机器学习识别LRRC15和MICB为类风湿关节炎的免疫诊断标志物  

Machine learning identification of LRRC15 and MICB as immunodiagnostic markers for rheumatoid arthritis

作　　者：田彦虎 黄心岸 郭桐桐 如斯坦木·阿合坦木 罗江淼 肖遥 王超[2] 王维山[2] Tian Yanhu;Huang Xinan;Guo Tongtong;Rusitanmu·Ahetanmu;Luo Jiangmiao;Xiao Yao;Wang Chao;Wang Weishan(Medical College,Shihezi University,Shihezi 832008,Xinjiang Uygur Autonomous Region,China;Department of Orthopedics,the First Affiliated Hospital of Shihezi University Medical College,Shihezi 832008,Xinjiang Uygur Autonomous Region,China)

机构地区：[1]石河子大学医学院,新疆维吾尔自治区石河子市832008 [2]石河子大学第一附属医院骨科,新疆维吾尔自治区石河子市832008

出　　处：《中国组织工程研究》2025年第11期2411-2420,共10页Chinese Journal of Tissue Engineering Research

基　　金：国家自然科学基金项目(82160423),项目负责人:王维山。

摘　　要：背景:类风湿关节炎是一种慢性的自身免疫性疾病,早期诊断对于预防疾病进展和治疗至关重要。因此探究类风湿关节炎的诊断特征和免疫细胞浸润具有重要意义。目的:基于Gene Expression Omnibus(GEO)数据库,通过机器学习算法,筛选类风湿关节炎潜在重要的诊断标记物,并探讨类风湿关节炎的诊断特征与免疫细胞浸润的关系。方法:从GEO数据库获取类风湿关节炎相关的滑膜组织的基因表达数据集,采用批量效应去除法对数据集进行合并,采用R软件进行差异表达基因的鉴定和功能相关性分析,通过生物信息学分析和3种机器学习算法进行疾病特征基因的提取,筛选出类风湿关节炎相关的关键基因。此外,对所有差异表达基因进行免疫细胞浸润分析,分析类风湿关节炎的炎症状态,并对其诊断特征与浸润性免疫细胞的关系进行研究。结果与结论:(1)在类风湿关节炎和正常滑膜组织中,获得了179个差异表达基因,其中124个基因表达上调,55个基因表达下调;(2)富集分析显示类风湿关节炎和免疫反应之间存在良好的相关性;(3)通过3种机器学习算法分析发现,LRRC15和MICB可能是类风湿关节炎潜在的标记物;(4)LRRC15(曲线下面积=0.964,95%CI:0.924-0.992)和MICB(曲线下面积=0.961,95%CI:0.923-0.990)在验证数据集上有着较强的诊断能力;(5)13种免疫细胞浸润发生改变,以巨噬细胞为主;(6)在类风湿关节炎中,免疫细胞功能的多数促炎途径被激活;(7)免疫相关性分析发现LRRC15和MICB与M1型巨噬细胞的相关性最强;(8)结果发现LRRC15和MICB被确定为类风湿关节炎潜在的诊断标记物,具有较强的诊断性能并且与免疫细胞浸润具有显著的相关性;通过机器学习及生物信息学分析加深了对类风湿关节炎免疫浸润的理解,为类风湿关节炎的诊断和治疗提供新的思路。BACKGROUND:Rheumatoid arthritis is a chronic autoimmune disease.Early diagnosis is crucial for preventing disease progression and for effective treatment.Therefore,it is of significance to investigate the diagnostic characteristics and immune cell infiltration of rheumatoid arthritis.OBJECTIVE:Based on the Gene Expression Omnibus(GEO)database,to screen potential diagnostic markers of rheumatoid arthritis using machine learning algorithms and to investigate the relationship between the diagnostic characteristics of rheumatoid arthritis and immune cell infiltration in this pathology.METHODS:The gene expression datasets of synovial tissues related to rheumatoid arthritis were obtained from the GEO database.The data sets were merged using a batch effect removal method.Differential expression analysis and functional correlation analysis of genes were performed using R software.Bioinformatics analysis and three machine learning algorithms were used for the extraction of disease signature genes,and key genes related to rheumatoid arthritis were screened.Furthermore,we analyzed immune cell infiltration on all differentially expressed genes to examine the inflammatory state of rheumatoid arthritis and investigate the correlation between their diagnostic characteristics and infiltrating immune cells.RESULTS AND CONCLUSION:In both rheumatoid arthritis and normal synovial tissues,we identified 179 differentially expressed genes,with 124 genes upregulated and 55 genes down-regulated.Enrichment analysis revealed a significant correlation between rheumatoid arthritis and immune response.Three machine learning algorithms identified LRRC15 and MICB as potential biomarkers of rheumatoid arthritis.LRRC15(area under the curve=0.964,95%confidence interval:0.924-0.992)and MICB(area under the curve=0.961,95%confidence interval:0.923-0.990)demonstrated strong diagnostic performance on the validation dataset.The infiltration of 13 types of immune cells was altered,with macrophages being the most affected.In rheumatoid arthritis,the major

关 键 词：类风湿关节炎 机器学习 免疫浸润 诊断标记物 差异表达基因 

分 类 号：R459.9[医药卫生—治疗学] R318[医药卫生—临床医学] R684.3