Atomically accurate de novo design of antibodies with RFantibody
Peer-reviewed| Поле | Значение |
|---|---|
| Authors | NR |
| Year | 2025/2026 |
| Venue / status | Peer-reviewed journal; Nature (Q1); issue 2026 |
| DOI | 10.1038/s41586-025-09721-5 |
| URL | https://doi.org/10.1038/s41586-025-09721-5 |
| Preprint URL | NR |
| Tools | RFantibody |
Задача и claimed novelty
| Поле | Значение |
|---|---|
| Main biological task | Epitope-conditioned de novo VHH and scFv CDR backbone/sequence design |
| Claimed novelty | NR |
Архитектура
| Поле | Значение |
|---|---|
| Model architecture | Antibody-fine-tuned RFdiffusion → ProteinMPNN CDR sequence → fine-tuned RoseTTAFold2 filter |
| Input representation | NR |
| Encoder / decoder | NR |
| Generative state | CDR backbones, CDR sequences, antibody–target complex proposals |
| Conditioning | Fixed framework, hotspots, CDR lengths, VHH/scFv format |
| Sampling | NR |
Training
| Поле | Значение |
|---|---|
| Objectives / losses | NR |
| Dataset | NR |
| Preprocessing | NR |
Inference
NR
Baselines и retrospective benchmarks
| Поле | Значение |
|---|---|
| Baselines | NR |
| Retrospective | NR |
Prospective experimental validation
| Поле | Значение |
|---|---|
| Prospective | Correct epitope and designed geometry for selected structures; TcdB neutralization after maturation; initial affinities can be modest |
| Tested designs | 95-design low-throughput or 9,000-design display campaigns per target; overall target success 0–2%; multiple confirmed VHH/scFv binders |
| Success criterion | Correct epitope and designed geometry for selected structures; TcdB neutralization after maturation; initial affinities can be modest |
| Main results | Correct epitope and designed geometry for selected structures; TcdB neutralization after maturation; initial affinities can be modest |
Limitations
Framework fixed; target holo conformation/hotspots critical; lower hit rates than general miniprotein binder pipelines; CDR humanness not optimized.
Что статья показывает
- Correct epitope and designed geometry for selected structures; TcdB neutralization after maturation; initial affinities can be modest
Чего статья не показывает
- Predicted confidence, likelihood, pLDDT, docking or developability scores are not experimental affinity.
Размеченные утверждения
| Утверждение | Тип | Локатор |
|---|---|---|
| Atomically accurate de novo design of antibodies with RFantibody | Interpretation | DOI/URL from landscape matrix |
| Publication status: Peer-reviewed journal; Nature (Q1); issue 2026 | Author-reported | landscape matrix, publication status column |
Review queue
- Bibliographic title must be verified against the PDF.
Source-grounded разбор статьи
Недоступные или неполные источники
| Источник | Почему недоступен | Что использовано вместо |
|---|---|---|
| Full Methods/SI Nature | Не выгружены полностью | HTML results narrative |
| bioRxiv 10.1101/2024.03.14.585103 | 429 | Nature |
Вопрос: можно ли epitope-conditioned de novo VHH/scFv получить antibody-fine-tuned RFdiffusion + MPNN + RF2 filtering с экспериментальным подтверждением.
Библиография
[PAPER]Bennett N. et al. / Watson J.L. et al. Atomically accurate de novo design of antibodies with RFdiffusion. Nature. DOI 10.1038/s41586-025-09721-5. Preprint 10.1101/2024.03.14.585103. Code https://github.com/RosettaCommons/RFantibody. Tool RFantibody.
Исследовательский вопрос
[PAPER]De novo CDR design on fixed frameworks against specified epitopes, with atomic-level experimental structures for successes.
Модель
[PAPER]Три модуля: Ab-RFdiffusion, ProteinMPNN, Ab-RF2. AF3 retrospective. См. frame diffusion.
Training / inference
NRAntibody fine-tune details NR. Inference: many trajectories, experimental campaigns 95 or ~9000 designs. Default steps NR.
Baselines / ablations / retrospective
NRСравнение hit rate с miniprotein RFdiffusion pipelines (qualitative, authors). Ablation tables NR this pass.
Prospective evidence
[PAPER]Binding, cryo-EM, modest initial Kd, OrthoRep maturation / TcdB neutralization for selected lineage. Это prospective binding + function, не proof общего превосходства над всеми antibody generators.
Поддержанные утверждения (осторожно)
[PAPER]Некоторые de novo designs bind intended epitope and match designed geometry (selected). Pipeline can yield starting points for experimental maturation. Hit rates low.
Не поддержано
- RF2/AF3 confidence = affinity
- High miniprotein-like hit rates
- CDR humanness / full developability
- Generator alone without filters/assays
Ограничения авторов
[PAPER]Fixed framework; hotspot/target conformation critical; lower hit rates than miniproteins; initial affinities modest.
Интерпретация
[INTERPRETATION]OrthoRep — experimental module. IgG conversion selected scFvs ≠ что все designs manufacturable.
SI и repo
NRSI cryo-EM — не разобран. Repo commit NR (timeout).
Источники
- [1] Nature 10.1038/s41586-025-09721-5 · DOI · источник
- [2] https://github.com/RosettaCommons/RFantibody · источник
Нерешённые научные вопросы
- Methods hyperparameters
- Independent replication