Identifying a Gene Classifier between normal and Breast Cancer from information available within Publically Available Gene Expression repositories

Introduction: There are large volumes of data available within public gene expression repositories (PGER), however it is not possible to search these repositories using relevant clinical terms such as hormonal profile or stage... [ view full abstract ]

Introduction: There are large volumes of data available within public gene expression repositories (PGER), however it is not possible to search these repositories using relevant clinical terms such as hormonal profile or stage of breast cancer. Previous research at the University of Limerick has permitted clinically relevant searches of these datasets based upon specific clinical parameters

Aim: The aim of this study is to exploit the potential of the data available from publicly available gene expression repositories to develop a potential gene expression classifier to differentiate between normal tissue and breast cancer.

Methods: Gene Expression Omnibus (GEO) and ArrayExpress which were searched using the phrase, “breast". The datasets were subsequently reviewed and only those which obtained a 5* in Minimum Information About a Microarray Experiment (MIAME) compliance score within ArrayExpress were imported into Chipster for further analysis.

Results: This search yielded 29 datasets which obtained a 5* MIAME compliance score within ArrayExpress, 6 of these datasets GSE 42568, GSE 42568, GSE 50428, GSE 50567, GSE 57297 and GSE 59246 compared the microarray expression between normal and breast cancer tissue.

GSE 24556, GSE 50428, GSE 50567, GSE57297 and GSE 59246 were used to identify the gene classifier between normal breast tissue and breast cancer, which includes the genes C2orf40, FABP4, KRT 5 and MMP1. GSE 42568 was used to externally validate this classifier, which has an average sensitivity and specificity of 97% and 85% respectfully of identifying the dysregulation between normal breast tissue and breast cancer.

Conclusions: This demonstrates that it is possible to use the microarray expression data available within public gene expression repositories in relation to breast cancer to generate a gene classifier containing only 4 genes that can help distinguish between normal and breast cancer.