Research map / Image generation / Autoregressive and latent generation
Constrained Diffusion for Protein Design with Hard Structural Constraints
Jacob K Christopher, Austin Seamann, Jingyi Cui, Sagar D Khare, Ferdinando Fioretto
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| use ADMM-based constrained diffusion instead of RFDiffusion baselines | RFDiffusion · PDZ domain | Usable percentage (%): 0.0 -> 21.0 | Improved |
| use ADMM-based constrained diffusion instead of best baseline (Recenter + CGD) | RFDiffusion · Molecule encapsulation | Usable percentage (%): 24.2 -> 97.8 | Improved |
| use ADMM-based constrained diffusion instead of Recenter + CGD | RFDiffusion · Molecule encapsulation | Constraint satisfaction (%): 27.4 -> 100.0 | Improved |
| add constraint guidance (CGD) to standard RFDiffusion | RFDiffusion · Molecule encapsulation | Constraint satisfaction (%): 0.0 -> 21.6 | Improved |
| add recentering to CGD | RFDiffusion CGD · Molecule encapsulation | Constraint satisfaction (%): 21.6 -> 27.4 | Improved |
| add constrained diffusion over CGD baseline | CGD · Molecule encapsulation | Radius of gyration (A): 23.9 -> 14.8 | Improved |
| use our method instead of RFDiffusion Standard | RFDiffusion · PDZ domain | Secondary structure (%): 80.8 -> 85.0 | Improved |
| use our method instead of CGD on PDZ | CGD · PDZ domain | Structure realism (%): 38.2 -> 21.0 | Hurt |
| use ADMM with local+global constraints in projected diffusion (ablation) | Projected Diffusion Model · PDZ domain | Constraint satisfaction (%): 100.0 (Global only) -> 0.0 (Failed to Converge) | Hurt |
| replace projection (Global PDM) with our method | Projected Diffusion Model Global · PDZ domain | Structure realism (%): 0.0 -> 97.8 | Improved |
| increase constraint weight schedule from 10/t to 100/t (ablation) | Ours · PDZ 6ubh | Usable percentage (%): 80.0 -> 100.0 | Improved |
| use lambda=inf instead of 100/t (ablation) | Ours · PDZ 6ubh | Usable percentage (%): 100.0 -> 90.0 | Hurt |
Closest papers by meaning
- Protein Design with Guided Discrete Diffusion (NeurIPS 2023 · similarity 0.89)
- Generating Novel, Designable, and Diverse Protein Structures by Equivariantly Diffusing Oriented Residue Clouds (ICML 2023 · similarity 0.88)
- Uncertainty-Aware Multi-Objective Reinforcement Learning-Guided Diffusion Models for 3D De Novo Molecular Design (NeurIPS 2025 · similarity 0.87)
- ProtPainter: Draw or Drag Protein via Topology-guided Diffusion (ICLR 2025 · similarity 0.86)
- 4D Diffusion for Dynamic Protein Structure Prediction with Reference and Motion Guidance (AAAI 2025 · similarity 0.86)
- Floating Anchor Diffusion Model for Multi-motif Scaffolding (ICML 2024 · similarity 0.86)
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.
