1 were all greater than 0.8,which showed good diagnostic efficiency.Immune infiltration analysis showed that 3 genes were significantly positively correlated with immune cell infiltration such as CD8^(+)T cells.Conclusion:The key differential genes of disulfide death in bladder cancer,ACTN4,FLNA,and IQGAP1,were screened through bioinformatics combined with machine learning.These genes play an impo

关 键 词：膀胱癌 双硫死亡 生物信息学分析 机器学习 

分 类 号：R737.14[医药卫生—肿瘤]