通过113组合机器学习框架鉴定细胞衰老相关基因以预测肺腺癌的诊断、预后和免疫治疗反应

Identification of cellular senescence-associated genes for predicting the diagnosis, prognosis and immunotherapy response in lung adenocarcinoma via a 113-combination machine learning framework.

作者信息

Ge Ting, He Guixin, Cui Qian, Wang Shuangcui, Wang Zekun, Xie Yingying, Tian Yuanyuan, Zhou Juyue, Yu Jianchun, Hu Jinmin, Li Wentao

机构信息

Central Laboratory, First Teaching Hospital of Tianjin University of Traditional Chinese Medicine, Tianjin, China.

Graduate School of Tianjin University of Traditional Chinese Medicine, Tianjin, China.

出版信息

Discov Oncol. 2025 Apr 1;16(1):440. doi: 10.1007/s12672-025-02262-3.

DOI:10.1007/s12672-025-02262-3

PMID:40169440

原文链接:https://pmc.ncbi.nlm.nih.gov/articles/PMC11961801/

Abstract

BACKGROUND

Lung adenocarcinoma (LUAD) is a prevalent malignant tumor of the respiratory system, with high incidence and mortality rates. Cellular senescence (CS) widely affects the tumor microenvironment (TME) and tumor growth, and is related to the invasion and immune escape of tumor cells. This study aims to develop a robust CS-related signature of LUAD.

METHODS

Using the GSE140797, GSE42458, GSE75037, and GSE85841 datasets, in combination with cellular senescence databases, 75 LUAD CS-related differentially expressed genes (LUAD-CSDEGs) were identified through the weighted gene co-expression network analysis (WGCNA) method. Subsequently, we developed a novel machine learning framework that incorporated 12 machine learning algorithms and their 113 combinations to construct a LUAD CS-related signature (LUAD-CSRS), which were assessed in both training and validation cohorts. A LUAD-CSRS-integrated nomogram was constructed to provide a quantitative tool for predicting prognosis in clinical practice. Finally, the difference of immune infiltration and response to immunotherapy in patients with high and low risk of LUAD were evaluated.

RESULTS

Based on a 113-combination machine learning framework, we finally identified a LUAD-CSRS containing eight genes: RECQL4, TIMP1, ANLN, SFN, MDK, KIF2C, AGR2, ITGB4. We also confirmed that it was significantly associated with survival, immune cell infiltration, prognosis, and response to immunotherapy in LUAD patients. Additionally, we found it is related to the activation of immune responses and may be involved in regulating the balance between immune cells in the TME.

CONCLUSION

In summary, our study constructed a novel LUAD-CSRS, which is not only expected to be a powerful tool for assisting diagnosis and prognosis evaluation of LUAD, but also may provide guidance for personalized immunotherapy programs.

摘要

背景

肺腺癌（LUAD）是呼吸系统中一种常见的恶性肿瘤，发病率和死亡率都很高。细胞衰老（CS）广泛影响肿瘤微环境（TME）和肿瘤生长，并且与肿瘤细胞的侵袭和免疫逃逸有关。本研究旨在开发一种强大的与肺腺癌相关的细胞衰老特征。

方法

使用GSE140797、GSE42458、GSE75037和GSE85841数据集，结合细胞衰老数据库，通过加权基因共表达网络分析（WGCNA）方法鉴定出75个与肺腺癌细胞衰老相关的差异表达基因（LUAD-CSDEGs）。随后，我们开发了一种新颖的机器学习框架，该框架纳入了12种机器学习算法及其113种组合，以构建与肺腺癌细胞衰老相关的特征（LUAD-CSRS），并在训练和验证队列中进行评估。构建了一个整合了LUAD-CSRS的列线图，为临床实践中的预后预测提供了一种定量工具。最后，评估了肺腺癌高低风险患者免疫浸润的差异以及对免疫治疗的反应。

结果

基于一个113种组合的机器学习框架，我们最终鉴定出一个包含8个基因的LUAD-CSRS：RECQL4、TIMP1、ANLN、SFN、MDK、KIF2C、AGR2、ITGB4。我们还证实它与肺腺癌患者的生存、免疫细胞浸润、预后以及对免疫治疗的反应显著相关。此外，我们发现它与免疫反应的激活有关，并且可能参与调节TME中免疫细胞之间的平衡。