We have combined laser capture microdissection (LCM) sample collection of oral epithelial tissue with microarray analysis to study gene expression in oral cancer. Matched normal and malignant oral epithelial samples were collected from five patients. Each tissue sample was reduced from millions of mixed normal and malignant cells to a homogenous sample of thousands of normal or malignant cells using LCM. A combination of the total RNA method and multiple rounds of in vitro transcription was used to derive sufficient quality and quantities of labeled targets. We hybridized labeled target on HuGeneFL arrays containing probes representing approximately 6,800 full-length human genes. We derived a list of candidate genes using three analysis tools: GeneChip, GeneCluster and Matlab; 27 candidate genes were identified by all three methods. Several genes involved in cell proteolysis, migration and xenobiotic pathways were differentially expressed. Other genes with altered patterns included transcription factors, oncogenes and tumor suppressor genes, and differentiation markers. For several genes involved in known metastatic or invasion pathways, results were validated by quantitative polymerase chain recation with reverse transcription.