Generative graph transformer?

I’m going to keep this as brief as possible: basically, I am trying to develop a generative deep learning model for de novo drug design. The challenging part is that I want to train it on complex 3D structural data rather than the SMILES strings and amino acid sequences that are typically used. I’m using the PDBBind dataset which has about 20k known protein-ligand (drug molecule) pairs, where protein structures are in .pdb files and ligands are in mol2 files. …

acid data deep learning deeplearning design drug design generative graph part strings train transformer