基于机器学习评估曲妥珠单抗耐药相关基因在胃癌中的诊断和预后效能  

Evaluation of diagnostic and prognostic relevance of genes related to trastuzumab resistance in gastric cancer based on machine learning

作　　者：刘涛 李桐桐 余春燕 黄翌楚 姜雷[2] LIU Tao;LI Tongtong;YU Chunyan;HUANG Yichu;JIANG Lei(The First School of Clinical Medicine,Lanzhou University,Lanzhou 730000,China;Department of General Surgery,the First Hospital of Lanzhou University,Lanzhou 730000,China)

机构地区：[1]兰州大学第一临床医学院,甘肃兰州730000 [2]兰州大学第一医院普外科,甘肃兰州730000

出　　处：《中国医学物理学杂志》2025年第4期525-533,共9页Chinese Journal of Medical Physics

基　　金：国家自然科学基金(82060527);兰州大学医学教育创新发展项目优秀青年支持计划(lzuyxcx-2022-175)。

摘　　要：目的:利用机器学习算法探索曲妥珠单抗耐药与敏感相关基因在胃癌中的诊断和预后效能。方法:从GEO数据库中下载耐药和敏感基因数据,进行功能富集分析。利用TCGA数据以及GEO数据进行交集分析,筛选出与胃癌耐药相关的特征基因。采用LASSO以及SVM-RFE方法进行特征基因的筛选。在测试组和验证组中评估特征基因的表达情况,并通过受试者工作特征曲线分析这些基因的诊断价值。利用在线数据库分析SH3GL2的预后价值,进一步探讨其在胃癌患者生存期中的作用;采用CIBERSORT算法评估SH3GL2与胃癌免疫细胞浸润的关系,分析其对免疫微环境的影响。结果:得到15个耐药相关基因,基于机器学习筛选出12个与胃癌相关的诊断生物标志物,包括MMP7、COCH、VCAN、SH3GL2、SYNM、KLK6、STC2、PPP1R1B、CDH3、WNT11、PMEPA1和BCAT1。SH3GL2在测试组和验证组中均表现为低表达,其高表达与胃癌的较差预后相关(P<0.01)。SH3GL2的表达水平与多种免疫细胞(激活的CD8+T细胞、激活的DC细胞)相关,与免疫抑制因子(如TGFB1、VTCN1)呈正相关,与免疫刺激因子(如CD70、CD80)呈负相关。结论:12个筛选出的特征基因可能成为胃癌的潜在诊断生物标志物。SH3GL2在胃癌中低表达,其高表达可能通过抑制抗肿瘤免疫以缩短胃癌患者的生存期。Objective To explore the diagnostic and prognostic relevance of genes associated with trastuzumab resistance and sensitivity in gastric cancer using machine learning algorithms.Methods The data on resistant and sensitive genes were downloaded from the GEO database and subjected to functional enrichment analysis.Intersection analysis was performed using TCGA and GEO data to identify feature genes related to gastric cancer drug-resistance.LASSO and SVM-RFE methods were used for feature gene selection.The expressions of these feature genes were detected in both test and validation groups,and their diagnostic value was analyzed using receiver operating characteristic curves.The prognostic value of SH3GL2 was assessed using online databases,and its role in patient survival was further explored.CIBERSORT algorithm was used to evaluate the relationship between SH3GL2 and immune cell infiltration in gastric cancer,and analyze its effect on immune microenvironment.Results Fifteen resistance-related genes were identified,and 12 diagnostic biomarkers related to gastric cancer were selected through machine learning,including MMP7,COCH,VCAN,SH3GL2,SYNM,KLK6,STC2,PPP1R1B,CDH3,WNT11,PMEPA1,and BCAT1.SH3GL2 showed low expression in both test and validation groups,and its high expression was associated with poorer prognosis in gastric cancer(P<0.01).SH3GL2 expression level was related to various immune cells(activated CD8+T cells,activated DC cells)and showed positive correlations with immune suppressive factors(such as TGFB1,VTCN1)and negative correlations with immune stimulatory factors(such as CD70,CD80).Conclusion The 12 selected feature genes can serve as potential diagnostic biomarkers for gastric cancer.SH3GL2 has a low expression in gastric cancer,and its high expression might shorten patient survival by inhibiting anti-tumor immunity.

关 键 词：胃癌 耐药 曲妥珠单抗 机器学习 生物信息学 

分 类 号：R318[医药卫生—生物医学工程] R735.2[医药卫生—基础医学]