Unsupervised analysis of follicular thyroid tumours transcriptome by oligonucleotide microarray gene expression profiling

Introduction: Mechanisms driving the invasiveness of follicular thyroid cancer (FTC) are not fully understood. In our study, we undertook an unsupervised analysis of the set of follicular thyroid tumours (adenomas (FTA) and carcinomas) to verify whether the malignant phenotype influences major sources of variability in our dataset. Material and methods: The core set of samples consisted of 52 tumours (27 FTC, 25 FTA). Total RNA was analysed by oligonucleotide microarray (HG-U133 Plus 2.0). Principal Component Analysis (PCA) was applied as a main method of unsupervised analysis. Results: An analysis of biological character of genes correlated to the first six PCs was performed. When genes correlated to the first PC were used to cluster FTC and FTA, they appeared in two branches; one, relatively enriched in adenomas, with homogenous expression of subset of genes, and the other containing mainly carcinomas, with down-regulation of these genes and heterogeneous up-regulation in a smaller cluster of transcripts. Genes highly up-regulated in adenomas included some thyroid-specific transcripts. The second cluster of genes, up-regulated in carcinomas, contained mainly immunity-related transcripts. Immune response genes were found in the first, third and sixth principal components, improving the discrimination between carcinomas and adenomas. Conclusions: Our unsupervised analysis indicates that invasiveness of follicular tumours might be considered as the major source of variability in transcriptome analysis. However, the distance between both groups is small and the clusters are overlapping, thus, unsupervised analysis is not sufficient to properly classify them.