Identification of high-quality cancer prognostic markers and metastasis network modules

Cancer patients are often overtreated because of a failure to identify low-risk cancer patients. Thus far, no algorithm has been able to successfully generate cancer prognostic gene signatures with high accuracy and robustness in order to identify these patients. In this paper, we developed an algorithm that identifi es prognostic markers using tumour gene microarrays focusing on metastasis-driving gene expression signals. Application of the algorithm to breast cancer samples identifi ed prognostic gene signature sets for both estrogen receptor (ER) negative ( 12 ) and positive ( + ) subtypes. A combinatorial use of the signatures allowed the stratifi cation of patients into low-, intermediate- and high-risk groups in both the training set and in eight independent testing sets containing 1,375 samples. The predictive accuracy for the low-risk group reached 87 \u2013 100 % . Integrative network analysis identifi ed modules in which each module contained the genes of a signature and their direct interacting partners that are cancer driver-mutating genes. These modules are recurrent in many breast tumours and contribute to metastasis.Peer reviewed: YesNRC publication: Ye