CSE Community Seminar | October 18, 2024

Abstract

Engineering molecules to exhibit precise 3D intermolecular interactions with their environment forms the basis of chemical design. In ligand-based drug design, bioisosteric analogues of known bioactive compounds are traditionally identified by virtually screening pre-enumerated chemical libraries with shape, electrostatic, and pharmacophore similarity scoring functions. We instead hypothesize that a generative model which learns the joint distribution over 3D molecular structures and their interaction profiles may facilitate 3D interaction-aware chemical design. In this seminar, I will motivate and introduce ShEPhERD, a new SE(3)-equivariant diffusion model which jointly diffuses/denoises 3D molecular graphs and explicit representations of their shapes, electrostatic potential surfaces, and directional pharmacophores to/from Gaussian noise. Inspired by traditional approaches to ligand discovery, we compose 3D similarity scoring functions to demonstrate ShEPhERD’s ability to conditionally generate novel molecules in specific 3D conformations that yield desired interaction profiles. We then assess ShEPhERD’s potential for impact via challenging yet exemplary drug design tasks including (in silico) natural product ligand hopping, protein-blind bioactive hit diversification, and bioisosteric fragment merging.