Heatmap and hierarchical clustering of meta analysis data on further candidate marker genes. As shown in the heatmap, there are a range of different genes with good discriminatory power between PTC and benign nodules. Therefore, removing SERPINA1 from the meta analysis dataset leads to a range of possible expression signatures with 99% classification accuracy in leave-one-out crossvalidation (see main text). In all cases, the same one sample is misclassified (see discussion for details). Columns correspond to samples and rows to genes, the red/green color bar on top of the heatmap corresponds to the histological classification (Red: benign, Green: PTC).