Abstract: We used hierarchical clustering to examine gene expression profiles generated by serial analysis of gene expression (SAGE) in a total of nine normal lung epithelial cells and non-small cell lung cancers (NSCLC). Separation of normal and tumor samples, as well as histopathological subtypes, was evident using the 3,921 most abundant transcript tags. This distinction remained when just 115 highly differentially expressed transcript tags were used. Furthermore, these 115 transcript tags clustered into groups that were suggestive of the unique biological and pathological features of the different tissues examined. Adenocarcinomas were characterized by high-level expression of small airway-associated or immunologically related proteins, while squamous cell carcinomas overexpressed genes involved in cellular detoxification or antioxidation.