Microarray profiling of 15 adjacent normal/tumor-matched esophageal squamous cell carcinoma (ESCC) specimens identified 40 up-regulated and 95 down-regulated genes. Verification of the microarray measurement by quantitative real-time reverse transcription PCR in the same set of samples as well as an additional 15 normal/tumor-matched samples revealed >95% consistency. These signatures can also be used to classify a recently reported ESCC microarray dataset. Moreover, these molecular signatures were used as templates to elucidate their corresponding protein-protein interaction (PPI) networks using the PPI databases, POINT and POINeT. As a result, 18 genes, of which six were not disclosed in the initial expression profile analysis, were found to be able to serve as the minimal discriminators for distinguishing ESCC tumors from normal specimens. Of these discriminators, ten (BGN, COL1A1, COL1A2, MMP9, CD44, FN1, TGFBI, PXN, SPARC and VWF) were associated with tumor metastasis and formed a highly interactive network with the first four molecules as 'hubs'. Our study not only reveals how novel insights can be obtained from gene expression profiling, but also highlights a group of highly interacting genes associated with metastasis in ESCC.