利用生物信息学方法探索Pumilio（PUM）蛋白作为胃癌潜在生物标志物的可能性。

从UCSC Xena官网下载TCGA-STAD队列作为训练集，从GEO数据库下载GSE15459作为验证集。基于训练集中胃癌样本的生存时间和生存状态，按照PUM1和PUM2的最佳截断值将胃癌样本分为高、低表达组，分别对PUM1和PUM2进行生存分析，并选取与胃癌预后生存显著相关的PUM；然后分别对该PUM高、低表达组以及胃癌和正常样本组进行差异分析，筛选出PUM-差异表达基因（PUM-DEGs）和DEGs，两者取交集获得候选基因，单因素Cox和套索算法（LASSO）回归分析后得到风险模型基因，对胃癌和正常样本风险模型基因的表达量进行分析，并构建相关预后预测模型；最后进行免疫细胞浸润分析和化疗药物敏感性分析。

基于PUM2的高、低表达组间患者生存率差异有统计学意义（P<0.05），PUM2与胃癌的预后生存显著相关。通过差异分析共得到307个PUM2-DEGs和4 176个DEGs，获得209个候选基因，筛选出33个预后相关基因和LBP、CST2、IGFBP1、C5orf46共4个风险模型基因，其在胃癌样本中的表达量均大于正常样本。成功构建可有效预测胃癌患者1、3、5年生存率的预后预测模型并验证了其有效性。免疫细胞浸润分析发现，记忆B细胞、M0巨噬细胞、浆细胞等6种免疫细胞的丰度以及18种免疫检查点分子在高、低风险组间差异有统计学意义（P<0.05）。药物敏感性分析显示，低风险组达沙替尼、多西他赛、银胶菊内酯等药物的50%抑制浓度（IC₅₀）高于高风险组。

PUM2与胃癌患者的预后相关，是胃癌的潜在生物标志物，以4个PUM2-DEGs构建的模型可用于胃癌预后预测，基于预后预测模型进行的免疫细胞分析及药物敏感性分析提示了胃癌潜在的免疫和化疗药物治疗方向。

To explore the possibility of Pumilio (PUM) protein as a potential biomarker in gastric carcinoma (GC) by bioinformatics methods.

The TCGA-STAD queue was downloaded from UCSC Xena’s official website as the training set, and GSE15459 was downloaded from Gene Expression Omnibus (GEO) as the verification set. Based on the survival time and state of GC samples in the training set, GC samples were divided into high and low expression groups according to the optimal cut-off values of PUM1 and PUM2. Survival analysis was performed at PUM1 and PUM2 respectively, and PUM associated with GC prognostic survival significantly were selected. PUM-differentially expressed genes (PUM-DEGs) and DEGs were entially screened out to obtain candidate genes in the high and low expression groups of PUM, as well as in the GC and normal samples, respectively. Through univariate Cox and least absolute shrinkage and selection operator (LASSO) analyses, the risk model genes were further analyzed, and a prognosis prediction model was established. Finally, immunocyte infiltration analysis and chemotherapy drug sensitivity analysis were performed.

There were significant differences in survival rates between the groups with high and low PUM2-based expression (P<0.05), indicating that PUM2 was significantly correlated with the prognostic survival of GC. A total of 307 PUM2-DEGs and 4 176 DEGs were obtained by difference analysis, and 209 candidate genes were obtained, 33 prognostic related genes and 4 risk model genes (LBP, CST2, IGFBP1, C5orf46) were screened out. The expressions of LBP, CST2, IGFBP1 and C5orf46 in GC samples were higher than those in normal samples. A prognostic prediction model for predicting effectively the survival rates of GC patients at 1-, 3- and 5-year was constructed and verified. Immunocyte infiltration analysis showed significant differences in the abundance of six immune cells such as memory B cells, M0 macrophages, plasma cells and 18 immune checkpoint molecules between the high and low risk groups (P<0.05). The inhibitory concentration (IC₅₀) of drugs including dasatinib, docetaxel, parthenolide in the low risk group was higher than that in the high risk group.

PUM2 is associated with the prognosis of patients with GC and is a potential biomarker of GC. A prognostic prediction model constructed with four PUM2-DEGs can be used to predict the prognosis of GC. Immunoanalysis and drug sensitivity analysis based on the prognostic prediction model suggest the potential direction of immune and chemotherapy drug treatment of GC.