Fig 1.
iPiDA-GCN mainly contains three modules: (i) Heterogeneous network construction (Fig 1A). Three kinds of edges are collected in the heterogeneous piRNA-disease association network, including piRNA-piRNA similarities, disease-disease similarities and piRNA-disease interactions. (ii) GCN-based node feature extraction (Fig 1B). Asso-GCN and Sim-GCN modules are designed to continuously learn node features from different subnetworks of piRNA-disease association network. Specifically, Asso-GCN captures hidden associated features of heterogeneous nodes from piRNA-disease interaction subnetwork, while Sim-GCN captures hidden associated features of homogeneous nodes from two similarity subnetworks. (iii) Association prediction for piRNAs and diseases (Fig 1C). Three fully connected layers are employed to learn the low-dimensional representations of piRNAs and diseases. Finally, association scores between piRNAs and diseases are computed through inner product operation.
Table 1.
The impact of GCN layers on the predictive performance of iPiDA-GCN on .
Table 2.
The performance of iPiDA-FN, iPiDA-AssoGCN, iPiDA-SimGCN and iPiDA-GCN on .
Table 3.
Performance comparison among different methods on .
Fig 2.
The prediction visualization of iPiDA-GCN and compared methods.
These figures are plotted with the help of Gephi [49]. The nodes shown in orange and purple represent diseases and piRNAs, respectively. Pink lines denote the similarity associations between piRNAs, and black and red lines denote piRNA-disease associations in the training set and test set, respectively. The piRNA-disease associations predicted by different models are represented by blue lines.
Table 4.
The top 5 piRNAs associated with different diseases predicted by iPiDA-GCN.