Tumor initiation and progression is a complex process that differs not only between different tumors but also between distinct areas of a single tumor. Therefore it is important to analyze key players in cellular metabolism in tumors simultaneously. Lung cancers are one of the dominant causes of tumor-related death. The prospect for understanding the genetic background of these tumors and their complex pathological picture makes lung cancer a particularly interesting target for functional genomic analysis. We compared tissue samples from lung adenocarcinomas and squamous cell carcinomas with samples from adjacent tumor-free resection margins. Tissue samples were collected at surgery and immediately frozen in liquid nitrogen, and all sample material was histologically classified. For expression array analysis we purified total RNA using an optimized CsCl-cushion centrifugation protocol. We hybridized complementary DNA array membranes (Clontech Atlas Arrays and UniGene-based membranes from RZPD, Berlin) with 33P cDNA and collected data by phosphoimaging. Comparison of the expression patterns led us to the following observations: (1) Tumor and tumor-free resection margins show a high degree of correlation in their expression patterns. (2) Comparisons between tumors or between tumor-free tissue samples show no obvious correlation. (3) The similarity between tumors and normal tissue from one patient is more pronounced in adenocarcinoma than in squamous cell carcinoma.(4) Filtering genes that represent plate epithelial carcinoma versus adenocarcinoma led to two potential discriminating candidate genes, those coding for interleukin-1α and granzyme A.