Department of Gastrointestinal Surgery, Changhai Hospital, Shanghai 200433, P.R. China.
Department of Gastroenterology, Changhai Hospital, Shanghai 200433, P.R. China.
Mol Med Rep. 2018 May;17(5):6345-6356. doi: 10.3892/mmr.2018.8722. Epub 2018 Mar 9.
Gastric cancer (GC) has a high incidence and mortality rate. If discovered late, GC tends to have a poor prognosis. Improvements in the prognostic accuracy of GC through combined analysis of multiple relevant genes and clinical factors may solve this problem. In the present study, GSE62254 (including 300 GC tissues), obtained from the Gene Expression Omnibus database, was used as a training set, and the mRNA‑sequencing data of GC (including 384 GC tissues) downloaded from the Cancer Genome Atlas database served as a validation set. Based on the t‑test and Wilcoxon test, the significantly differentially expressed genes (DEGs) were obtained by screening the intersecting DEGs. The prognosis-associated genes and clinical factors were identified using Cox regression analysis in the R survival package. The optimal prognosis‑associated pathways were examined using the Cox‑proportional hazards (Cox‑PH) model in the R penalized package. Finally, risk prediction models were constructed and validated using the Cox‑PH model and the Kaplan‑Meier method, respectively. There were a total of 382 significant DEGs, including 268 upregulated genes and 114 downregulated genes. A total of 50 prognosis‑associated genes were identified, 16 optimal prognosis‑associated pathways (including mitochondrial pathway and the tyrosine‑protein kinase JAK‑signal transducer and activator of transcription signaling pathway, which involve caspase 7, phosphoinositide‑3‑kinase regulatory subunit 3, peroxisome proliferator‑activated receptor γ and collagen triple helix repeat containing 1) and four prognosis‑associated clinical factors [including Pathologic_N, Pathologic_stage, mutL homolog 1 (MLH1) mutation and recurrence]. The pathway‑ and clinical‑factor‑based risk prediction model exhibited marked prognostic accuracy. The clinical‑factor‑based risk prediction model with improved P‑values for prognosis prediction may be superior to the pathway‑based risk prediction model in predicting the prognosis of GC patients.
胃癌（GC）发病率和死亡率均较高，如果发现较晚，GC 预后往往较差。通过综合分析多个相关基因和临床因素来提高 GC 的预后准确性，可能会解决这个问题。本研究以基因表达综合数据库（GEO）中的 GSE62254（包含 300 例 GC 组织）为训练集，癌症基因组图谱（TCGA）数据库中下载的 GC 的 mRNA 测序数据（包含 384 例 GC 组织）作为验证集。通过 t 检验和 Wilcoxon 检验，筛选出交集差异表达基因（DEGs）。利用 R survival 包中的 Cox 回归分析筛选出与预后相关的基因和临床因素。利用 R 惩罚包中的 Cox 比例风险（Cox-PH）模型检测最佳预后相关通路。最后，分别利用 Cox-PH 模型和 Kaplan-Meier 法构建和验证风险预测模型。共得到 382 个显著差异表达基因，其中 268 个上调基因和 114 个下调基因。共鉴定出 50 个与预后相关的基因，16 个最佳预后相关通路（包括线粒体通路和酪氨酸蛋白激酶 JAK-信号转导和转录激活因子信号通路，涉及半胱氨酸蛋白酶 7、磷脂酰肌醇 3-激酶调节亚基 3、过氧化物酶体增殖物激活受体 γ 和胶原三螺旋重复包含 1）和 4 个与预后相关的临床因素[包括病理 N、病理分期、错配修复蛋白 1（MLH1）突变和复发]。基于通路和临床因素的风险预测模型具有显著的预后准确性。基于临床因素的风险预测模型，其预后预测的 P 值有所提高，可能优于基于通路的风险预测模型，有助于预测 GC 患者的预后。
