Gemcitabine is a commonly used therapy for many solid tumors. Acquired resistance to this nucleoside analogue, however, diminishes the long-term effectiveness in a majority of patients. To better define the molecular background of gemcitabine resistance, a mouse colon tumor was selected during successive rounds of transplantation with continued treatment of gemcitabine. Expression microarray analysis was applied to determine which genes are consistently and highly overexpressed or underexpressed in the resistant versus the nonresistant tumor. For the statistical interpretation of the microarray data, a parametric model was implemented, which returns model-based differential gene expression (log-) ratios and their uncertainties. This defined a set of 13 genes, putatively responsible for the gemcitabine resistance in solid tumors. One of these, RRM1, was previously identified as an important marker for gemcitabine resistance in human cell lines. Five of the 13 genes, including RRM1, are located within a 3 Mb region at chromosome 7E1 of which four are highly overexpressed, suggesting a chromosomal amplification. Therefore, chromosomal copy number changes were measured, using oligo array comparative genomic hybridization. A narrow and high amplification area was identified on 7E1 that encompassed all five genes. In addition, reduced RNA expression of two other genes at 8E1 encoding COX4I1 and RPL13 could be explained by a decrease in chromosomal copy number on chromosome 8. In conclusion, the array comparative genomic hybridization biologically validates our statistical approach and shows that gemcitabine is capable to select for chromosomally aberrant tumor cells, where changed gene expression levels lead to drug resistance.