Abstract
Heat shock proteins (HSPs) are a large family of molecular chaperones, which have shown to be implicated in various hallmarks of cancer such as resistance to apoptosis, invasion, angiogenesis, induction of immune tolerance, and metastasis. Several studies reported aberrant expression of HSPs in liquid biopsies of cancer patients and this has opened new perspectives on the use of HSPs as biomarkers of cancer. However, no specific diagnostic, predictive, or prognostic HSP chaperone-based urine biomarker has been yet discovered. On the other hand, divergent expression of HSPs has also been observed in other pathologies, including neurodegenerative and cardiovascular diseases, suggesting that new approaches should be employed for the discovery of cancer-specific HSP biomarkers. In this study, we propose a new strategy in identifying cancer-specific HSP-based biomarkers, where HSP networks in urine can be used to predict cancer. By analyzing HSPs present in urine, we could predict cancer with approximately 90% precision by machine learning approach. We aim to show that coupling the machine learning approach and the understanding of how HSPs operate, including their functional cycles, collaboration with and within networks, is effective in defining patients with cancer, which may provide the basis for future discoveries of novel HSP-based biomarkers of cancer.