Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

DecompDiff: Diffusion Models with Decomposed Priors for Structure-Based Drug Design

About

Designing 3D ligands within a target binding site is a fundamental task in drug discovery. Existing structured-based drug design methods treat all ligand atoms equally, which ignores different roles of atoms in the ligand for drug design and can be less efficient for exploring the large drug-like molecule space. In this paper, inspired by the convention in pharmaceutical practice, we decompose the ligand molecule into two parts, namely arms and scaffold, and propose a new diffusion model, DecompDiff, with decomposed priors over arms and scaffold. In order to facilitate the decomposed generation and improve the properties of the generated molecules, we incorporate both bond diffusion in the model and additional validity guidance in the sampling phase. Extensive experiments on CrossDocked2020 show that our approach achieves state-of-the-art performance in generating high-affinity molecules while maintaining proper molecular properties and conformational stability, with up to -8.39 Avg. Vina Dock score and 24.5 Success Rate. The code is provided at https://github.com/bytedance/DecompDiff

Jiaqi Guan, Xiangxin Zhou, Yuwei Yang, Yu Bao, Jian Peng, Jianzhu Ma, Qiang Liu, Liang Wang, Quanquan Gu• 2024

Related benchmarks

TaskDatasetResultRank
structure-based drug designMolGenBench In(RM.): Proteins in CrossDock, remove SMILES in CrossDock (train)
Hit Recovery200
10
structure-based drug designMolGenBench Proteins in CrossDock
Pass Rate22.9
10
structure-based drug designMolGenBench Not: Proteins not in CrossDock
Pass Rate20.77
10
structure-based drug designCrossDock 2020 (test)
PB Valid71.7
6
structure-based drug designCrossDocked 2020
Avg Binding Affinity-8.39
4
Showing 5 of 5 rows

Other info

Follow for update