基于机器学习筛选骨关节炎坏死性凋亡的特征基因RHOB及实验验证  

Machine learning-based screening of signature genes RHOB for necrotic apoptosis in osteoarthritis and experimental validation

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

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

出　　处：《生物骨科材料与临床研究》2024年第5期1-8,共8页Orthopaedic Biomechanics Materials and Clinical Study

基　　金：2023年广西中医药大学青年科学基金项目(2022QN012);2023年广西中医药大学第一附属医院青年科学基金项目(院字[2023]29号);国家自然科学基金(82160912);国家自然科学基金(82060875)。

摘　　要：目的 运用机器学习方法筛选骨关节炎坏死性凋亡的特征基因RHOB并加以验证,以期为OA治疗提供新的思路与方法。方法 从GEO数据库下载GSE55235、GSE1919、GSE82107、GSE98918微阵列数据集及GeneCard网站获取坏死性凋亡相关基因,采用R语言对OA数据进行批次校正,提取OA坏死性凋亡基因及差异分析,并对DEGs进行功能GO分析及KEGG信号通路分析,同时运用机器学习(LASSO回归、SVM-RFE、random Forest)筛选OA坏死性凋亡的特征基因,进一步运用荧光定量PCR试验验证。结果 经批次校正及PCA分析后共获得OA基因8 492个,同时获得坏死性凋亡相关基因657个,经分析后得到48个OA坏死性凋亡DEGs,其中上调基因18个,下调基因30个;主要涉及调节炎症反应、白细胞细胞间黏附等生物过程;涉及膜筏、膜微区等细胞组分;涉及细胞因子活性、整合素结合等分子功能;同时与TNF、IL-17、AGE-RAGE等信号通路有关。运用机器学习(LASSO回归、SVM-RFE、Random Forest)分别筛选出8个基因、11个基因及8个基因。其交集后获得特征基因RHOB,验证后发现RHOB作为OA坏死性凋亡特征基因准确性较高(AUC>0.5),且滑膜组织中试验组RHOB表达高于对照组(P=0.36),而半月板组织中试验组RHOB表达高于对照组(P=0.033)。同时运用体外软骨细胞培养及荧光定量PCR试验进一步证实,试验组RHOB mRNA的表达高于对照组(P=0.001)。结论 运用机器学习方法获得OA坏死性凋亡的特征基因及潜在因素,从新型细胞死亡角度为阐明其发病机制,为临床上更好治疗OA提供新方向。Objective A machine learning approach was used to screen and validate RHOB,a gene that characterizes necrotic apoptosis in osteoarthritis,with the aim of providing new ideas and methods for OA.Methods Genes related to necrotic apoptosis were obtained by downloading GSE55235,GSE1919,GSE82107,GSE98918 microarray datasets from the GEO database and GeneCard website,and the OA data were batch corrected,OA necrotic apoptosis genes extracted,and variance analyzed by using R.Functional GO analysis was performed on DEGs and KEGG signaling the DEGs were subjected to functional GO analysis and KEGG signaling pathway analysis,and machine learning(LASSO regression,SVM-RFE,random Forest)was applied to screen the characteristic genes of OA necrotic apoptosis,which were further validated by fluorescence quantitative PCR assay and analyzed by immune infiltration.Results A total of 8492 OA genes were obtained after batch correction and PCA analysis,and 657 necrotic apoptosis-related genes were obtained at the same time.Forty-eight OA necrotic apoptosis DEGs were obtained after analysis,including 18 upregulated genes and 30 downregulated genes;mainly involved in the regulation of inflammatory response,leukocyte intercellular adhesion and other biological processes;involved in the cellular components,such as membrane rafts,membrane microregions,and other cellular components;involved in cytokine activity,integrin binding and other molecular functions;also related to TNF,IL-17,AGE-RAGE signaling pathway and other signaling pathways.Machine learning(LASSO regression,SVM-RFE,Random Forest)was applied to screen 8 genes,11 genes and 8 genes.Respectively,the feature gene RHOB was obtained after intersection,and the validation revealed that RHOB was more accurate as the feature gene of necrotic apoptosis in OA(AUC>0.5),and at the same time,the expression of RHOB of the test group was higher than that of the control group in the synovial tissue(P=0.36),while the expression of RHOB in meniscus tissues was higher than that of the control group

关 键 词：机器学习 骨关节炎 坏死性凋亡 特征基因 

分 类 号：R318[医药卫生—生物医学工程]