Skip to main content
Fig. 5 | BMC Bioinformatics

Fig. 5

From: PIKE-R2P: Protein–protein interaction network-based knowledge embedding with graph neural network for single-cell RNA to protein prediction

Fig. 5

The method of PIKE-R2P. a is the whole pipeline and the model structure. The pipeline includes data denoising, model training, and testing. The green matrix represents denoised RNA data; the blue matrix is the high-dimensional representation of the RNA data; the orange vectors are the features of proteins and the orange lines correspond to the edges in the PPI network; the purple vectors are representations of the prior knowledge. b The knowledge embedding structure. The prior knowledge of each protein is mapped into a high-dimensional feature matrix. Then the attention mechanism is used to select the features. After that, the weights of protein interaction pairs linked to the same protein are adjusted. Finally, these feature vectors are concatenated together into one matrix as the prior knowledge embedding

Back to article page