The incidence of thyroid cancer is rapidly increasing worldwide. Whilst outcome in thyroid cancer is generally good, up to 25% of patients develop recurrence, and have a significantly reduced life expectancy. We hypothesise those thyroid tumours which subsequently recur display a distinct pattern of driver events. Whole exome sequencing data were downloaded from The Cancer Genome Atlas (TCGA). Bioinformatic analysis of data on N=43 patients whose tumours recurred was performed, using a Platypus, Annovar and SIFT/PolyPhen2/MutationTaster filtering pipeline. This identified mutations in biologically significant genes, including Inosine-5’-monophosphate dehydrogenase 2 (IMPDH2), 6-Phosphofructo-2-Kinase/Fructose-2,6-Biphosphatase 4 (PFKFB4) and Dicer 1 ribonuclease type III (DICER1). As in-silico analysis suggested these variants to be potentially pathogenic, we recapitulated mutations from TCGA. Subcellular localisation, proliferation, cellular migration and invasion were investigated in cell lines which represent the most common background driver mutations of papillary thyroid cancer (TPC1: RET/PTC; SW1736: BRAF; Cal62: KRas). In TPC1 cells IMPDH2 mutation significantly increased cell migration at 4, 8 and 24 hrs vs. WT (P=0.0068, P=0.0008, P=0.0088 respectively) and DICER1 mutation induced increased cell migration at 24 hours vs. vector-only (P=0.0094). Overexpression of IMPDH2 resulted in altered intracellular localisation into intracellular discrete bodies known as cytoophidia. As recurrence may also affect altered gene expression, we analysed the RNA profile of the recurrent patients (N=43) compared to the non-recurrent (N=457). In particular, genes involved in matrix adhesion and thyroid cancer pathogenesis were most differentially expressed in recurrence patients, including fibronectin 1 (FN1), thyroglobulin (TG), α3 integrin (ITGA3), SPARC-like protein 1 (SPARCL1), Integral Membrane Protein 2A (ITM2A) and the proto-oncogene mesenchymal-epithelial transition factor (MET) (P=0.00376, P=0.00311, P=0.00757, P=0.01874, P=0.00003, P=0.00003 respectively). Overall, we propose that rare somatic mutations on top of established driver events, as well as specifically altered RNA expression levels, may be key to predicting thyroid cancer prognosis.