Gene amplification is one of the major mechanisms of oncogene activation in tumorigenesis. Currently, however, direct identification of amplified genes in cancer remains difficult. The most important technique allowing genome-wide screening of gene copy-number changes is comparative genomic hybridization (CGH). CGH is based on the hybridization of equal amounts of normal and tumour DNA in situ on normal metaphase chromosomes. DNAs are labelled with different fluorochromes and differences in the intensity ratios indicate copy-number differences in the amount of sample and reference DNA ref. 1. CGH only allows the detection of copy-number changes on a megabase scale, however, and the genes resident in amplicons must be identified using other techniques. We have developed a technique based on the hybridization of total genomic DNA on cDNA microarrays. These microarrays contain thousands of known genes and unknown ESTs. Compared with the previously published genomic microarray techniques where genomic DNA is hybridized on genomic clones (CGH microarrays), hybridization on cDNAs offers some advantages. The most important is the possibility to directly identify amplified genes, not just amplified genomic regions. Due to the complexity of genomic DNA and the small size of the arrayed probe (0.5–2 kb), however, the hybridization signals are weak. To facilitate signal detection, we have used a tyramide-based technique that allows amplification of a fluorescent signal up to 1,000-fold. We hybridized seven different cancer cell lines on cDNA microarrays. We detected the amplification of several genes, including C-myc, N-myc, SAS, MDM2, ERBB2 and OS4. The sensitivity of this assay was studied by diluting tumour DNA with normal genomic DNA, and the results suggest that amplifications of fivefold and higher can be detected. An additional advantage of this technique is the potential to measure copy number and gene expression for the same set of genes by hybridizing either DNA or cDNA on identical arrays. This will eventually allow correlation of gene amplification with gene expression on a genome-wide scale. We think that this technique is very likely to become an important tool for screening and discovery of genes important in tumorigenesis.