Cai Lijun, Tang Shuang, Liu Yin, Zhang Yingwan, Yang Qin
Department of Pathophysiology, College of Basic Medical Sciences, Guizhou Medical University, Guiyang, Guizhou, China.
Department of Neurology, The Affiliated Hospital of Guizhou Medical University, Guiyang, Guizhou, China.
Front Mol Neurosci. 2023 Oct 16;16:1274268. doi: 10.3389/fnmol.2023.1274268. eCollection 2023.
This study aims to utilize Weighted Gene Co-expression Network Analysis (WGCNA) and Support Vector Machine (SVM) algorithm for screening biomarkers and constructing a diagnostic model for Parkinson's disease.
Firstly, we conducted WGCNA analysis on gene expression data from Parkinson's disease patients and control group using three GEO datasets (GSE8397, GSE20163, and GSE20164) to identify gene modules associated with Parkinson's disease. Then, key genes with significantly differential expression from these gene modules were selected as candidate biomarkers and validated using the GSE7621 dataset. Further functional analysis revealed the important roles of these genes in processes such as immune regulation, inflammatory response, and cell apoptosis. Based on these findings, we constructed a diagnostic model by using the expression data of FLT1, ATP6V0E1, ATP6V0E2, and H2BC12 as inputs and training and validating the model using SVM algorithm.
The prediction model demonstrated an AUC greater than 0.8 in the training, test, and validation sets, thereby validating its performance through SMOTE analysis. These findings provide strong support for early diagnosis of Parkinson's disease and offer new opportunities for personalized treatment and disease management.
In conclusion, the combination of WGCNA and SVM holds potential in biomarker screening and diagnostic model construction for Parkinson's disease.
本研究旨在利用加权基因共表达网络分析（WGCNA）和支持向量机（SVM）算法筛选生物标志物并构建帕金森病诊断模型。
首先，我们使用三个基因表达综合数据库（GEO）数据集（GSE8397、GSE20163和GSE20164）对帕金森病患者和对照组的基因表达数据进行WGCNA分析，以识别与帕金森病相关的基因模块。然后，从这些基因模块中选择具有显著差异表达的关键基因作为候选生物标志物，并使用GSE7621数据集进行验证。进一步的功能分析揭示了这些基因在免疫调节、炎症反应和细胞凋亡等过程中的重要作用。基于这些发现，我们以FLT1、ATP6V0E1、ATP6V0E2和H2BC12的表达数据作为输入，使用SVM算法构建诊断模型，并对该模型进行训练和验证。
预测模型在训练集、测试集和验证集中的曲线下面积（AUC）均大于0.8，从而通过合成少数类过采样技术（SMOTE）分析验证了其性能。这些发现为帕金森病的早期诊断提供了有力支持，并为个性化治疗和疾病管理提供了新的机会。
总之，WGCNA和SVM的结合在帕金森病生物标志物筛选和诊断模型构建方面具有潜力。
