Prediction Of Disease Genes Based On Nonlinear Induction Matrix Completion Model

The research on disease genes is of great significance to clinical medicine and the research and development on drug target.Traditional biological experiments are timeconsuming and painstaking.Using computer technology to study genetic diseaserelated issues has its own advantages and has gradually become the main research method.Based on the hypothesis that similar genes of disease-causing genes are more likely to be related to diseases,many related algorithms have been proposed.However,many methods only use the linear characteristics of the data,while the information of genetic disease data is more complicated and may contain many non-linear relationships.In this thesis,a nonlinear induction matrix completion model is constructed based on the neural network structure.The neural network model SDAE is used in the model to extract the nonlinear correlation features of the samples in the feature matrix.In order to make better use of the nonlinear relationship between nodes in the network,we use a frame combined with Graph Convolutional Network(GCN)and Non-linear matrix completion.In addition,because the biological data is too sparse and contains a lot of noise,in order to avoid the influence of a single data set on the model,we use feature data sets from multiple sources.In order to verify the effect of the model,we used three-fold cross-validation to compare our method with the other two inductive matrix filling models from multi-angle and multi-index.The results show that the performance of our proposed model is better than the other two methods.