Meta-analysis of microarray data on pancreatic cancer defines a set of commonly dysregulated genes

Abstract

Pancreatic ductal adenocarcinoma is the eighth most common cancer with the lowest overall 5-year relative survival rate of any tumor type today. Expression profiling using microarrays has been widely used to identify genes associated with pancreatic cancer development. To extract maximum value from the available gene expression data, we applied a meta-analysis to search for commonly differentially expressed genes in pancreatic ductal adenocarcinoma. We obtained data sets from four different gene expression studies on pancreatic cancer. We selected a consensus set of 2984 genes measured in all four studies and applied a meta-analysis approach to evaluate the combined data. Of the genes identified as differentially expressed, several were validated using RT-PCR and immunohistochemistry. Additionally, we used a class discovery algorithm to identify a gene expression signature. Our meta-analysis revealed that the pancreatic cancer gene expression data sets shared a significant number of up- and downregulated genes, independent of the technology used. This interstudy crossvalidation approach generated a set of 568 genes that were consistently and significantly dysregulated in pancreatic cancer. Of these, 364 (64.1%) were upregulated and 204 (35.9%) were downregulated in pancreatic cancer. Only 127 (22%) were described in the published individual analyses. Functional annotation of the genes revealed that genes presumably associated with the cell adhesion-mediated drug resistance pathway are frequently overexpressed in pancreatic cancer. Meta-analysis is an important tool for the identification and validation of differentially expressed genes. These could represent good candidates for novel diagnostic and therapeutic approaches to pancreatic cancer.