Yes, I’ve got an entire network consisting of drugs, proteins and diseases and they are all connected via drug-protein, drug-disease, protein-protein and protein-disease edges. So from this I just split the training from the test graph (which only contains drug-disease edges). In this way, the edges connecting proteins to drugs or diseases should be helpful, right ?
Yes, I’ve got an entire network consisting of drugs, proteins and diseases and they are all connected via drug-protein, drug-disease, protein-protein and protein-disease edges. So from this I just split the training from the test graph (which only contains drug-disease edges). In this way, the edges connecting proteins to drugs or diseases should be helpful, right ?
For both training and test you should work on a same graph. In your case, it should be the entire network with test drug-disease edges removed.
Hi @sopkri,
For lengthy discussions like what you have here now, we would like to invite you to join the slack channel: https://join.slack.com/t/deep-graph-library/shared_invite/zt-eb4ict1g-xcg3PhZAFAB8p6dtKuP6xQ. There you may get more timely replies and it will be easier to keep track of the discussions. Thanks!
1 Like
Just a feedback about that, I am working on the same problem, and if this conversation had happened in slack, I would have never found it.
1 Like
Can you open a new discussion thread for the problem? You can also join the slack channel with this link: https://join.slack.com/t/deep-graph-library/shared_invite/zt-eb4ict1g-xcg3PhZAFAB8p6dtKuP6xQ.