Prior laboratory prediction of individual drug response is of key importance in esophageal squamous cell carcinoma (ESCC), because of the extremely narrow therapeutic index of chemotherapy. However, very few critical markers have been validated to date for ESCC. We previously demonstrated that simultaneous performance of two different types of comprehensive gene expression analysis might provide a way to identify potent marker genes for drug sensitivity from the expression-sensitivity correlation analysis alone, but the screening method appeared not to be always effective. Therefore, we attempted to identify novel potent marker genes using a new statistical analysis of oligonucleotide microarray expression data, based on a two-dimensional mixed normal model, and selected 3 and 7 novel candidates for 5-fluorouracil (5-FU) and cis-platinum (CDDP), respectively. Interferon induced transmembrane protein 1 (IFITM1) gene alone, being suggested as a key gene of Wnt pathway, was commonly selected in both screening methods. The transfection analyses and siRNA-mediated knock-down experiments revealed that expression of IFITM1 closely related to cellular sensitivity to CDDP. Considering the fact that drug sensitivity is determined by multiple genes, we established the best linear model using quantified expression data of a set of all the selected marker genes including IFITM1, which converted the quantified expression data of ESCC cell lines into an IC50 value of each drug. In the same way, using the representative genes selected in vitro, we developed highly predictive formulae for disease-free survival (DFS) of the CDDP/5-FU combination after curative operation in esophageal cancer patients (R=0.917). A two-dimensional mixed normal model can be a powerful tool to identify novel drug-response determinants, and the IFITM1 gene selected by the statistical method a novel critical biomarker of CDDP response in ESCC.