机器学习鉴定KDELR3作为骨关节炎缺氧特征基因的实验验证  

Experimental validation of machine learning identification of KDELR3 as a signature gene for osteoarthritis hypoxia

作　　者：徐文飞 明春玉 梅其杰[1] 袁长深[1] 郭锦荣[1] 曾超[1] 段戡[1] Xu Wenfei;Ming Chunyu;Mei Qijie;Yuan Changshen;Guo Jinrong;Zeng Chao;Duan Kan(Orthopedic Department of the Limbs,First Affiliated Hospital of Guangxi University of Chinese Medicine,Nanning 530023,Guangxi Zhuang Autonomous Region,China;Department of Geriatrics,Ruikang Affiliated Hospital of Guangxi University of Chinese Medicine,Nanning 530011,Guangxi Zhuang Autonomous Region,China)

机构地区：[1]广西中医药大学第一附属医院四肢骨伤科,广西壮族自治区南宁市530023 [2]广西中医药大学附属瑞康医院老年病科,广西壮族自治区南宁市530011

出　　处：《中国组织工程研究》2024年第21期3431-3437,共7页Chinese Journal of Tissue Engineering Research

基　　金：国家自然科学基金(82160912),项目负责人:段戡;国家自然科学基金(82060875),项目负责人:袁长深;2018年广西一流学科建设项目重点课题(2018xk074),项目负责人:段戡。

摘　　要：背景:缺氧与骨关节炎软骨细胞损伤的发生、发展密切相关,但具体作用靶点及调控机制尚不清楚。目的:运用机器学习方法鉴定KDEL(Lys-Asp-Glu-Leu)受体3(KDELR3)作为骨关节炎缺氧的特征基因及免疫浸润分析,以期为骨关节炎的治疗提供新的思路与方法。方法:从GEO数据库下载骨关节炎相关的数据集和GSEA网站中获取缺氧相关基因;采用R语言对骨关节炎数据集进行批次校正及免疫浸润分析,并提取骨关节炎缺氧基因进行差异分析,对差异表达基因进行GO功能及KEGG信号通路分析;同时运用加权基因共表达网络分析(Weighted correlation network analysis,WGCNA)及机器学习筛选骨关节炎缺氧的特征基因,并进行体外细胞实验,运用数据集及qPCR验证表达并行相关免疫浸润分析。结果与结论:①经批次校正及主成分分析获得骨关节炎基因8492个,主要与Macrophages M2和Mast cells resting等免疫细胞密切相关;同时获得缺氧基因200个,进而得到41个骨关节炎缺氧差异表达基因。②GO分析主要涉及对营养水平、糖皮质激素反应等生物过程;涉及溶酶体腔、高尔基内腔等细胞组分;涉及14-3-3蛋白结合、DNA结合转录激活剂活性等分子功能。③KEGG分析骨关节炎缺氧差异表达基因与PI3K-Akt、FoxO及癌症中的微小RNA等信号通路有关。④运用WGCNA分析及机器学习筛选后获得特征基因KDELR3。⑤通过基因芯片验证后发现KDELR3基因在滑膜中实验组基因表达高于对照组(P=0.014),而半月板中实验组基因的表达却低于对照组(P=0.024)。⑥体外软骨细胞实验显示KDELR3基因在软骨中实验组表达高于对照组(P=0.005),同时KDELR3基因与Macrophages M0(P=0.014),T cells follicular helper(P=0.014)等密切相关。运用机器学习方法证实KDELR3可作为骨关节炎缺氧特征基因,可能通过改善缺氧来干预骨关节炎发病,期待能为更好地治疗骨关节炎提供新方向。BACKGROUND:Hypoxia is strongly associated with the development and progression of osteoarthritic chondrocyte injury,but the specific targets and regulatory mechanisms are unclear.OBJECTIVE:A machine learning approach was used to identify KDEL(Lys-Asp-Glu-Leu)receptor 3(KDELR3)as a characteristic gene for osteoarthritis hypoxia and immune infiltration analysis,to provide new ideas and methods for the treatment of osteoarthritis.METHODS:The osteoarthritis-related datasets were downloaded from the GEO database and the GSEA website to obtain hypoxia-related genes.The osteoarthritis datasets were batch-corrected and immune infiltration analyzed using R language,and osteoarthritis hypoxia genes were extracted for differential analysis.Differentially expressed genes were analyzed for GO function and KEGG signaling pathway.Weighted correlation network analysis(WGCNA)and machine learning were also used to screen osteoarthritis hypoxia signature genes,and in vitro cellular experiments were performed to validate expression and correlate immune infiltration analysis using the datasets and qPCR.RESULTS AND CONCLUSION:(1)8492 osteoarthritis genes were obtained by batch correction and principal component analysis,mainly strongly associated with immune cells such as Macrophages M2 and Mast cells resting;200 hypoxia genes were also obtained,resulting in 41 osteoarthritis hypoxia differentially expressed genes.(2)GO analysis involved mainly biological processes such as response to nutrient levels and glucocorticoids;cellular components such as lysosomal lumen and Golgi lumen;and molecular functions such as 14-3-3 protein binding and DNA-binding transcriptional activator activity.(3)KEGG analysis of osteoarthritis hypoxia differentially expressed genes was associated with signaling pathways such as PI3K-Akt,FoxO,and microRNAs in cancer.(4)The characteristic gene KDELR3 was obtained after using WGCNA analysis and machine learning screening.(5)The gene expression of KDELR3 was found to be higher in the test group than in the control

关 键 词：骨关节炎 机器学习 缺氧 特征基因 软骨细胞 生物标志物 免疫浸润分析 

分 类 号：R452[医药卫生—治疗学] R318[医药卫生—临床医学] R364