Research map / Probabilistic and scientific modeling / Molecules and proteins
DecompDiff: Diffusion Models with Decomposed Priors for Structure-Based Drug Design
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add decomposed arm/scaffold priors (vs. standard Gaussian prior, TargetDiff) (ablation) | atom-only DecompDiff · CrossDocked2020 subset, 16 pockets | Vina Dock Avg.: -7.80 to -9.08 (pocket prior) | Helped |
| use reference-molecule-derived decomposed prior instead of pocket prior (ablation) | atom-only DecompDiff · CrossDocked2020 subset, 16 pockets | Vina Dock Avg.: -7.53 vs -9.08 for pocket prior | Mixed |
| add bond diffusion (atoms and bonds jointly diffused) (ablation) | DecompDiff · CrossDocked2020 | Success Rate Avg.: 11.21% to 15.38% | Helped |
| add bond diffusion (ablation) | DecompDiff · CrossDocked2020 | Vina Dock Avg.: -7.53 to -7.10 | Hurt |
| add bond diffusion (ablation) | DecompDiff · CrossDocked2020 | QED and SA Avg.: QED 0.48 to 0.51; SA 0.61 to 0.66 | Helped |
| add validity guidance (arm-scaffold connection and clash drifts) in sampling (ablation) | atom-only DecompDiff · CrossDocked2020 subset | Complete Avg.: 0.89 to 0.94 | Helped |
| add validity guidance (all drifts) in sampling (ablation) | atom-only DecompDiff · CrossDocked2020 subset | Vina Score Avg.: -4.75 to -6.11 | Helped |
| add only clash drift in sampling (ablation) | atom-only DecompDiff · CrossDocked2020 subset | Vina Score Avg.: -4.75 to -5.79 | Helped |
| add only arm-scaffold connection drift in sampling (ablation) | atom-only DecompDiff · CrossDocked2020 subset | Complete Avg.: 0.89 to 0.94 | Helped |
| full DecompDiff vs TargetDiff on Vina Dock and High Affinity | DecompDiff · CrossDocked2020 | Vina Dock Avg.: -7.80 to -8.39 | Helped |
| full DecompDiff vs Pocket2Mol on Success Rate | DecompDiff · CrossDocked2020 | Success Rate Avg.: 24.5% vs 24.4% | No clear effect |
| full DecompDiff vs Pocket2Mol and TargetDiff on QED/SA | DecompDiff · CrossDocked2020 | QED Avg.: 0.45 vs Pocket2Mol 0.56 | Hurt |
Closest papers by meaning
- DecompOpt: Controllable and Decomposed Diffusion Models for Structure-based Molecular Optimization (ICLR 2024 · similarity 0.92)
- Navigating the Design Space of Equivariant Diffusion-Based Generative Models for De Novo 3D Molecule Generation (ICLR 2024 · similarity 0.90)
- ShEPhERD: Diffusing shape, electrostatics, and pharmacophores for bioisosteric drug design (ICLR 2025 · similarity 0.88)
- SubgDiff: A Subgraph Diffusion Model to Improve Molecular Representation Learning (NeurIPS 2024 · similarity 0.87)
- Equivariant Blurring Diffusion for Hierarchical Molecular Conformer Generation (NeurIPS 2024 · similarity 0.87)
- Geometric-Facilitated Denoising Diffusion Model for 3D Molecule Generation (AAAI 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.