Research map / Image generation / Autoregressive and latent generation
Diffusion for protein and antibody design
Generate protein structures, sequences, and antibodies using latent and language-model diffusion.
48 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2022–2026) · an approach in Autoregressive and latent generation, keyword cluster “protein · molecular · molecule”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
14 papers in 2023–24 and 33 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 3.0% to 3.7%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2022 · Antigen-Specific Antibody Design and Optimization with Diffusion-Based Generative Models for Protein Structures (NeurIPS 2022 · 119 citations)
- 2023 · SE(3) diffusion model with application to protein backbone generation (ICML 2023 · 71 citations)
- 2024 · Proteus: Exploring Protein Structure Generation for Enhanced Designability and Efficiency (ICML 2024 · 18 citations)
- 2025 · Diffusion on Language Model Encodings for Protein Sequence Generation (ICML 2025 · 7 citations)
- 2026 · TD3B: Transition-Directed Discrete Diffusion for Allosteric Binder Generation (ICML 2026 · 0 citations)
Earliest papers here
- Antigen-Specific Antibody Design and Optimization with Diffusion-Based Generative Models for Protein Structures (NeurIPS 2022 · 119 citations)
- SE(3) diffusion model with application to protein backbone generation (ICML 2023 · 71 citations)
- Graph Denoising Diffusion for Inverse Protein Folding (NeurIPS 2023 · 21 citations)
What has been tried
23 papers here report 236 measured changes (a component added, removed or swapped, compared with a baseline): helped 75%, no clear effect 3%, hurt 20%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add single-chain training data derived from oligomers | Unconditional monomer generation · Success rate | 36.9% -> 59.2% | Helped | Proteus: Exploring Protein Structure Generation for Enhanced Designability and Efficiency (ICML 2024 · 18 citations) |
| Add force guidance to ConfDiff-Base | Fast-folding proteins · Val-CA | 0.89/0.91 -> 0.97/0.98 | Helped | Protein Conformation Generation via Force-Guided SE(3) Diffusion Models (ICML 2024 · 12 citations) |
| Add long skip connections | SwissProt · FD-seq | 0.45 -> 0.34 | Helped | Diffusion on Language Model Encodings for Protein Sequence Generation (ICML 2025 · 7 citations) |
| Use CDR-level preference instead of residue-level preference | 1a14 · optimization efficiency | Hurt | Antigen-Specific Antibody Design via Direct Energy-based Preference Optimization (NeurIPS 2024 · 7 citations) | |
| Add RFT and DPO feedback training | CASP few-shot · TM-Score | +8.5 | Helped | MSAGPT: Neural Prompting Protein Structure Prediction via MSA Generative Pre-Training (NeurIPS 2024 · 5 citations) |
| Increase noise scale from γ=0.35 to γ=0.5 | Unconditional backbone generation · diversity (Cluster) | 0.49 (239) -> 0.71 (323) | Helped | Proteina: Scaling Flow-based Protein Structure Generative Models (ICLR 2025 · 4 citations) |
| add dihedral fine-tuning | D2R-MD · lDDT All | 0.719 -> 0.683 | Hurt | Generative Modeling of Full-Atom Protein Conformations using Latent Diffusion on Graph Embeddings (NeurIPS 2025 · 2 citations) |
| Replace TSP sequence construction with random connection | MS Benchmark, 2 motifs, average across lengths · in silico success rate | 73.05 -> 60.67 | Hurt | Floating Anchor Diffusion Model for Multi-motif Scaffolding (ICML 2024 · 2 citations) |
| Add retrieval augmentation | CDR-H3 inverse folding · AAR (%) | Helped | Retrieval Augmented Diffusion Model for Structure-informed Antibody Design and Optimization (ICLR 2025 · 1 citation) | |
| add IPR conditioning to GO | GO function generation · macro F1 | 0.370 -> 0.458 | Helped | CFP-Gen: Combinatorial Functional Protein Generation via Diffusion Language Models (ICML 2025 · 0 citations) |
| Replace density-ratio score correction with standard full fine-tuning of all EGNN parameters | PROTAC-DB test · Valid (%) | 82.5 ± 0.4 -> 94.8 ± 0.4 | Helped | Domain-Adapted Diffusion Model for PROTAC Linker Design Through the Lens of Density Ratio in Chemical Space (ICML 2025 · 0 citations) |
| Increase noise scale η from 0.0 to 1.0 | 1hhp, 1exr, 1hhp assembly, and 4tgl · % designable | 1hhp: 58.2 -> 42.7; 1exr: 53.3 -> 46.4; 1hhp assembly: 20.3 -> 15.8; 4tgl: 11.5 -> 8.2 | Hurt | NMA-tune: Generating Highly Designable and Dynamics Aware Protein Backbones (ICML 2025) |
What reviewers will expect
Datasets and baselines named in the 17 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- PDB · 2 of 17 papers
- CATH · 2 of 17 papers
Baselines
- RFDiffusion · 3 of 17 papers
- DPLM · 2 of 17 papers
- ESM3 · 2 of 17 papers
- ProteinMPNN · 2 of 17 papers
- Chroma · 2 of 17 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 47 papers; from OpenAlex and the papers' affiliations).
- Chinese Academy of Sciences · 4 papers
- University of Cambridge · 4 papers
- University of California, Los Angeles · 3 papers
- ByteDance · 3 papers
- Shanghai Jiao Tong University · 3 papers
- Tsinghua University · 2 papers
- Massachusetts Institute of Technology · 2 papers
- Institute of Computing Technology · 2 papers
- Chan Zuckerberg Biohub - San Francisco · 2 papers
Use it
Download BibTeX of 24 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Equivariant diffusion for 3D molecules (71 papers)
Elsewhere in the topic
- Classical generative image synthesis (147 papers)
- High-resolution diffusion synthesis (140 papers)
- Retrieval and matching with generative features (140 papers)
- Generative latent space tooling (129 papers)
- Generalizable fake image detection (119 papers)
- Latent diffusion with learned tokenizers (114 papers)
- Synthetic training data from generators (111 papers)
- Compositional scene generation (105 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).