Prediction of Cancer Driver Genes Using a Deep Convolutional Network

Abstract

The paper describes a method for predicting genes associated with the development of cancer. The method applies the convolutional neural network for the purpose of predicting disease driver genes. Distinctive features of the method are the use of gene expression data to determine the topological structure of the network, the efficiency of prediction with limited information about genes associated with the disease, and the possibility of jointly including information on mutations and similarity of gene expression profiles to improve the accuracy of prediction.