Pharmacophore-guided diffusion generates 3D molecules without needing a protein structure
- PharmaDiff is a diffusion-based generator that builds 3D molecules steered by a pharmacophore hypothesis, which a transformer encodes as atom-level features. This suits targets with little structural or functional data.
- In testing, its outputs fit the requested pharmacophore constraints better than existing ligand-based design approaches.
- In a structure-based setting across several proteins, it also reached better docking scores, even though it never saw the target structure.
💡 Points Based on the abstract
Better docking scores from a model that never saw the pocket are a flag, not a win: docking rewards size and lipophilicity, so I'd check whether the outputs are simply bigger or greasier than the baselines. On my own project I'd run it against known actives for a ligand-only target and ask what fraction are synthesizable, novel scaffolds rather than near-copies of the pharmacophore's source ligands.
📄 Paper · Journal of Chemical Information and Modeling (ACS) ↗ Pharmacophore-Conditioned Diffusion Model for Ligand-Based De Novo Drug Design