This repository contains the data and the codes for the manuscript "Relational graph convolutional networks for predicting blood-brain barrier penetration of drug molecules".
-
Run get_large_files.sh to download the big data files.
-
Calculate the drug-drug similarity by running drug_similarity.py. The results will be stored in drug_similarity.csv.
-
Collect the drug-protein interactions with drug_protein_interaction.py. The results will be saved in drug_protein_interaction.csv.
-
Run graph.py to generate the drug features and to structure the data into graphs. Two graphs will be built and be saved separately in the following files.
-
graph.pt includes the drug-protein interactions as the edges and the Mordred descriptors as the node features.
-
graph_drugsim.pt includes the drug-protein interactions and the drug-drug similarity as the edges, and the Mordred descriptors as the node features.
-
-
Run rgcn.py to train and evaluate the RGCN model with graph.pt as the input. Similarly, rgcn_drugsim.py trains and evaluates the RGCN model using graph_drugsim.pt as the input.