Establishment and Investigation of a Multiple Gene Expression Signature to Predict Long-Term Survival in Pancreatic Cancer.

Zhiqiang Zhang,Ying Li,Gang Wei,Di Wang,Menghong Yin,Chi Ma,Jian Du,Siling Hu,Guang Tan,Zhikun Lin,Xuelong Wang,Haifeng Luo,Jiangning Gu

doi:10.1155/2020/1570862

Abstract

Pancreatic cancer remains a lethal type of cancer with poor prognosis. Molecular classification enables in-depth, precise prognostic assessment. This study is aimed at identifying a robust and simple mRNA signature to predict the overall survival (OS) of pancreatic cancer (PC) patients. Differentially expressed genes (DEGs) between 45 paired pancreatic tumor samples and adjacent healthy tissues were selected. For risk determination, a LASSO Cox regression model with DEGs was used to generate the OS-associated risk score formula for the training cohort containing 177 PC patients. Another five independent datasets were used as the testing cohort to determine the predictive efficiency for further validation. In total, 441 DEGs were selected after considering the enrichment of classical pathways, such as EMT, cell cycle, cell adhesion, and PI3K-AKT. A five-gene signature for risk discrimination was established with high efficacy using LASSO Cox regression in the training group. External validation showed that patients identified by the gene expression signature to be in the high-risk group had poorer prognosis compared with the low-risk patients. Further investigation identified the differential epigenetic modification patterns of the five genes, which indicated their roles in tumor progression and their effect on therapy. In conclusion, we constructed a robust five-gene expression signature that could predict the OS of PC patients, offering a new insight for risk discrimination in daily clinical practice.

Highlights

Great improvements have been achieved in detection and treatment of many types of highly malignant tumors, such as lung and breast cancers, the overall survival (OS) and prognosis of pancreatic cancer (PC) remain poor, with a five-year survival rate of only around 8% [1]
The dataset of GSE28735 was used for the selection of differentially expressed genes (DEGs) and functional enrichment analysis
We used the TCGA-PAAD dataset, which contained 177 PC patients with detailed survival data, combined with the selected Differentially expressed genes (DEGs) to construct a gene expression signature prognostic risk score model based on the Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression

Summary

Introduction

Great improvements have been achieved in detection and treatment of many types of highly malignant tumors, such as lung and breast cancers, the overall survival (OS) and prognosis of pancreatic cancer (PC) remain poor, with a five-year survival rate of only around 8% [1]. Surgical resection offers the only chance for long-term survival, since PC is naturally resistant to chemo- and radiotherapy. About 20% of patients have the opportunity to receive surgical resection, and the median OS is only around 24 months [2]. TNM staging formulated by AJCC is used to determine the course of the treatment. Even two patients at the same TNM stage may have totally different prognoses [3]. This means that clinical histopathological classification has inherent limitations in predicting the prognosis for PC patients, and identification of new biomarkers for prognostic assessment is urgently required [4]

Objectives

Methods

Results

Conclusion