Background The present study was designed to examine and analyze the global gene expression changes during the tumorigenesis of a human immortalized oral epithelial cell line, and search for the possible genes that may play a role in the carcinogenesis of oral cancer associated with benzo (a) pyrene.Methods The human immortalized oral epithelial cells, which have been established through transfection of E6/E7 genasof human papillomavirus type 16 and proved to be non-tumorigenic in nude mice, were treated with benzo (a) pyrene.Tumorigenesity of the treated cells were examined through nude mice subcutaneous injection. The global gene expression profiles of immortalized cells and the tumorigenic cells were acquired through hybridization of a microarray of Affymetrix U133 plus 2.0. The data were analyzed using Spring 7.0 software and treated statistically using one-way analysis of variance (ANOVA). The differentially expressed genes were classified using a Venn diagram and annotated with gene ontology. Several highlighted genes were validated in cells using a real-time polymerase chain reaction.Results There were 883 differentially expressed genes during the tumorigenesis and most of them changed expression in the early stage of tumorigenesis. These genes mainly involved in macromolecule metabolism and signal transduction,possessed the molecular function of transition metal ion binding, nucleotide binding and kinase activity; their protein products were mainly integral to membranes or localized in the nucleus and cytoskeleton. The expression patterns of IGFBP3, S100A8, MAP2K, KRT6B, GDF15, MET were validated in cells using a real-time polymerase chain reaction; the expression of IGFBP3 was further validated in clinical oral cancer specimens.Concluslona This study provides the global transcription profiling associated with the tumorigenesis of oral epithelial cells exposed to benzo (a) pyrene; IGFBP3 may play a potential role in the initiation of oral cancer related to benzo (a) pyrene exposure.