Biological systems exhibit complex responses to xenobiotics varying from generic stress responses to very specific changes closely associated with the mechanism of toxicity. Until recently our view of this complexity was obscured by the simplicity of available analysis tools which allowed determination of only a few genes in any one study. Then genome sequencing and high throughput library screening projects delivered data on the genome sequence of many organisms, and clones were collected and made available to researchers in a previously unparalleled quantity. To exploit this new resource the microarray was developed from its predecessor the dot blot. Further development has expanded the number of clones contained on any one microarray to a point where the expression of many tens of thousands of genes in a biological system can be determined in a short period of time. What these data are revealing is the full complexity of the gene expression response to stimuli such as xenobiotic exposure. Toxicogenomics seeks to use the complexity of this response as a fingerprint or signature characteristic of that xenobiotic exposure. There are though two major experimental challenges that need to be dealt with for toxicogenomics to be successful. The first is technical and relates to the intrinsic difficulties associated with the accurate measurement of gene expression. For microarrays, this problem is multiplied by the number of genes on the microarray itself. To overcome this technical variability correct experimental design is critical. The second challenge concerns the biological system used. What genetic background, time point and dose of xenobiotic should be chosen? For in vitro systems should cell lines or primary cells be used? These factors, and more, could affect the gene expression profile obtained in response to the same xenobiotic exposure. Using both our data and data from public databases these issues are explored in this paper.