【摘要】 目的 基于机器学习构建肺腺癌中嘧啶代谢相关基因的预后风险模型，并探索其免疫表型特征及药物敏感性，旨在寻求新的治疗策略。方法 从癌症基因组图谱和高通量基因表达数据库下载肺腺癌患者的基因表达谱，并从GeneCards数据库获取嘧啶代谢相关基因，筛选肺腺癌中嘧啶代谢相关的差异表达基因。用一致性聚类分析，探讨嘧啶相关基因在肺腺癌不同亚型中的表现。基于嘧啶代谢基因，用最小绝对收缩和选择算子回归构建预后风险评分模型，并在测试组中进行模型验证。依据风险评分将患者分为高、低风险组，进一步分析免疫相关特征及模型基因的蛋白表达。通过药物敏感性分析，评估常用化疗药物在不同风险组中的疗效。基于生物信息学分析，进一步研究模型基因环氧化物水解酶2(EPHX2)在肺腺癌中的表达谱及预后相关性，并通过人类蛋白图谱数据库和实时荧光定量聚合酶链反应技术进行验证。最后通过中医药证候关联数据库预测具有潜在靶向调控EPHX2作用的中药。结果 构建了基于嘧啶代谢基因的预后风险模型，识别出15个关键调控基因。高风险组患者的总体生存率显著低于低风险组(P<0.001)。在免疫分析中发现，低风险组的免疫评分[(1 934±760)vs(1 418±747)分]和基质评分[(600±645)vs(296±677)分]均显著高于高风险组(均P<0.001)。高风险组中静息自然杀伤细胞(NK)细胞和巨噬细胞浸润较多，低风险组中CD8~+T细胞和单核细胞浸润较多。药物敏感性分析显示，奥沙利铂和伊立替康对高风险组患者可能更有效(P<0.001)。生物信息学结果提示，EPHX2高表达与较好的预后相关(P<0.01)。HPA数据库的免疫组化分析证实了EPHX2的差异表达。EPHX2在正常肺上皮细胞中的高表达及在肺腺癌细胞中的低表达(16HBE vs H1299=1.00±0.12 vs 0.50±0.11,P<0.01)。中药预测表明，清热解毒、收涩理气类中药可能靶向调控EPHX2。结论 基于嘧啶代谢基因构建风险预后模型可作为评估肺腺癌预后的潜在工具，并为其诊疗提供新的思路。更多还原

【Abstract】 Objective To construct a prognostic risk model for pyrimidine metabolism-related genes in lung adenocarcinoma utilizing machine learning techniques,while also explore the immune phenotypic characteristics and drug sensitivity associated with these genes,thereby seeking novel therapeutic strategies.Methods Gene expression profiles from patients diagnosed with lung adenocarcinoma were retrieved from The Cancer Genome Atlas and Gene Expression Omnibus databases.Pyrimidine metabolism-related genes were sourced from the GeneCards database.Differentially expressed genes pertinent to pyrimidine metabolism in lung adenocarcinoma were identified.Consensus clustering analysis was employed to assess the expression patterns of pyrimidine-related genes across various subtypes of lung adenocarcinoma.A prognostic risk score model was developed based on these pyrimidine metabolism genes using Least Absolute Shrinkage and Selection Operator regression,followed by validation in an independent test cohort.Patients were stratified into high-risk and low-risk groups according to their risk scores,facilitating further analysis of immune-related features and protein expressions of the model genes.Drug sensitivity analyses evaluated the efficacy of commonly used chemotherapeutic agents across different risk categories.Additionally,bioinformatics approaches elucidated the expression profile and prognostic significance of epoxide hydrolase 2(EPHX2),which was subsequently validated through data obtained from the Human Protein Atlas database as well as real-time quantitative polymerase chain reaction.Finally,potential traditional Chinese medicines that may target EPHX2 regulation were predicted using data from the Traditional Chinese Medicine Syndrome Association database.Results A robust prognostic risk model predicated on pyrimidine metabolism gene signatures was established,identifying 15 key regulatory genes.Notably,patients classified within the high-risk group exhibited significantly lower overall survival rates compared to those in the low-risk group(P <0.001).Immune profiling revealed that both immune scores(1 934±760 vs 1 418±747) and stromal scores(600±645 vs 296±677) were markedly elevated in low-risk patients relative to their high-risk counterparts(all P <0.001).Furthermore,increased infiltration levels of resting natural killer cells and macrophages characterized the high-risk group;conversely,enhanced infiltration by CD8~+T cells and monocytes predominated within low-risk individuals.Drug sensitivity assessments indicated that oxaliplatin and irinotecan might confer greater therapeutic benefits for patients categorized as high risk(P <0.001).Bioinformatics analyses suggested a correlation between elevated EPHX2 expression levels and improved prognosis outcomes(P <0.01).Immunohistochemical evaluations corroborated differential EPHX2 expression patterns observed via HPA database investigations.Heightened EPHX2 levels in normal lung epithelial cells contrasted with reduced expressions noted in lung adenocarcinoma cell lines [16HBE vs H1299=1.00±0.12 vs 0.50±0.11,P <0.01].Predictions regarding traditional Chinese medicine indicated that formulations aimed to clearing heat,detoxifying effects,astringing properties,or regulating qi could potentially modulate EPHX2 activity.Conclusion The construction of a risk prognostic model based on pyrimidine metabolism genes can be used as a potential tool for evaluating the prognosis of lung adenocarcinoma and provide new ideas for its diagnosis and treatment.更多还原