广西医学

目的基于生物信息学方法探讨巴雷特食管（BE）的铜死亡相关基因（CRGs），并进行潜在中药预测。方法在GEO数据库中选取BE相关数据集GSE39491（训练集）和GSE36223（验证集）。从FerrDb数据库和相关文献中获取25个CRGs。运用R语言软件筛选训练集中BE组和对照组之间的铜死亡差异表达基因（DECRGs），对DECRGs进行相关性分析、京都基因与基因组百科全书通路富集分析、共识聚类分析，构建机器学习模型［支持向量机（SVM）模型、极限梯度上升模型和广义线性模型］，以预测效能最佳的模型作为最佳机器学习模型，以最佳机器学习模型中重要性排名前5的基因作为特征基因，并对特征基因进行基因本体论功能富集分析。通过列线图、校准曲线、决策曲线和验证集评估最佳机器学习模型的有效性。利用COREMINE数据库及中医传承计算平台进行中药预测和用药规律分析。结果筛选得到13个DECRGs，DECRGs之间存在着不同的相关性。DECRGs主要富集在硫辛酸代谢、三羧酸循环、2⁃氧代羧酸代谢、碳代谢、丙酮酸代谢等信号通路。共识聚类将BE组织分为4个亚型。SVM模型为最佳的机器学习模型。ATP7B、PDHA1、ADM、LIPT1、SLC31A1为BE的特征基因。基于特征基因构建的列线图模型具有较优的预测准确度和临床实用性，当模型总得分>120分时，BE患病率超过90%。特征基因主要富集在铜离子稳态、铜离子转移、铜离子的平衡等生物过程，主要涉及基底外侧膜、基膜、基底细胞、次级内体等细胞组分，分子功能方面主要包括铜离子结合力、过渡金属离子跨膜转运活性等。基于特征基因预测得到的中药共63味，药性以温性、寒性、平性为主，药味以苦味、甘味、辛味为主，主归肝经、肺经、脾经、胃经，功效分类以清热类、补虚类、活血化瘀类为主，其中丹参、郁金、姜黄、黄芩、肉桂等为主要药物。结论 ATP7B、PDHA1、ADM、LIPT1、SLC31A1可作为BE的诊断性标志物，丹参、郁金、姜黄、黄芩、肉桂等中药可能通过调控CRGs发挥治疗BE的作用。

Objective To explore cuproptosis-related genes (CRGs) in Barrett esophagus (BE), and to perform potential Traditional Chinese Medicines prediction based on bioinformatics methods. Methods BE⁃related datasets GSE39491 (training set) and GSE36223 (validation set) were selected from the GEO database. Twenty⁃five CRGs were obtained from the FerrDb database and related literature. R language software was adopted to screen for differentially expressed CRGs (DECRGs) between the BE group and the control group in the training set. Correlation analysis, Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis, and consensus clustering analysis were performed on the DECRGs. Machine learning models including support vector machine (SVM) model, eXtreme gradient boosting model, and generalized linear model were constructed. The model with the optimal predictive performance was selected as the optimal machine learning model. The top 5 genes by importance in the optimal machine learning model were identified as feature genes, and Gene Ontology functional enrichment analysis was conducted for these feature genes. The effectiveness of the optimal machine learning model was evaluated using a nomogram, calibration curve, decision curve, and the validation set. The COREMINE database and the Traditional Chinese Medicine Inheritance Computing Platform were used for Traditional Chinese Medicine prediction and medication pattern analysis. Results Thirteen DECRGs were screened. Different correlations existed among the DECRGs. The DECRGs were mainly enriched in pathways such as lipoic acid metabolism, tricarboxylic acid cycle, 2⁃oxocarboxylic acid metabolism, carbon metabolism, and pyruvate metabolism. Consensus clustering divided BE tissues into four subtypes. The SVM model was the optimal machine learning model. ATP7B, PDHA1, ADM, LIPT1, and SLC31A1 were identified as feature genes for BE. The nomogram model constructed based on the feature genes demonstrated good predictive accuracy and clinical utility. When the total model score exceeded 120 points, the BE prevalence rate was over 90%. The feature genes were mainly enriched in biological processes such as copper ion homeostasis, copper ion transport, and copper ion equilibrium, in cellular components containing basal lamina, basal membrane, basal cell, secondary endosome, and in molecular functions such as copper ion binding and transition metal ion transmembrane transporter activity. Based on the feature genes, 63 flavors of Traditional Chinese Medicine were predicted. Their medicinal properties were mainly warm, cold, and moderate, flavors were mainly bitter, sweet, and acrid, primary meridians were the liver, lung, spleen, and stomach meridians, efficacy classifications were mainly heat⁃clearing, tonifying, and blood⁃activating and stasis⁃resolving. Important herbs included Salvia miltiorrhiza, Curcuma aromatica, Curcuma longa, Scutellaria baicalensis, and Cinnamomum cassia. Conclusion ATP7B, PDHA1, ADM, LIPT1, and SLC31A1 can serve as diagnostic markers for BE. Traditional Chinese Medicines such as Salvia miltiorrhiza, Curcuma aromatica, Curcuma longa, Scutellaria baicalensis, and Cinnamomum cassia may exert therapeutic effects on BE by regulating CRGs.