Predicting distant metastatic sites of cancer using perturbed correlations of miRNAs with competing endogenous RNAs

Abstract

Cancer metastasis is the dissemination of tumor cells from the primary tumor site to other parts of the body via the lymph system or bloodstream. Metastasis is the leading cause of cancer associated death. Despite the significant advances in cancer research and treatment over the past decades, metastasis is not fully understood and difficult to predict in advance. In particular, distant metastasis is more difficult to predict than lymph node metastasis, which is the spread of cancer cells to nearby lymph nodes. Distant metastatic sites is even more difficult to predict than the occurrence of distant metastasis because the problem of predicting distant metastatic sites is a multi-class and multi-label classification problem; there are more than two classes for distant metastatic sites (bone, liver, lung, and other organs), and a single sample can have multiple labels for multiple metastatic sites. This paper presents a new method for predicting distant metastatic sites based on correlation changes of miRNAs with competing endogenous RNAs (ceRNAs) in individual cancer patients. Testing the method on independent datasets of several cancer types demonstrated a high prediction performance. In comparison of our method with other state of the art methods, our method showed a much better and more stable performance than the others. Our method can be used as useful aids in determining treatment options by predicting if and where metastasis will occur in cancer patients at early stages.