Based on the paper for DiffDock (https://arxiv.org/abs/2210.01776 https://arxiv.org/abs/2210.01776) it looks like it was a great use case for a diffusion model.

Based on the paper for DiffDock (https://arxiv.org/abs/2210.01776 https://arxiv.org/abs/2210.01776) it looks like it was a great use case for a diffusion model. > We thus frame molecular docking as a generative modeling problem—given a ligand and target protein structure, we learn a distribution over ligand poses. I just hope work on these very valid use cases doesn’t get negatively impacted when the AI bubble inevitably bursts.