%0 Journal Article %T 基于WGCNA和ssGSEA的胰腺癌预后模型构建 %T Construction of pancreatic cancer prognosis model based on WGCNA and ssGSEA %A 张峻烽,刘淞淞,王槐志 %A ZHANG,Jun feng %A LIU,Song song %A WANG,Huai zhi %J 中国临床新医学 %J CHINESE JOURNAL OF NEW CLINICAL MEDICINE %@ 1674-3806 %V 13 %N 11 %D 2020 %P 1084-1090 %K 胰腺癌;加权基因共表达网络分析;单样本基因富集分析 %K Pancreatic cancer;Weighted gene co-expression network analysis(WGCNA);Single sample gene enrichment analysis(ssGSEA) %X [摘要] 目的 通过加权基因共表达网络分析(WGCNA)和单样本基因富集分析(ssGSEA)构建胰腺癌预后模型。方法 获取胰腺癌组织转录组测序数据和患者预后信息,ssGSEA对各肿瘤组织17 810条通路进行评分,利用WGCNA、单因素分析和Lasso回归筛选各通路参数,构建多因素Cox比例风险回归模型对胰腺癌患者的预后情况进行预测,并对模型进行评价。结果 经ssGSEA发现,不同胰腺癌组织各通路评分不同,如GO RECEPTOR REGULATOR ACTIVITY、GO MONOSACCHARIDE BINDING等通路,提示不同患者各通路活性有差异,这可能是引起肿瘤向不同结局发展的重要原因。经WGCNA分析采用动态剪切树法合并表达相似的通路模块,最终得到7个通路模块。其中,Blue模块中包含的通路最多,含3 098条通路。Red模块与胰腺癌患者生存状态有较高的正相关性(r=0.33,P<0.001),Black模块也与其生存有显著的相关性(r=0.11,P=0.03)。同时,Red模块(r=0.30,P<0.001)和Blue模块(r=0.24,P=0.002)与胰腺癌组织学分期有显著的正相关性;Blue模块(r=0.27,P<0.001)、Yellow模块(r=-0.17,P=0.03)、Black模块(r=0.30,P<0.001)和Magenta模块(r=0.15,P=0.03)与胰腺癌肿瘤分期有明显的相关性。使用单因素分析、Lasso回归筛选Red模块中与生存预后相关的通路,共得到11条与胰腺癌患者生存预后相关的通路。进一步利用多因素Cox回归建立多通路预后预测模型,最终得到由5条通路构成的胰腺癌预后预测模型,C指数为0.7,赤池信息准则(AIC)为776.49,且该模型具有良好的预测效能(AUC=0.742)。结论 WGCNA和ssGSEA技术在胰腺癌预后模型的构建中有重要作用。 %X [Abstract] Objective To construct a prognostic model of pancreatic cancer through weighted gene co-expression network analysis(WGCNA) and single sample gene enrichment analysis(ssGSEA). Methods The pancreatic cancer tissue transcriptome sequencing data and the patients′prognosis information were obtained. Seventeen thousand eight hundred and ten pathways in each tumor tissue were scored by ssGSEA. WGCNA, univariate analysis and Lasso regression were used to screen the parameters of each pathway. A multifactorial Cox risk proportional regression model was constructed to predict the prognosis of the patients with pancreatic cancer, and the model was evaluated. Results It was found by ssGSEA that the scores of various pathways in different pancreatic cancer tissues were different, such as GO RECEPTOR REGULATOR ACTIVITY, GO MONOSACCHARIDE BINDING and other pathways, suggesting that the activities of various pathways in different patients were different, which might be an important reason for the development of different tumor outcomes. After WGCNA analysis, the dynamic shearing tree method was used to merge the similar pathway modules, and 7 pathway modules were obtained. Among them, the Blue module contained the most pathways, including 3098 pathways. The Red module had a high positive correlation with the survival status of the patients with pancreatic cancer(r=0.33, P<0.001), and the Black module also had a significant correlation with the patients′survival(r=0.11, P=0.03). At the same time, the Red module(r=0.30, P<0.001) and the Blue module(r=0.24, P=0.002) had a significant positive correlation with histological staging of pancreatic cancer. The Blue module(r=0.27, P<0.001), the Yellow module(r=-0.17, P=0.03), the Black module(r=0.30, P<0.001) and the Magenta module(r=0.15, P=0.03) were significantly related to the tumor staging of pancreatic cancer. After univariate analysis and Lasso regression were used to screen the pathways related to survival and prognosis in the Red module, a total of 11 pathways related to the survival and prognosis of pancreatic cancer patients were obtained. The multi-factor Cox regression analysis was further used to establish a multi-channel prognostic prediction model, and finally a pancreatic cancer prognostic prediction model composed of 5 pathways was obtained. The C-index was 0.7 and Akaike′s information criterion(AIC) was 776.49. The model had good prediction efficiency(AUC=0.742). Conclusion WGCNA and ssGSEA techniques play a significant role in the construction of prognostic model of pancreatic cancer. %R 10.3969/j.issn.1674-3806.2020.11.03 %U http://www.zglcxyxzz.com/ch/reader/view_abstract.aspx %1 JIS Version 3.0.0