MiRNA-disease association prediction method based on GCN coding and DNN decoding

The invention discloses a miRNA-disease association prediction method based on GCN (Genetic Content Network) coding and DNN (Determined Neural Network) decoding. According to the method, a graph convolutional network (GCN) is utilized to absorb and learn high-order features of nodes in the network, so that the complex relationship between miRNA and diseases is effectively captured. In research, mixed high-order neighborhood information of a layer in a multi-modal network is fused, and a learning representation process of mirna is introduced by taking Cm as an example. In addition, the GCN also considers the neighborhood information of the miRNA at different distances, so that the feature expression ability of the miRNA is further enhanced. Through an innovative GCN coding method, mixed high-order neighborhood information in a multi-modal network is fused to learn the characterization of miRNA and diseases; the DNN decoding part is used for decoding the characteristics of the GCN codes by using a deep neural n