Fig 1. Bioinformatics analysis workflow of BE datasets for biomarker discovery.

A. Three publicly available microarray datasets of BE data containing 33 BE samples with progression information were used to interrogate the expression levels of 12719 unique genes. B. After data normalization with frozen robust multi-array (fRMA) we first performed C. differential gene expression analysis, which allowed the identification of 799 differentially expressed genes. Normalized fRMA data was subsequently submitted to D. gene expression barcode which further restricted the number of selected candidates to 19. The combined usage of a E. systems biology approach plus F. manual literature curation selected the two most promising candidates.