| 乳酸化相关基因在结直肠癌奥沙利铂化疗响应中的预测价值 |
| 投稿时间：2025-09-03 修订日期：2025-09-22 点此下载全文 |
| 引用本文：杨晔宏,张轵贻,孔杰,杨俊涛.乳酸化相关基因在结直肠癌奥沙利铂化疗响应中的预测价值[J].医学研究杂志,2025,54(11):135-140, 170 |
| DOI：
10.11969/j.issn.1673-548X.2025.11.024 |
| 摘要点击次数: 38 |
| 全文下载次数: 20 |
|
| 基金项目:中国医学科学院医学与健康科技创新工程基金资助项目(CIFMS2021-I2M-1-057,CIFMS2022-I2M-2-001) |
|
| 中文摘要:目的 探讨乳酸化相关基因集合对结直肠癌患者奥沙利铂化疗敏感度的预测价值。方法 从基因表达数据集(Gene Expression Omnibus,GEO)数据库选取3个独立结直肠癌奥沙利铂化疗应答数据集(GSE28702、GSE19860、GSE69657)。整合乳酸化修饰机制相关的5大功能通路(核心乳酸代谢、乳酸化修饰系统、乳酸信号通路、耐药机制、表观调控),构建包含70个关键基因的综合集合。在每个数据集独立筛选差异表达基因,取其与乳酸化相关基因集合的交集作为特征基因,采用多种机器学习算法建模,通过特征选择算法和交叉验证确定最优模型。观察模型的准确率、曲线下面积(area under the curve, AUC)、敏感度和特异性。结果 3个数据集的预测模型均表现出稳定性能。在GSE69657数据集中,逻辑回归和岭回归模型性能最优(准确率为80.00%,AUC为0.7250)；在GSE28702数据集中,线性判别分析准确率为65.00%,AUC为0.6792；在GSE19860数据集中,极端梯度提升树准确率为72.50%,AUC为0.7733。分析发现,不同数据集的最优特征集合存在明显异质性,但均位于乳酸化相关通路框架内,分别凸显了免疫微环境调控、代谢-表观遗传交叉对话和药物转运-炎性反应等不同生物学模块在驱动耐药中的重要作用。结论 乳酸化相关基因集合可有效预测结直肠癌奥沙利铂化疗的敏感度,其整体功能状态是预测化疗响应的稳健基础,但具体分子实现方式因患者群体而异,提示存在多种分子亚型,值得多中心、大样本量研究进一步验证。 |
| 中文关键词:乳酸化相关基因 结直肠癌 化疗敏感度 机器学习预测 |
| |
| Predictive Value of Lactylation-related Genes in the Response to Oxaliplatin Chemotherapy in Colorectal Cancer. |
|
|
| Abstract:Objective To explore the predictive value of the set derived from lactylation-related genes for oxaliplatin chemosensitivity in patients with colorectal cancer (CRC). Methods Three independent CRC oxaliplatin response datasets (GSE28702, GSE19860, GSE69657) were obtained from the Gene Expression Omnibus (GEO) database. Five functional pathways related to lactate modification (core lactate metabolism, lactate modification system, lactate signaling pathway, drug resistance mechanisms, and epigenetic regulation) was integrated to construct a comprehensive set of 70 key genes. Differentially expressed gene (DEG) were independently screened in each dataset, and their intersection with the lactylation-related gene set was used as feature gene. Multiple machine learning algorithms are adopted for modeling, with optimal models selected via feature selection algorithms and cross-validation. The accuracy rate, area under the curve (AUC), sensitivity, and specificity were observed. Results Predictive models demonstrated consistent performance across all three datasets. In GSE69657 dataset, the logistic regression and ridge regression models had the best performance (with an accuracy rate of 80.00% and an AUC of 0.7250). In GSE28702 dataset, linear discriminant analysis attained an accuracy rate of 65.00% and an AUC of 0.6792. In GSE19860 dataset, random forest achieved an accuracy rate of 72.50%, while XGBoost attained an AUC of 0.7733. Analysis revealed significant heterogeneity in the optimal feature sets across different datasets. However, all features resided within the framework of lactylation-related pathways, highlighting the important roles of various biological modules such as immune microenvironment regulation, metabolism-epigenetic crosstalk, and drug transport-inflammatory response in driving drug resistance. Conclusion The lactylation-related gene set can effectively predict oxaliplatin chemosensitivity for CRC. The overall functional state of this gene set provides a robust basis for predicting chemotherapy response. However, the specific molecular mechanisms involved vary across patient cohorts, suggesting the existence of distinct molecular subtypes. These findings warrant further validation through multi-center, large-sample studies. |
| keywords:Lactylation-related gene Colorectal cancer Chemotherapy sensitivity Machine learning prediction |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
