@kjelljorner This may seem a bit redundant, because we are using output from docking (the ligand's pose) to predict the docking score. However, our ultimate goal is not score prediction, but de novo molecule generation :)
Check out our new poster at the 🤖💊 3rd RSC-BMCS/RSC-CICAG Artificial Intelligence in Chemistry Symposium 💊🤖, where we show how to improve VAE representations of molecules with structural information #AIChem20posters
@kjelljorner I'm glad you like it @kjelljorner ! We voxelized molecular poses that had been previously docked to the D2 receptor, so rotations and translations were "standardized" to a certain degree by the position of D2, which was the same for all ligands and served as a reference.