ShEPhERD: Shape, Electrostatics, and Pharmacophore Explicit Representation Diffusion

A really interesting preprint caught my attention from Connor Coley’s group at MIT. ShEPhERD diffusing shape, electrostatics, and pharmacophores for bioisosteric drug design

https://arxiv.org/abs/2411.04130v1

… a new 3D molecular generative model that facilitates interaction-aware chemical design by learning the joint distribution over 3D molecular structures and their shapes, electrostatics, and pharmacophores. Empirically, ShEPhERD can sample chemically diverse molecules with highly enriched interaction-similarity to target structures, as assessed via custom 3D similarity scoring functions. In bioisosteric drug design, ShEPhERD can design small-molecule mimics of complex natural products, diversify bioactive hits while enriching docking scores despite having no knowledge of the protein, and merge fragments from experimental fragment screens into bioisosteric ligands.

All code is available on GitHub

https://github.com/coleygroup/shepherd code to train and sample from ShEPhERD’s diffusion generative model, which learns the joint distribution over 3D molecular structures and their shapes, electrostatics, and pharmacophores.

https://github.com/coleygroup/shepherd-score code for generating/optimizing conformers, extracting interaction profiles, aligning interaction profiles, and differentiably scoring 3D similarity.

Related Posts