We apply the oligonucleotide microarray technology to study the expression profile of differentiated thyroid cancer (DTC) and to select transcripts which differentiate between its subclasses. We use state-of-the-art bioinformatic techniques, based on Support Vector Machines algorithms, to select not only the single solitary markers but rather the sets of genes which are taken into account cooperatively to enhance the diagnosis accuracy. Simultaneously, we try to measure the molecular distance between various histotypes of DTC.
We used high density oligonucleotide microarrays (HG-U133A, Affymetrix). We started our analysis from the simplest model, the comparison between papillary thyroid cancer and normal thyroid tissue. In total, 73 thyroid samples were investigated: 48 PTC samples and 25 samples of macroscopically unchanged thyroid tissue. We observed that the difference between PTC and benign thyroid was large, with thousands of genes differentiating these two classes. We further extended this analysis to the set of differentiated thyroid cancer samples (57 tumors: 11 FTC, 9 FA follicular adenomas and 37 PTC, both classic and follicular variant). We used Singular Value Decomposition to select 579 major variability genes in this dataset. Based on these genes, we were able to classify 18 of 20 FTCs or FAs as follicular and 36 of 37 PTCs as papillary. Next, we included also datasets obtained by other groups and publicly available to evaluate the genomic distance between various histotypes of differentiated thyroid cancer.
Conclusion: The distance between molecular profiles of papillary and follicular thyroid cancer is large. This indicates that both histotypes differ not only by their initiating events but also by further steps of neoplastic transformation. Simultaneously, we show the heterogeneity of follicular variant of papillary thyroid ca.
01 - 05 Apr 2006
European Society of Endocrinology