Table_3_Integrative analysis of DNA methylation and gene expression data for the diagnosis and underlying mechanism of Parkinson’s disease.XLSX

BackgroundParkinson’s disease (PD) is the second most common progressive neurodegenerative disorder and the leading cause of disability in the daily activities. In the management of PD, accurate and specific biomarkers in blood for the early diagnosis of PD are urgently needed. DNA methylation is one of the main epigenetic mechanisms and associated with the gene expression and disease initiation of PD. We aimed to construct a methylation signature for the diagnosis of PD patients, and explore the potential value of DNA methylation in therapeutic options.
Materials and methodsWhole blood DNA methylation and gene expression data of PD patients as well as healthy controls were extracted from Gene Expression Omnibus database. Next, differentially expressed genes (DEGs) and differentially methylated genes (DMGs) between PD patients and healthy controls were identified. Least absolute shrinkage and selection operator cox regression analysis was carried out to construct a diagnostic signature based on the overlapped genes. And, the receiver operating characteristic (ROC) curves were drawn and the area under the curve (AUC) was used to assess the diagnostic performance of the signature in both the training and testing datasets. Finally, gene ontology and gene set enrichment analysis were subsequently carried out to explore the underlying mechanisms.
ResultsWe obtained a total of 9,596 DMGs, 1,058 DEGs, and 237 overlapped genes in the whole blood between PD patients and healthy controls. Eight methylation-driven genes (HIST1H4L, CDC42EP3, KIT, GNLY, SLC22A1, GCM1, INO80B, and ARHGAP26) were identified to construct the gene expression signature. The AUCs in predicting PD patients were 0.84 and 0.76 in training dataset and testing dataset, respectively. Additionally, eight methylation-altered CpGs were also identified to construct the CpGs signature which showed a similarly robust diagnostic capability, with AUCs of 0.8 and 0.73 in training dataset and testing dataset, respectively.
ConclusionWe conducted an integrated analysis of the gene expression and DNA methylation data, and constructed a methylation-driven genes signature and a methylation-altered CpGs signature to distinguish the patients with PD from healthy controls. Both of them had a robust prediction power and provide a new insight into personalized diagnostic and therapeutic strategies for PD.

背景
帕金森病（Parkinson’s disease, PD）是第二大常见的进行性神经退行性疾病，也是导致日常活动功能障碍的首要诱因。在帕金森病的临床管理中，亟需能够用于早期诊断的精准特异性血液生物标志物。DNA甲基化是主要的表观遗传调控机制之一，与帕金森病的基因表达及疾病发生密切相关。本研究旨在构建用于帕金森病患者诊断的甲基化特征，并探究DNA甲基化在治疗方案选择中的潜在价值。

材料与方法
从基因表达综合数据库（Gene Expression Omnibus, GEO）中提取帕金森病患者与健康对照者的全血DNA甲基化及基因表达数据。随后，筛选出帕金森病患者与健康对照者之间的差异表达基因（differentially expressed genes, DEGs）与差异甲基化基因（differentially methylated genes, DMGs）。基于重叠基因，采用最小绝对收缩与选择算子（Least absolute shrinkage and selection operator, LASSO）Cox回归分析构建诊断特征模型。绘制受试者工作特征（receiver operating characteristic, ROC）曲线，以曲线下面积（area under the curve, AUC）评估特征模型在训练集与测试集中的诊断效能。最后，通过基因本体（gene ontology, GO）富集分析与基因集富集分析（gene set enrichment analysis, GSEA）探究潜在的分子机制。

结果
本研究共筛选得到帕金森病患者与健康对照者全血中的9596个差异甲基化基因、1058个差异表达基因，以及237个重叠基因。筛选出8个甲基化驱动基因（HIST1H4L、CDC42EP3、KIT、GNLY、SLC22A1、GCM1、INO80B、ARHGAP26）以构建基因表达特征模型。该模型在训练集与测试集中预测帕金森病患者的曲线下面积分别为0.84与0.76。此外，筛选得到8个甲基化改变的CpG位点以构建CpG特征模型，该模型同样具备优异的诊断效能，其在训练集与测试集中的曲线下面积分别为0.8与0.73。

结论
本研究对基因表达与DNA甲基化数据进行整合分析，构建了甲基化驱动基因特征模型与甲基化改变CpG位点特征模型，可有效区分帕金森病患者与健康对照者。两种特征模型均具备优异的预测性能，可为帕金森病的个性化诊断与治疗策略提供全新的研究思路。