Research map / Probabilistic and scientific modeling / Molecules and proteins
Generative protein structure design
Generates protein structures and sequences using flow matching, equivariant models and AlphaFold.
142 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Molecules and proteins, keyword cluster “proteins · binding · sequence”
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.
24 papers in 2023–24 and 109 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 4.9% to 11.8%.
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
- 2016 · MUST-CNN: A Multilayer Shift-and-Stitch Deep Convolutional Architecture for Sequence-Based Protein Structure Prediction (AAAI 2016 · 39 citations)
- 2018 · Generative modeling for protein structures (NeurIPS 2018 · 116 citations)
- 2019 · Learning Protein Structure with a Differentiable Simulator (ICLR 2019 · 93 citations)
- 2021 · Fold2Seq: A Joint Sequence(1D)-Fold(3D) Embedding-based Generative Model for Protein Design (ICML 2021 · 16 citations)
- 2022 · Proximal Exploration for Model-guided Protein Sequence Design (ICML 2022 · 17 citations)
- 2023 · Diffusion Probabilistic Modeling of Protein Backbones in 3D for the motif-scaffolding problem (ICLR 2023 · 95 citations)
- 2024 · AlphaFold Meets Flow Matching for Generating Protein Ensembles (ICML 2024 · 88 citations)
- 2025 · P(all-atom) Is Unlocking New Path For Protein Design (ICML 2025 · 11 citations)
- 2026 · SimpleFold: Folding Proteins is Simpler than You Think (ICLR 2026 · 3 citations)
Earliest papers here
- MUST-CNN: A Multilayer Shift-and-Stitch Deep Convolutional Architecture for Sequence-Based Protein Structure Prediction (AAAI 2016 · 39 citations)
- Generative modeling for protein structures (NeurIPS 2018 · 116 citations)
- Learning Protein Structure with a Differentiable Simulator (ICLR 2019 · 93 citations)
What has been tried
62 papers here report 664 measured changes (a component added, removed or swapped, compared with a baseline): helped 72%, no clear effect 7%, hurt 18%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| distill the generative process into a single forward pass | ATLAS test ensembles · Root mean W2-distance | 2.61 -> 3.70 | Hurt | AlphaFold Meets Flow Matching for Generating Protein Ensembles (ICML 2024 · 88 citations) |
| Use Riemannian OT paths instead of independent coupling | Protein backbone design · Designability fraction | 0.657 ± 0.042 -> 0.820 ± 0.037 | Helped | SE(3)-Stochastic Flow Matching for Protein Backbone Generation (ICLR 2024 · 15 citations) |
| add distillation | Co-design 1 · designability | 0.42 -> 0.86 | Helped | Generative Flows on Discrete State-Spaces: Enabling Multimodal Flows with Applications to Protein Co-Design (ICML 2024 · 12 citations) |
| use atom14 instead of hybrid14 representation | protein generation · DES-aa | 5% -> 87% | Helped | P(all-atom) Is Unlocking New Path For Protein Design (ICML 2025 · 11 citations) |
| Use coil cutoff r_c = 0.75 instead of r_c = 0.5 during training | Fast-folding proteins · JS-TIC | 0.397 -> 0.373 | Helped | Str2Str: A Score-based Framework for Zero-shot Protein Conformation Sampling (ICLR 2024 · 10 citations) |
| Add ViennaRNA structure context | Archive2 · F1 | 74.38 -> 74.72 | Helped | Bridging Sequence-Structure Alignment in RNA Foundation Models (AAAI 2025 · 7 citations) |
| remove structure MSE loss | Sequence Similarity Split · RMSD | +1.25 RMSD when removed | Helped | RNAFlow: RNA Structure & Sequence Design via Inverse Folding-Based Flow Matching (ICML 2024 · 7 citations) |
| add the sidechain packing objective | CATH 4.2 test set · sequence accuracy | Helped | Sidechain conditioning and modeling for full-atom protein sequence design with FAMPNN (ICML 2025 · 6 citations) | |
| Use non-i.i.d. rather than i.i.d. ensemble sampling | Ubiquitin · correlation with experimental N-H S² | r=0.52 -> r=0.72 | Helped | Inverse problems with experiment-guided AlphaFold (ICML 2025 · 5 citations) |
| add global guidance to local-only guidance | STP10 (inward) · all-atom RMSD | 3.860 -> 1.057 | Helped | Multiscale guidance of protein structure prediction with heterogeneous cryo-EM data (NeurIPS 2025 · 5 citations) |
| add a skip connection from the structure encoder to the decoder | Protein structure generation · Performance | Helped | Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Generation (NeurIPS 2024 · 5 citations) | |
| Add global landscape modeling | CATH 4.2 · recovery rate (%) | 15.51 -> 57.78 | Helped | SurfPro: Functional Protein Design Based on Continuous Surface (ICML 2024 · 4 citations) |
What reviewers will expect
Datasets and baselines named in the 66 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 · 11 of 66 papers
- ATLAS · 6 of 66 papers
- SAbDab · 5 of 66 papers
- EMDB · 3 of 66 papers
- AFDB · 3 of 66 papers
- RAbD · 3 of 66 papers
- CATH 4.2 · 3 of 66 papers
- mdCATH · 3 of 66 papers
Baselines
- ProteinMPNN · 10 of 66 papers
- ESM3 · 6 of 66 papers
- RFdiffusion · 5 of 66 papers
- PiFold · 5 of 66 papers
- MDGen · 5 of 66 papers
- DiffAb · 4 of 66 papers
- Chroma · 4 of 66 papers
- BioEmu · 4 of 66 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 133 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 7 papers
- Massachusetts Institute of Technology · 7 papers
- Peking University · 6 papers
- Mila · 5 papers
- Université de Montréal · 5 papers
- Fudan University · 4 papers
- Tencent · 4 papers
- Zhejiang University · 4 papers
- Westlake University · 4 papers
- McGill University · 4 papers
Use it
Download BibTeX of 31 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
- Protein fitness and variant effect prediction (71 papers)
- Protein interface and docking prediction (72 papers)
- Protein representation learning (40 papers)
- Biological sequence transformers (34 papers)
Elsewhere in the topic
- Prediction-powered and conformal inference (129 papers)
- Statistical analysis of ML practice (122 papers)
- Neuro-symbolic probabilistic layers (120 papers)
- Structure-based 3D molecule generation (107 papers)
- Sparse graphical and tensor models (97 papers)
- Flow matching generative modeling (97 papers)
- Machine learning for physical simulation (96 papers)
- Multimodal trajectory prediction (95 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).