目的：发掘急性胰腺炎（acute pancreatitis， AP）中铜死亡的特征基因。方法：提取GSE194331数据集中铜死亡相关基因（cuproptosis-related genes， CRG）的表达，进行差异分析和免疫细胞相关性分析。根据CRG表达分出不同亚型，并利用基因集变异分析（gene set variation analysis， GSVA）富集代谢通路。采用广义线性模型（generalized linear models， GLM）、随机森林（random forest， RF）、支持向量机（support vector machine， SVM）和极端梯度提升（extreme gradient boosting， XGB）4种机器学习算法筛选疾病特征基因。结果：分析得到AP中差异表达CRG共13个（P<0.05）。CRG之间不仅存在不同程度的相关性，且与多种免疫细胞也具有相关性（P<0.05）。通过一致性聚类分析得到的2个亚型间有4条免疫相关通路存在差异，其中T细胞受体信号通路值得注意。进一步分析发现多种T细胞在两亚型间有显著差异（P<0.05）。每种机器学习算法各筛选出5个特征基因，得到了可作为下一个研究目标的二氢脂酰胺脱氢酶（dihydrolipoamide dehydrogenase， DLD）。结论：基于CRG的机器学习和生物信息学分析，挖掘AP中铜死亡相关基因，发现潜在的生物标志物。

Objective To discover the characteristic genes of cuproptosis in acute pancreatitis (AP). Methods The expression of cuproptosis-related genes (CRG) in the GSE1943 dataset was extracted and performed differential analysis and immune cell correlation analysis. The different subtypes were classified according to the expression of CRG, and metabolic pathway enrichment was performed using gene set variation analysis. Four machine learning algorithms, including generalized linear models, random forest, support vector machine and extreme gradient boosting were used to screen disease characteristic genes. Results A total of 13 CRG were differentially expressed in AP (P<0.05), and CRG were not only correlated to each other in different degrees, but also had correlation with multiple immune cells (P<0.05). There were four immune-related pathways among the two subtypes obtained by cluster analysis, in which the T-cell receptor signaling pathway was noteworthy. Further analysis revealed significant difference between the two subtypes of multiple T cells (P<0.05). Each machine learning algorithm screened out five characteristic genes, and dihydrolipoamide dehydrogenase (DLD) was obtained as the next target of research. Conclusions CRG-based machine learning and bioinformatics analyses could be used to explore CRG in AP to discover potential biomarkers.