A Cancer Survival Prediction Method Based On Graph Convolutional Network

Cancer is one of the most colourful diseases in the history of human disease.It has always been one of the most important threats to human life and health.The high mortality of cancer is largely due to the complexity of cancer and the significant differences in clinical outcomes.Therefore,improving the accuracy of cancer survival prediction is of great significance and has become one of the main areas of cancer research.At present,many models for cancer survival prediction have been proposed,but most of them only use single genomic data or clinical data to generate prediction models,and there is no model for fusion of multi-genomic data and clinical data.In order to integrate multi-genomic data(including gene expression,copy number variation,gene methylation,exon expression)and clinical data effectively,and apply them to cancer survival prediction research,this paper proposes a method for cancer survival prediction based on graph convolution network which integrates the multi-genomic data and clinical data.First,integrating multi-genomic data and clinical data by using similarity network fusion algorithm,generating sample similarity matrix.Second,selecting the features from the multi-genomic data and clinical data through min-redundancy max-relevance algorithm to generate the feature matrix.Finally,a cancer survival prediction method,GCGCN,is obtained by semi-supervised training of the two matrices through the graph convolution network.The performance analysis of GCGCN model indicates that multi-genomic data and clinical data all play a key role in accurately predicting the survival of cancer patients.Furthermore,compared with the other existing methods of cancer survival prediction,the results demonstrate that the proposed GCGCN method,which fuse multi-genomic data with clinical data performs remarkably better than the existing methods of cancer survival prediction.In addition,we modify the GCGCN model,obtain the importance index of all features and get the topN feature,the accuracy and reliability of the method are further verified by analyzing these features.All the results of this study show that GCGCN is effective and superior in predicting cancer survival.