目的 通过生物信息学多芯片联合分析方法和机器学习技术探讨复发性阿弗他溃疡(recurrent aphthous ulcer,RAU)免疫特征基因在溃疡性结肠炎(ulcerative colitis,UC)中的临床价值.方法 从基因表达综合(Gene Expres-sion Omnibus,GEO)数据库下载RAU和UC的转录组数据,免疫相关基因列表从免疫学数据库和分析门户(Immunolo-gy Database and Analysis Portal,IMMPORT)获得.利用加权基因共表达网络分析(weighted gene co-expression net-work analysis,WGCNA)、差异分析和功能富集分析筛选评估RAU疾病基因及其功能,通过LASSO-Cox回归和随机森林筛选RAU免疫特征基因.验证RAU免疫特征基因在UC样本中的表达水平并进行Friends分析,通过多层感知机(multi-layer perception,MLP)人工神经网络鉴定其诊断效能.基于免疫特征基因使用一致性聚类算法对UC样本进行分子分型,探讨不同亚型免疫细胞浸润水平和生物学特征.最后利用药物转录图谱平台(Integrated Traditional Chinese Medicine,ITCM)进行靶向药物筛选,CB-dock2和SwissADME平台进行分子对接验证和模拟药物代谢动力学分析.结果 共鉴定出324个RAU疾病基因.生物功能富集分析显示,RAU疾病基因主要与免疫炎症、微生物感染和炎症性肠病等危险因素有关.进一步筛选验证确定防御素β4A(defensin beta 4A,DEFB4A)、骨髓基质细胞抗原2(bone marrow stromal cell antigen 2,BST2)、胸苷磷酸化酶(thymidine phosphorylase,TYMP)和神经胶质成熟因子 γ(glia maturation factor gamma,GMFG)为高表达免疫特征基因,并对UC具有诊断价值[曲线下面积(area under the curve,AUC)=0.915].RAU和UC存在相同的T细胞紊乱基础,免疫特征基因可将UC区分为代谢型和免疫型两个亚型,且免疫型具有更高的特征基因表达和T细胞亚群浸润水平.筛选获得10个天然小分子药物,分子对接显示其均与免疫特征基因翻译蛋白质有良好结合活性,且湖贝甲素具有良好的药代动力学参数,可能是潜在的UC防治药物.结论 基于RAU的免疫特征基因可以很好地区分UC样本并进行亚型鉴定,可为消化道炎症性疾病基础研究、临床疾病辅助筛查和潜在防治药物筛选提供一定理论依据和新思路.
Identification of Ulcerative Colitis Biomarkers and Potential Small Molecule Drug Screening Based on Immunosignature Genes of Recurrent Aphthous Ulcers
Objective To explore the clinical significance of immunosignature genes of recurrent aphthous ulcers(RAU)in ulcerative colitis(UC)through bioinformatics multi-chip joint analysis and machine learning techniques.Meth-ods Transcriptome data of RAU and UC were downloaded from the Gene Expression Omnibus(GEO)database,and a list of immune-related genes was obtained from the Immunology Database and Analysis Portal(IMMPORT).Weighted gene co-expression network analysis(WGCNA),differential analysis and functional enrichment analysis were used to screen and evaluate RAU disease genes and its functions,the RAU immunosignature genes were screened by LASSO-Cox regression and random forest.The expression levels of RAU immunosignature genes in UC samples were validated and subjected to Friends analysis,the diagnostic efficiency of these genes was identified by a multi-layer perception(MLP)artificial neural network.Based on im-munosignature genes,a consistency clustering algorithm was employed to classify UC samples molecularly,explo-ring different subtypes'levels of immune cell infiltration and biological features.Finally,targeted drug screening was per-formed by the Integrated Traditional Chinese Medicine(ITCM),and molecular docking validation and simulated pharma-cokinetic analysis were conducted by CB-dock2 and SwissADME platforms.Results A total of 324 disease genes related to RAU were identified.Functional enrichment analysis indicated that RAU disease genes were mainly associated with risk factors like immune inflammation,microbial infections and inflammatory bowel disease etc.Further screening,vali-dating and identifying defensin beta 4A(DEFB4A),bone marrow stromal cell antigen 2(BST2),thymidine phosphoryl-ase(TYMP)and glia maturation factor gamma(GMFG)as highly expressed immune characteristic genes,which demon-strating diagnostic value for UC[area under the curve(AUC)=0.915].RAU and UC shared a common T-cell disorder foundation,immunosignature genes could differentiate UC into metabolic and immune subtypes,with the immune subtype exhibiting higher expression of signature genes and levels of T-cell subset infiltration.Ten natural small molecule drugs were screened,and molecular docking showed good binding activity with the translation proteins of immune characteristic genes,moreover,hupehenine demonstrated favorable pharmacokinetic parameters,which might become potential UC therapeutic drugs.Conclusion Immunosignature genes based on RAU can effectively differentiate UC samples and identi-fy subtypes,which providing a theoretical basis and new perspectives for basic research on digestive inflammatory disea-ses,clinical disease auxiliary screening,potential preventive and therapeutic drug screening.
NETL
NSTL
万方数据
