基于机器学习的胃黏液分泌性腺癌特征基因筛选及免疫浸润分析  

Machine learning-based gene screening for gastric mucinous adenocarcinoma and immune infiltrationanalysis

作　　者：张原 张鹏善 罗再 徐亦天 蔡谦谦 黄陈 Zhang Yuan;Zhang Pengshan;Luo Zai;Xu Yitian;Cai Qianqian;Huang Chen(Department of Gastrointestinal Surgery,Shanghai General Hospital,Shanghai Jiaotong University School of Medicine,SShanghai 201620,China)

机构地区：[1]上海交通大学医学院附属第一人民医院胃肠外科,上海201620

出　　处：《中华实验外科杂志》2023年第5期939-942,共4页Chinese Journal of Experimental Surgery

基　　金：国家自然科学基金(82072662、82203751);上海市申康三年行动计划项目(SHDC2020CR4022);上海交通大学医院高峰高原计划-"研究型医师"(第二轮,20191425);上海市抗癌协会"雏鹰"计划(SACA-CY21C08)。

摘　　要：目的通过机器学习结合生物信息学方法分析胃黏液分泌性腺癌(GMA)的特征基因及特征通路,探索GMA的诊断标志物及其与免疫细胞浸润的关联。方法从癌症基因组图谱(TCGA)数据库中下载胃癌(GC)测序数据,以P<0.05为标准筛选99例GMA与311例胃普通型腺癌(GTA)样本差异表达基因(DEGs)。通过功能富集方法分析GMA特征通路和功能途径,进一步通过机器学习算法,使用套索回归(LASSO)结合支持向量机递归特征消除(SVM-RFE)获取23个特征基因,并通过绘制受试者工作特征曲线(ROC)分析DEGs的诊断效能,及其与免疫细胞浸润的相关性。在基因表达谱数据库(GEO)数据集GSE113255中对DEGs的表达和诊断效能进行验证。结果在TCGA数据库中检测到1769个DEGs(P<0.05),其中759个在GMA中上调,而1010个在GTA中上调。这些基因与磷脂酰肌醇3激酶(PI3K)-蛋白激酶B(Akt)信号通路、细胞黏附、细胞外基质、细胞骨架调节等途径显著相关(P<0.01)。GMA组中CMTM8(11.416比17.845,P<0.001)、RTKN(8.456比13.845,P<0.001)、TOMM34(23.219比36.991,P<0.001)平均表达量均低于GTA组,是GMA的诊断标志物。GMA组中M2巨噬细胞(P<0.01)、静息肥大细胞(P<0.001)浸润水平高于GTA组,而激活CD4+T细胞(P<0.01)浸润水平低于GTA组。结论CMTM8、RTKN和TOMM34是GMA的特征基因,与M2巨噬细胞和静息肥大细胞浸润等显著相关。Objective To analyze the signature genes and pathways of gastric mucinous adenocar-cinoma(GMA)by machine learning combined with bioinformatics methods,and to explore the diagnostic markers of GMA and their association with immune cell infiltration.Methods Gastric cancer(GC)se-quencing data were downloaded from the cancer genome atlas(TCGA)database.Differentially expressed genes(DEGs)of 99 patients with GMA and 311 patients with gastric traditional adenocarcinoma(GTA)were screened at P<0.05.The GMA feature pathways and functional pathways were analyzed by functional enrichment methods,and further analyzed by machine learning algorithms using the least absolute selection and shrinkage operator(LASSO)combined with support vector machine recursive feature elimination(SVM-RFE)to obtain 23 signature genes,then the diagnostic efficacy of DEGs and their correlation with immune cell infiltration were analyzed by plotting the receiver operating characteristic(ROC)curve.The expression and diagnostic efficacy of DEGs were validated in the gene expression omnibus(GEO)dataset GSE113255.Results A total of 1769 significant DEGs were detected in the TCGA database:759 were upregulated in GMA,and 1 O1Owere upregulated in GTA.These genes were significantly associated with phosphatidylinositol 3 kinase(PI3K)-protein kinase B(Akt)signaling pathway,cell adhesion,extracellu-lar matrix,cytoskeleton regulation and other pathways(P<0.01).CMTM8(11.416 vs.17.845,P<0.001),RTKN(8.456 vs.13.845,P<0.001),T0MM34(23.219 vs.36.991,P<0.001)in the GMA group had lower average expression than in the GTA group,and were diagnostic markers of GMA.The levels of M2 macrophage(P<0.01)and resting mast cell(P<0.001)infiltration were higher in the GMA group than in the CTA group,while the levels of activated CD4*T cell infiltration(P<O.O1)were lower than in the GTA group.Conclusion CMTM8,RTKN and TOMM34 are signature genes of CMA,which are correlated with M2 macrophages and resting mast cell infiltration.

关 键 词：胃癌 黏液分泌性腺癌 机器学习 免疫浸润 生物标志物 

分 类 号：R735.2[医药卫生—肿瘤]