基于转录组数据与机器学习的非小细胞肺癌放疗敏感性基因筛选及预后模型建立OA

Objective To identify key genes associated with radiotherapy prognosis in non-small cell lung cancer(NSCLC)and construct a risk prediction model using public datasets and machine learning algorithms.Methods Radiotherapy-relat-ed datasets(GSE197236,GSE253564)were obtained from the GEO database.Common radiation response genes(RRGs)were identified through differential expression analysis and functional enrichment.In the TCGA cohort of 175 NSCLC patients treated with radiotherapy,three machine learning algorithms were applied to screen overall survival(OS)-related RRGs.Prognostic value was assessed using survival analysis and multivariate Cox regression,and a risk score model was developed and validated.Results Fifty-eight differential RRGs were identified,mainly enriched in cell cycle and mitosis pathways.Six key RRGs were selected via three machine learning algorithms.High expression of BCL2L15,CENPE,KRT6A,and FGA was associated with better OS,while MSMB and HIST1H3G indicated worse OS(P＜0.05).Multivariate analysis results showed that FGA,BCL2L15,and KRT6A were independent prognostic factors(P＜0.05).The risk model demonstrated good predictive performance,with AUC of 0.745,0.752,and 0.706 for 1-year,3-year,and 5-year OS,respectively.Conclusion This study identified six radiotherapy-related genes and established a validated prognostic model,providing a potential tool for assessing radiosensitivity and guiding personalized treatment in NSCLC.