基于SVM机器学习筛选多发性骨髓瘤的特征基因及免疫浸润分析  

Screening for feature genes and immune infiltration of multiple myeloma:a study based on support vector machine

作　　者：石蕾 张红宾[1] Shi lei;Zhang Hongbing(Department of Hematology,The First Affiliated Hospital of Chongqing Medical University)

机构地区：[1]重庆医科大学附属第一医院血液科,重庆400042

出　　处：《重庆医科大学学报》2025年第1期135-144,共10页Journal of Chongqing Medical University

摘　　要：目的:应用生物信息学技术探究多发性骨髓瘤(multiple myeloma,MM)的遗传异质性以及免疫细胞在其病理生理中发挥的重要调节作用。方法:通过美国国家生物技术信息中心(National Center for Biotechnology Information,NCBI)公共高通量基因表达(gene expression omnibus,GEO)数据库对MM相关的数据集GSE125364、GSE72213采用生物信息学及机器学习等方法筛选MM诊断的关键基因,探索MM差异基因相关通路计算免疫细胞浸润情况,并通过分子生物学实验进行验证。结果:基于公共数据库的多发性骨髓瘤基因芯片数据采用生物信息学方法分析筛选差异基因410个,其中MM患者较对照组下调259个,上调151个。通过GO富集分析发现差异基因主要参与生物学过程包括DNA复制、染色体分离及有丝分裂;细胞定位主要富集在染色体区、纺锤体;分子功能主要富集在单链DNA螺旋酶活性,作用于DNA的催化和依赖ATP的活性等。KEGG通路富集分析显示的主要信号通路包括细胞周期、p53信号通路、细胞衰老和DNA复制等。GSEA分析对照组主要富集细胞周期、DNA复制、嘌呤代谢及核糖体等通路,MM组主要富集脂肪细胞因子信号通路、细胞粘附分子、核糖核酸多聚酶及抗坏血酸和醛酸代谢等通路。通过支持向量机递归特征消除(support vector machine-recursive feature elimination,SVM-RFE)算法筛选出2个MM诊断基因CPXM1和UROD,通过CIBERSORTx进行了免疫浸润分析表明CPXM1和UROD与免疫浸润相关,并通过MM.1S细胞进行了qRT-PCR验证(P<0.05)。结论:采用生物信息学方法能够有效分析MM与正常对照人群差异表达的基因,本次研究筛选出多发性骨髓瘤诊断的关键基因CPXM1和UROD,其表达与免疫浸润相关,可作为后续MM基础和临床实验研究的新靶点。Objective:To investigate the genetic heterogeneity of multiple myeloma(MM)and the important regulatory role of immune cells in its pathophysiology by using bioinformatics techniques.Methods:The datasets of GSE125364 and GSE72213 associated with MM were obtained from the gene expression omnibus database of National Center for Biotechnology Information,and bioinformatics and machine learning methods were used to identify the key genes for the diagnosis of MM.Pathways associated with the differentially expressed genes in MM were analyzed to calculate immune cell infiltration,and molecular biology experiments were used for validation.Results:In this study,a total of 410 differentially expressed genes were obtained by the bioinformatics methods based on the gene microarray data of MM from public databases,among which 259 were downregulated and 151 were upregulated in MM patients compared with controls.The gene ontology enrichment analysis showed that the differentially expressed genes were mainly involved in the biological processes such as DNA replication,chromosome segregation,and mitosis;as for cellular localization,they were mainly enriched in chromosomal region and the spindle apparatus;as for molecular function,they were mainly enriched in single-stranded DNA helicase activity,DNA catalysis,and ATP-dependent activity.The KEGG pathway enrichment analysis showed that the main signaling pathways included cell cycle,the p53 signaling pathway,cellular senescence,and DNA replication.The GSEA analysis showed that in the control group,the genes were mainly enriched in cell cycle,DNA replication,purine metabolism,and ribosomes,while in the MM group,the genes were mainly enriched in the adipokine signaling pathway,cell adhesion molecules,ribonucleic acid polymerase,and ascorbate and aldarate metabolism pathways.Two genes,CPXM1 and UROD,were obtained for the diagnosis of MM by support vector machine-recursive feature elimination algorithm,and the immune infiltration analysis via CIBERSORTx showed that CPXM1 and UROD were a

关 键 词：多发性骨髓瘤 生物标志物 生物信息学 免疫浸润 

分 类 号：R733.3[医药卫生—肿瘤]