While the survival rate is close to 100% for men who are diagnosed with local prostate cancer disease, the survival rate for advanced stage patients is only about 28%. Approximately 10-20% of men present with metastatic disease, and many others progress to metastasis after hormone treatment, surgery or radiotherapy. However, few biomarkers have been effective at predicting which patients will metastasise, especially for patients who have biochemically recurred. As such, understanding why patients metastasise or become resistant to treatments, as well as identifying biomarkers that will predict response to therapy or propensity for more aggressive disease, is crucial to increasing survival of patients. We show that SOX2 is a marker for patients who metastasise after biochemical recurrence, and that patients with a SOX2-positive tumour will have a shorter time to metastasis and prostate cancer specific survival. Additionally, we show that SOX2 is not acting in its canonical stem cell-like manner in prostate cancer, and in this context does not confer a neuroendocrine phenotype. In castration-resistant prostate cancer, we see that SOX2 regulates both glycolysis and mitochondrial respiration, perhaps through regulation of PFKP, a key glycolytic enzyme, or increasing mitochondrial mass. We also observed that cells without SOX2 were more sensitive to enzalutamide, with decreased growth and a G1 arrest. This may be due to increased DNA damage, and a deficiency in DNA repair pathways. In this dissertation, we show for the first time a role for SOX2 in prostate cancer metabolism, as well as a novel role for SOX2 in DNA replication and repair. SOX2 can be used to identify patients at a higher risk for metastasis, and can aid in treatment decisions such as a radical prostatectomy for lower risk patients with SOX2-positive tumour, or increased screening for metastases. Alternatively, due to its role in DNA replication and repair, it could be used to predict patient response to radiation therapy, and patients with SOX2-positive tumours may not respond to radiation therapy. Overall, the implications of this work reveal great potential for the use of SOX2 as a biomarker clinically to inform therapeutic decisions and improve patient outcome.