广西医学

目的基于生物信息学挖掘银屑病的自噬和免疫相关特征基因，并利用机器学习及循环算法构建银屑病的预测模型。方法（1）从GEO数据库下载与银屑病相关的基因表达数据集（GSE117239、GSE30999和GSE41662）。从ImmPort数据库获取免疫相关基因(IRGs)，从 HADb数据库、HAMdb数据库和GSEA数据库获取自噬相关基因(ATGs)。（2）针对GSE117239数据集，通过差异表达分析获取差异表达基因(DEGs)，通过免疫细胞浸润特征分析获取差异免疫细胞，通过加权基因共表达网络分析获取差异免疫细胞相关基因模块和银屑病相关基因模块，通过最小绝对值收敛和选择算子算法获取自噬和免疫相关特征基因。针对自噬和免疫相关特征基因进行基因本体论（GO）功能富集分析和京都基因与基因组百科全书（KEGG）通路富集分析。（3）基于自噬和免疫相关特征基因，采用循环算法构建银屑病预测模型。绘制受试者工作特征曲线评估该模型的预测效能，并利用GSE30999和GSE41662数据集进行验证。结果（1）共筛选出1 071个DEGs和9种差异免疫细胞。（2）将银屑病相关基因模块包含的基因、差异免疫细胞相关基因模块包含的基因分别与IRGs和DEGs、ATGs和DEGs取交集，得到95个Hub IRGs和29个Hub ATGs。分别在Hub IRGs和Hub ATGs中获得42个免疫相关特征基因和14个自噬相关特征基因。GO功能富集分析结果显示，免疫和自噬相关特征基因分别涉及814个、706个GO条目；KEGG通路富集分析结果显示，免疫相关特征基因主要富集在趋化因子信号通路、Janus激酶-信号转导和转录激活因子信号通路和Th17分化等信号通路，自噬相关特征基因主要富集在白细胞介素17信号通路、肿瘤坏死因子信号通路、流体剪切应力与动脉粥样硬化、NOD样受体信号通路等信号通路。（3）循环算法结果显示，基于包含14个基因（S100A8、CCL2、EGF、WNT5A、RORC、TNFRSF4、PAK3、TRBC1、PRKCQ、MTCL1、SVIP、LAMP3、SPTLC2、PDK4）的基因组合构建的银屑病预测模型具有较高预测性能。结论42个免疫相关特征基因和14个自噬相关特征基因与银屑病密切相关，其中包含S100A8、CCL2、EGF、WNT5A、RORC、TNFRSF4、PAK3、TRBC1、PRKCQ、MTCL1、SVIP、LAMP3、SPTLC2、PDK4的基因组合对银屑病具有较高的预测性能，这14个基因有望成为诊断银屑病的潜在生物标志物。

ObjectiveTo extract autophagy- and immune-related identity genes of psoriasis based on bioinformatics, and to establish the prediction model of psoriasis by employing machine learning and round-robin algorithm. Methods(1) Gene expression datasets (GSE117239, GSE30999, and GSE41662) related to psoriasis were downloaded from the GEO database. Immune-related genes (IRGs) were obtained from the ImmPort database, and autophagy-related genes (ATGs) were obtained from the HADb, HAMdb, and GSEA databases. (2) For GSE117239 dataset, differentially expressed genes (DEGs) were obtained by differentially expressed analysis, differentially immune cells were acquired by characteristic analysis of immune cell infiltration, and modules of differentially immune cell-related genes and psoriasis-related genes were acquired by analysis of weighted gene co-expression network, as well as autophagy- and immune-related identity genes were obtained by the least absolute shrinkage and selection operator algorithm. The Gene Ontology (GO) functional enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis were performed on autophagy- and immune-related identity genes. (3) The prediction model of psoriasis was established by using round-robin algorithm based on autophagy- and immune-related identity genes. The predictive efficiency of this model was evaluated by drawing the receiver operating characteristic curve, and was validated by employing GSE30999 and GSE41662 datasets. Results(1) A total of 1071 DEGs and 9 categories of differentially immune cells were screened. (2) Intersections of contained genes from psoriasis-related gene module and differentially immune cell-related gene module with IRGs and DEGs, ATGs and DEGs were obtained, respectively, and 95 Hub IRGs and 29 Hub ATGs were acquired. A total of 42 immune-related identity genes and 14 autophagy-related identity genes were obtained from Hub IRGs and Hub ATGs, respectively. The results of GO functional enrichment analysis revealed that immune- and autophagy-related identity genes involved 814 and 706 GO catalogues, respectively. The results of KEGG pathway enrichment analysis interpreted that immune-related identity genes were mainly enriched in chemokine signaling pathway, Janus kinase-signal transducer and activator of transcription signaling pathway, Th17 differentiation signaling pathway, etc., and autophagy-related identity genes were mainly enriched in interleukin 17 signaling pathway, tumor necrosis factor signaling pathway, fluid shear stress and atherosclerosis, NOD-like receptor signaling pathway, etc. (3) The results of round-robin algorithm indicated that the prediction model of psoriasis established based on gene combinations containing 14 genes (S100A8, CCL2, EGF, WNT5A, RORC, TNFRSF4, PAK3, TRBC1, PRKCQ, MTCL1, SVIP, LAMP3, SPTLC2, PDK4) exerted relatively high predictive efficiency. ConclusionForty-two immune-related identity genes and 14 autophagy-related identity genes are closely related to psoriasis, therein, gene combinations containing S100A8, CCL2, EGF, WNT5A, RORC, TNFRSF4, PAK3, TRBC1, PRKCQ, MTCL1, SVIP, LAMP3, SPTLC2, PDK4 exert relatively high predictive efficiency on psoriasis, and these 14 genes may be regarded as potential biomarkers for diagnosing psoriasis.