使用加权基因共表达网络分析（WGCNA）和机器学习鉴定甲状腺癌生物标志物

Identification of thyroid cancer biomarkers using WGCNA and machine learning.

作者信息

Hu Gaofeng, Niu Wenyuan, Ge Jiaming, Xuan Jie, Liu Yanyang, Li Mengjia, Shen Huize, Ma Shang, Li Yuanqiang, Li Qinglin

机构信息

Wenzhou Medical University, Wenzhou, Zhejiang, China.

Zhejiang Cancer Hospital, Hangzhou, Zhejiang, China.

出版信息

Eur J Med Res. 2025 Apr 5;30(1):244. doi: 10.1186/s40001-025-02466-x.

DOI:10.1186/s40001-025-02466-x

PMID:40186253

原文链接:https://pmc.ncbi.nlm.nih.gov/articles/PMC11971869/

Abstract

OBJECTIVE

The incidence of thyroid cancer (TC) is increasing in China, largely due to overdiagnosis from widespread screening and improved ultrasound technology. Identifying precise TC biomarkers is crucial for accurate diagnosis and effective treatment.

METHODS

TC patient data were obtained from TCGA. DEGs were analyzed using DESeq2, and WGCNA identified gene modules associated with TC. Machine learning algorithms (XGBoost, LASSO, RF) identified key biomarkers, with ROC and AUC > 0.95 indicating strong diagnostic performance. Immune cell infiltration and biomarker correlation were analyzed using CIBERSORT.

RESULTS

Four key genes (P4HA2, TFF3, RPS6KA5, EYA1) were found as potential biomarkers. High P4HA2 expression was associated with suppressed anti-tumor immune responses and promoted disease progression. In vitro studies showed that P4HA2 upregulation increased TC cell growth and migration, while its suppression reduced these activities.

CONCLUSION

Through bioinformatics and experimental validation, we identified P4HA2 as a key potential thyroid cancer biomarker. This finding provides new molecular targets for diagnosis and treatment. P4HA2 has the potential to be a diagnostic or therapeutic target, which could have significant implications for improving clinical outcomes in thyroid cancer patients.

摘要

目的

在中国，甲状腺癌（TC）的发病率正在上升，这在很大程度上归因于广泛筛查和超声技术改进导致的过度诊断。确定精确的TC生物标志物对于准确诊断和有效治疗至关重要。

方法

从TCGA获取TC患者数据。使用DESeq2分析差异表达基因（DEG），并通过加权基因共表达网络分析（WGCNA）确定与TC相关的基因模块。机器学习算法（XGBoost、套索回归、随机森林）确定关键生物标志物，受试者工作特征曲线（ROC）和曲线下面积（AUC）>0.95表明具有强大的诊断性能。使用CIBERSORT分析免疫细胞浸润和生物标志物相关性。