Using a panel of cDNA microarrays comprising 47 650 transcript elements, we have carried out a dual-channel analysis of gene expression in 39 resected primary human non-small cell lung tumours versus normal lung tissue. Whilst approximately 11 000 elements were scored as differentially expressed at least twofold in at least one sample, 96 transcripts were scored as over-represented fourfold or more in at least seven out of 39 tumours and 30 sequences 16-fold in at least two out of 39 tumours, including 24 transcripts in common. Transcripts (178) were found under-represented fourfold in at least seven out of 39 tumours, 31 of which are under-represented 16-fold in at least two out of 39 lesions. The relative expression levels of representative genes from these lists were analysed by comparative multiplex RT-PCR and found to be broadly consistent with the microarray data. Two dramatically over-represented genes, previously designated as potential tumour suppressors in breast (maspin) and lung and breast (S100A2) cancers, were analysed more extensively and demonstrate the effectiveness of this approach in identifying potential lung cancer diagnostic or therapeutic targets. Whilst it has been reported that S100A2 is downregulated in NSCLC at an early stage, our microarray, cmRT-PCR, Western and immunohistochemistry data indicate that it is strongly expressed in the majority of tumours.