GeneExpressScore Signature: a robust prognostic and predictive classifier in gastric cancer. Issue 11 (28th September 2018)

• Record Type: Journal Article
• Authors: Zhu, Xiaoqiang; Tian, Xianglong; Sun, Tiantian; Yu, Chenyang; Cao, Yingying; Yan, Tingting; Shen, Chaoqin; Lin, Yanwei; Fang, Jing‐Yuan; Hong, Jie; Chen, Haoyan
• Is Part Of: Molecular oncology. Volume 12:Issue 11(2018)
• Page Start: 1871
• Page End: 1883
• Publication Date: 2018-09-28
• Subjects: gastric cancer -- gene expression -- LASSO -- prognosis -- signature; Cancer -- Molecular aspects -- Periodicals; 616.994005
• DOI: 10.1002/1878-0261.12351
• Languages: English
• ISSNs: 1574-7891

Abstract

Abstract : Although several prognostic signatures have been developed for gastric cancer (GC), the utility of these tools is limited in clinical practice due to lack of validation with large and multiple independent cohorts, or lack of a statistical test to determine the robustness of the predictive models. Here, a prognostic signature was constructed using a least absolute shrinkage and selection operator (LASSO) Cox regression model and a training dataset with 300 GC patients. The signature was verified in three independent datasets with a total of 658 tumors across multiplatforms. A nomogram based on the signature was built to predict disease‐free survival (DFS). Based on the LASSO model, we created a GeneExpressScore signature (GESGC ) classifier comprised of eight mRNA. With this classifier patients could be divided into two subgroups with distinctive prognoses [hazard ratio (HR) = 4.00, 95% confidence interval (CI) = 2.41–6.66, P < 0.0001]. The prognostic value was consistently validated in three independent datasets. Interestingly, the high‐GESGC group was associated with invasion, microsatellite stable/epithelial–mesenchymal transition (MSS/EMT), and genomically stable (GS) subtypes. The predictive accuracy of GESGC also outperformed five previously published signatures. Finally, a well‐performed nomogram integrating the GESGC and four clinicopathological factors was generated to predict 3‐ and 5‐year DFS. In summary, we describe an eight‐mRNA‐based signature, GESGC, as a predictive model for disease progression in GC. The robustness of this signature was validated across patient series, populations, and multiplatform datasets. Abstract : A total of 978 gene expression profiles of gastric cancer were used to construct a prediction model based on LASSO‐Cox regression model. This classifier was an independent prognostic factor. It also outperformed five reported gene signatures.