现代医药卫生2025，Vol.41Issue(11)：2506-2514,9.DOI:10.3969/j.issn.1009-5519.2025.11.002
基于临床肿瘤样本DNA甲基化组的机器学习分类器预测放疗反应
Machine learning classifiers based on DNA methylation profiles of clinical tumor samples predict radiotherapy responses
摘要
Abstract
Objective To construct a machine learning model using DNA methylation features to pre-dict radiotherapy(RT)response in cancer patients.Methods By integrating and analyzing 10 types of cancer and whole-genome DNA methylation data alongside RT efficacy data from 843 patients,differentially methyla-ted sites(DMSs)significantly associated with radio sensitivity were identified.Machine learning classifier models were developed based on these characteristic DMSs.Results The machine learning classifier construc-ted based on methylation signals from multiple CpG sites(AUC=0.889-1.000)was significantly superior to a single DMS(AUC=0.594-0.956,P<0.001)in distinguishing between radio-sensitive and radio-resistant patients.The model could also accurately predict the overall survival period of patients.Conclusion The ma-chine learning model based on DNA methylome features has clinical application value in predicting the efficacy and prognosis of RT.关键词
放射治疗/DNA甲基化/机器学习分类器/癌症治疗/放疗敏感性预测Key words
Radiotherapy/DNA methylation/Machine learning classifiers/Cancer treatment/Prediction of radio-sensitivity分类
医药卫生引用本文复制引用
施小龙,唐超,吴颜,綦俊..基于临床肿瘤样本DNA甲基化组的机器学习分类器预测放疗反应[J].现代医药卫生,2025,41(11):2506-2514,9.基金项目
重庆市科卫联合医学科研面上项目(2024MSXM065 （）
2023GGXM002) （）
重庆市长寿区科技局项目(2024023CSKJ). （2024023CSKJ）
