ABSTRACT

The work presented in this PhD dissertation was based on experiments performed at the Department of Clinical Biochemistry, Aarhus University Hospital, Skejby, Aarhus, Denmark. The aim of this PhD project was to identify gene expression patterns that characterize the different stages of colorectal cancer with the purpose of establishing the basis for a molecular classifier, and to evaluate custom cDNA microarrays to be used for verification of molecular classifiers (genes). We used the Affymetrix oligonucleotide microarrays to analyze the expression of more than 5000 genes in samples (normal mucosa and Dukes' stages A, B, C, and D) from the sigmoid and upper rectum of the left colon. In addition, three different software packages for analyzing Affymetrix oligonucleotide microarrays were evaluated. Technical reproducibility of custom cDNA microarrays, designed for classification of cancers, was also evaluated. A nearest neighbour classifier was used to classify normal, Dukes' B and C samples with less than 20% error, whereas Dukes' A and D could not be classified correctly. A number of interesting gene expression profiles showed a discriminating differences between Dukes' B and C samples. These included mitochondrial genes, stromal remodelling genes and heat shock genes. When comparing three different software packages there was little agreement in the absolute level of gene expression on the arrays (only 28% of the genes showed good correlation). However, there was a surprising good agreement between the software packages, when the relative (fold) changes of gene expression were used. When comparing the relative (fold) changes to those measured by RT-PCR, correlation coefficients above 0.8 were obtained. We also demonstrated that our custom cDNA arrays are of high technical quality.