Docking refinement with DiffDock

[DT6A]

Hello, thanks for your work!

Is it possible to run non-blind docking using partial diffusion, for example (maybe some code snippet is available)? E.g., I know some approximate position of the docking site but don't know it exactly and it is also needed to rotate ligand somehow.

Thank you

[DT6A]

Is it OK to remove initial random translations and to do something like this in inference? tr_schedule = get_t_schedule(inference_steps=args.inference_steps + 5)[-args.inference_steps:]

[key262yek]

It's quite late, but in my use, I saved ligand_center in inference_utils.py, and remove random_init function from inference.py and set the initial position to ligand_center - protein_center