Improvements in detection, treatment and prognosis for patients with non-small cell lung cancer (NSCLC) depend on the molecular understanding of tumor development and progression. Using Affymetrix GeneChips comprising 59,620 elements, we determined the gene expression profiles of 89 NSCLC and 15 normal lung samples. We found 187 (0.3%) genes, which are at least 2-fold overexpressed and 157 (0.3%) genes 2-fold less expressed in NSCLC compared with normal lung. Cell cycle regulation, cell adhesion and nucleotide metabolism were the major biological processes connected to a large proportion of genes up-regulated in NSCLC. Down-regulated genes were frequently involved in metabolic/catabolic processes and signal transduction. The expression of specific genes revealed reliable differentiation of tumor from normal lung tissues, as well as the classification of both NSCLC subtypes squamous cell carcinoma and adenocarcinoma. In this context, collagens (COL7, 17) and cytokeratins (CK6, 15, 17) are preferentially induced in squamous cell carcinoma, whereas several transcription factors (TTF1, DAT1, TF-2) are exclusively elevated in adenocarcinomas. Some gene products involved in the metastatic process [matrixmetalloproteinase 12 (MMP-12) and urokinase plasminogen activator alpha (uPA)] were found as prognostic markers for the recurrence free interval and survival. Particularly, the simultaneous use of the MMP-12 and uPA expression predicted relapse-free and global survival of the patients. Screening of NSCLC with a genome-wide array revealed diagnostic, prognostic and potential therapeutic targets that might be suitable for an individual risk profile by tumor specific arrays.