METHODS: Conjunctival cell samples were obtained by aspiration from patients with DED-MGD (n = 27) and asymptomatic controls (n = 22). RNA was purified, converted to cDNA, preamplified and analyzed using the Gene Expression Human Immune V2 panel (NanoString), which includes 579 target and 15 housekeeping genes. A machine learning (ML) algorithm was applied to design a signature associated with DED-MGD.
RESULTS: Forty-five immune genes were found upregulated in DED-MGD vs. controls, involved in eight signaling pathways, IFN I/II, MHC class I/II, immunometabolism, B cell receptor, T Cell receptor, and T helper-17 (Th-17) differentiation. Additionally, statistically significant correlations were found between 31 genes and clinical characteristics of the disease such as lid margin or tear osmolarity (Pearson\'s r < 0.05). ML analysis using a recursive feature elimination (RFE) algorithm selected a 4-gene mRNA signature that discriminated DED-MGD from control samples with an area under the ROC curve (AUC ROC) of 0.86 and an accuracy of 77.5%.
CONCLUSIONS: Multiplexed mRNA analysis of conjunctival cells can be used to analyze immune gene expression patterns in patients with DED-MGD and to generate diagnostic signatures.
方法:结膜细胞样本通过抽吸从DED-MGD患者(n=27)和无症状对照(n=22)获得。RNA被纯化,转化为cDNA,使用基因表达人类免疫V2面板(NanoString)进行预扩增和分析,其中包括579个靶基因和15个管家基因。应用机器学习(ML)算法来设计与DED-MGD相关联的签名。
结果:在DED-MGD与controls,涉及八种信号通路,IFNI/II,MHCI/II类,免疫代谢,B细胞受体,T细胞受体,和T辅助-17(Th-17)分化。此外,31个基因与眼睑边缘或泪液渗透压等疾病的临床特征之间存在统计学上的显着相关性(Pearson'sr<0.05)。使用递归特征消除(RFE)算法的ML分析选择了4基因mRNA标记,该标记将DED-MGD与对照样品区分开,ROC曲线下面积(AUCROC)为0.86,准确率为77.5%。
结论:结膜细胞的多重mRNA分析可用于分析DED-MGD患者的免疫基因表达模式并产生诊断特征。