Wang Junhao, Zhou Yan, Hu Jie, Han Jianpeng, Feng Jianyong, Guo Kuo, Chen Wenbin, Yun Yanrui, Li Yongzhang
Department of Urology, Hebei Provincial Hospital of Traditional Chinese Medicine, Shijiazhuang, Hebei Province, China.
Department of Urology, Langfang People's Hospital, Langfang City, Hebei Province, China.
PLoS One. 2025 May 28;20(5):e0320249. doi: 10.1371/journal.pone.0320249. eCollection 2025.
Interstitial cystitis (IC) was still a disease with the exclusive diagnosis and lacked an effective gold standard. It was of great significance to find diagnostic markers for IC. Our study was aimed to screen characteristic genes via machine learning algorithms, characterize the immune landscape of IC, and show correlations between characteristic genes and immune cell subtypes.
RNA sequencing data sets on IC were downloaded from Gene Expression Omnibus (GEO) database, including GSE57560, GSE11783 and GSE621, whose corresponding platforms were GPL16699, GPL570 and GPL262 respectively. Three machine learning algorithms were applied for identification of characteristic gene for IC. Single sample Gene Set Enrichment Analysis (ssGSEA) was applied to figure out the immune cell infiltration (ICI) of IC and normal tissue samples. Correlation analysis was performed via Spearman test. Receiver operator characteristic curve (ROC) was used to evaluate diagnostic efficacy of key genes.
CCL18, MMP10 and WIF1 were identified as characteristic gene via machine learning algorithms. MMP10 and CCL18 were with higher expression in IC tissues compared with normal bladder tissues, while WIF1 had lower expressionin IC tissues (P < 0.05). These three genes had good diagnostic efficacy for IC. Compared with normal bladder tissues, 18 immune cell subtypes were up-regulated in interstitial cystitis tissues (P < 0.05). MMP10 and CCL18 were positively correlated to immune scores in IC, while WIF1 was negatively correlated to immune scores (P > 0.05).
We screened the feature genes, CCL18, MMP10 and WIF1, among the differentially expressed genes (DEGs) by three different machine learning algorithms. They showed good diagnostic performance in both training and testing cohorts and were potential diagnostic markers for IC. We paint the immune landscape of IC. In IC tissue, immune cell subtypes infiltrated extensively. Most immune cell subtypes were up-regulated in IC tissue, including mast cells, activated CD4 T cells, and regulatory T cells that suppress immune responses. MMP10 and CCL18 had positive correlation to ICI, while WIF was negatively correlated with ICI. MMP10 and CCL18 may be the driving factors of immune response or their expression levels may be increased by immune response. The effect of characteristic genes of IC on immune cell subtypes still needed to be further explored.
间质性膀胱炎（IC）仍是一种依赖排他性诊断的疾病，缺乏有效的金标准。寻找IC的诊断标志物具有重要意义。我们的研究旨在通过机器学习算法筛选特征基因，描绘IC的免疫图谱，并展示特征基因与免疫细胞亚型之间的相关性。
从基因表达综合数据库（GEO）下载IC的RNA测序数据集，包括GSE57560、GSE11783和GSE621，其相应平台分别为GPL16699、GPL570和GPL262。应用三种机器学习算法鉴定IC的特征基因。采用单样本基因集富集分析（ssGSEA）来确定IC和正常组织样本中的免疫细胞浸润情况。通过Spearman检验进行相关性分析。使用受试者工作特征曲线（ROC）评估关键基因的诊断效能。
通过机器学习算法鉴定出CCL18、MMP10和WIF1为特征基因。与正常膀胱组织相比，MMP10和CCL18在IC组织中表达较高，而WIF1在IC组织中表达较低（P < 0.05）。这三个基因对IC具有良好的诊断效能。与正常膀胱组织相比，间质性膀胱炎组织中有18种免疫细胞亚型上调（P < 0.05）。MMP10和CCL18与IC中的免疫评分呈正相关，而WIF1与免疫评分呈负相关（P > 0.05）。
我们通过三种不同的机器学习算法在差异表达基因（DEGs）中筛选出特征基因CCL18、MMP10和WIF1。它们在训练和测试队列中均表现出良好的诊断性能，是IC的潜在诊断标志物。我们描绘了IC的免疫图谱。在IC组织中，免疫细胞亚型广泛浸润。大多数免疫细胞亚型在IC组织中上调，包括肥大细胞、活化的CD4 T细胞和抑制免疫反应的调节性T细胞。MMP10和CCL18与免疫细胞浸润呈正相关，而WIF与免疫细胞浸润呈负相关。MMP10和CCL18可能是免疫反应的驱动因素，或者它们的表达水平可能因免疫反应而升高。IC特征基因对免疫细胞亚型的影响仍需进一步探索。
