Fig. 5From: PDA-PRGCN: identification of Piwi-interacting RNA-disease associations through subgraph projection and residual scaling-based feature augmentationThe flowchart of PDA-PRGCN. PDA-PRGCN contains the following three parts. First, a median subgraph projection algorithm (A) was designed to construct the disease-disease (a1) and piRNA-piRNA subgraph (a2). Secondly, a residual scaling-based feature augmentation algorithm (B) was applied for normalized similarity profile information (b1) to enhance the node feature (b2). Thirdly, we combined the above two parts using a GCN (C) with two layers (c1, c2) by dual-loss mechanism. Finally, a three-layer fully-connected neural network predictor (c3) was used to optimize model loss and output the probabilities of potential PDAs for PDA predictionBack to article page