Fig. 4.

Embedding visualization. (a) Two-dimensional t-SNE visualization of the high-dimensional embeddings generated by GDRnet for the nodes in the four-layered heterogeneous graph. The left embedding plot shows the representation of all the nodes (around 42 000 nodes), which are colored according to their layer. The right plot focuses on the drugs and diseases used for testing. (b) Embeddings of the COVID-19 disease nodes (27 SARS-CoV-2 proteins and 6 coronavirus related diseases) and the predicted drugs by GDRnet. The drugs in the both the plots (a) and (b) are colored according to their first-level anatomical therapeutic chemical (ATC) categorization.