The prognosis of pancreatic adenocarcinoma is affected by early metastases and local tumour invasion beyond surgical margins. Gene expression profiling in pancreatic cancer tissue is complicated due to the high amount of RNAses being present in human tissue and that of suitable models. In order to demonstrate early metastases, the models should take into account the anatomical environment of the tumour. Using the orthotopic transplantation of pancreatic tumour cells in SCID (severe combined immunodeficiency) mice, these interactions are taken into consideration. In order to identify genes associated with local tumour invasion and metastases in ductal pancreatic cancer, we investigated a human pancreatic tumour cell line derived from an orthopic pancreatic tumour model in SCID mice. Differential gene expression was performed on the basis of microarray technique. The human MiaPaca-2 cell line was implanted orthotopically in SCID mice. Transcriptional profiling was performed on fresh frozen tissue derived from the primary tumour, the tumour invasion front and the liver metastases. Differentially expressed genes were identified using statistical analyses, and were validated with external databases and with immunohistochemistry. A total of 1066 of 14 500 genes were significantly differentially expressed. Comparing the primary tumour with the tumour invasion front, there were 614 statistically significant up- and 348 downregulated genes. Twenty-five statistically significant up- and 181 downregulated genes were identified comparing the liver metastases with the primary tumour. Eight genes (PAI-1, BNIP3l, VEGF, NSE, RGS4, HSP27, GADD45A, PTPN14) were chosen and validated in a semi-quantitative immunohistochemical analysis, which revealed a positive correlation to the array data. Overrepresentation analyses revealed a total of 66 significantly regulated pathways associated with cell proliferation, cell stress, cell communication metabolic and cytokine function. In conclusion, model marker genes for local invasion and liver metastases can be identified using transcriptional profiling in the SCID mouse. Overrepresentation analysis secures a good and fast overview about the significantly regulated genes and can assign genes to certain pathways. These marker genes can be related to the apoptotic cascade, angiogenesis and cell interaction.