To improve the prognosis of advanced esophageal cancer, neoadjuvant chemotherapy (NACT) followed by surgery is a promising treatment strategy. NACT has been shown to improve the prognosis of responders. However, non-responders not only suffer from side-effects, but they also lose precious time to take advantage of other possible treatments. Therefore, it is crucial to establish a reliable method that allows prediction of response before chemotherapy. A biopsy sample can provide valuable information on the biological characteristics of an individual esophageal cancer, which can affect chemosensitivity. Comprehensive gene expression profiling (GEP) using oligonucleotide microarray covering 30,000 human probes was performed in 50 pretreatment endoscopic biopsy samples from 25 patients with esophageal squamous cell cancer (ESCC) who underwent cisplatin-based chemotherapy (two samples per patient). Chemotherapeutic responses were evaluated by the reduction rate of the tumor area on CT scans. Responders were defined as patients with reduction rates of ≥50% and non-responders were defined as patients with <50% decrease. The diagnostic system, that predicts responses to chemotherapy, was constructed with the 199 most informative genes, and showed 82% of accuracy. Furthermore, the predictive performance of this system was confirmed using an additional ten samples with an accuracy of 80%. This study shows that GEP of pretreatment ESCC biopsy samples has the potential to predict responses to chemotherapy.