Thanks for your code sharing and analysis in paper.
In the main text of the paper and in the Supplementary, it says that you predicted protein-ligand binding using only the structures of the protein and ligand in the protein alone and ligand alone models, respectively.
So, for example, how does the protein alone model predict the binding affinity for multiple ligands for a given protein?
If you actually used only the structure of the protein, I don't think you would be able to distinguish the binding affinity values for each ligand.
Looking at the code, it looks like the protein and ligand alone model also used the structures of the protein-ligand complex, protein, and ligand as input, so I'm curious how you actually predicted multiple ligands using only the protein.
Thanks for your code sharing and analysis in paper.
In the main text of the paper and in the Supplementary, it says that you predicted protein-ligand binding using only the structures of the protein and ligand in the protein alone and ligand alone models, respectively.
So, for example, how does the protein alone model predict the binding affinity for multiple ligands for a given protein?
If you actually used only the structure of the protein, I don't think you would be able to distinguish the binding affinity values for each ligand.
Looking at the code, it looks like the protein and ligand alone model also used the structures of the protein-ligand complex, protein, and ligand as input, so I'm curious how you actually predicted multiple ligands using only the protein.