Endocrine Abstracts

We aim to distinguish malignant from benign thyroid nodules by their molecular profiles. To this end we employ a range of different genomic and epigenetic technologies such as classical whole genome microarray based transcriptomics, integration and meta analysis of published microarray data, DNA methylation testing and micro-RNA profiling. Our transcriptomics and meta analysis approaches include all major histological entities. The other approaches (DNA methylation and microRNA) focus on the distinction of follicular adenoma (FTA) and follicular carcinoma (FTC). We compare the performance of these profiles with respect to their classification task and present functional and genome-level integration of the data.

Transcriptomics and meta analysis: We conducted microarray analysis of 49 thyroid tumour nodules including all major histological classes. From this data we calculated inference statistics and employed different feature selection algorithms for classification of (1) malignant versus benign and (2) FTA versus FTC (using only a subset of 24 samples). This yielded two set of genes comprising of 20 and 22 genes, respectively. Classification accuracies in our data were 100 and 92%, respectively. Then, these gene sets were tested on seven different publicly available datasets (totalling 197 samples), two of those included FTA and FTC nodules. The classification accuracy of the FTA/FTC gene set was: 96 and 100%; the accuracy of the benign/malignant gene set was: 92, 87, 90, 92, 100, 94 and 100%.

Micro RNA: The 24 follicular samples used for mRNA analyses were also used for measuring their miRNA levels in a genome wide assay. We identified a set of six miRNAs which are potentially capable of distinguishing FTA from FTC.

DNA methylation: We used an in-house developed assay, based on methylation sensitive restriction digestion of DNA followed by microarray readout to assess the methylation status of 360 selected CpG islands. As little as two CpG islands were sufficient to distinguish between follicular adenoma and carcinoma.