End-to-End Full-Atom Antibody Design
ICML 2023 main track (A*)| Поле | Значение |
|---|---|
| Authors | Xiangzhe Kong, Wenbing Huang, Yang Liu |
| Year | 2023 |
| Venue / status | ICML 2023 main track (A*) |
| DOI | NR |
| URL | https://proceedings.mlr.press/v202/kong23c.html |
| Preprint URL | https://arxiv.org/abs/2302.00203 |
| Tools | dyMEAN |
Задача и claimed novelty
| Поле | Значение |
|---|---|
| Main biological task | End-to-end full-atom antibody CDR-H3 design, docking and affinity-oriented optimization |
| Claimed novelty | NR |
Архитектура
| Поле | Значение |
|---|---|
| Model architecture | E(3)-equivariant multi-channel graph network with iterative sequence/structure updates |
| Input representation | NR |
| Encoder / decoder | NR |
| Generative state | Full-atom antibody sequence/structure and docked complex |
| Conditioning | Epitope, incomplete antibody, CDR design region |
| Sampling | NR |
Training
| Поле | Значение |
|---|---|
| Objectives / losses | NR |
| Dataset | NR |
| Preprocessing | NR |
Inference
NR
Baselines и retrospective benchmarks
| Поле | Значение |
|---|---|
| Baselines | NR |
| Retrospective | Retrospective structure/design metrics only |
Prospective experimental validation
| Поле | Значение |
|---|---|
| Prospective | NR |
| Tested designs | No prospective wet-lab validation in primary paper |
| Success criterion | Retrospective structure/design metrics only |
| Main results | Retrospective structure/design metrics only |
Limitations
Predicted affinity optimization is not measured improvement; training-data scarcity and epitope representation limit generalization.
Что статья показывает
- Retrospective structure/design metrics only
Чего статья не показывает
- Predicted confidence, likelihood, pLDDT, docking or developability scores are not experimental affinity.
Размеченные утверждения
| Утверждение | Тип | Локатор |
|---|---|---|
| End-to-End Full-Atom Antibody Design | Author-reported | DOI/URL from landscape matrix |
| Publication status: ICML 2023 main track (A*) | Author-reported | landscape matrix, publication status column |
Source-grounded разбор статьи
Исследовательский вопрос
[MATRIX]End-to-end full-atom antibody CDR-H3 design, docking and affinity-oriented optimization
Утверждения и поддержка evidence
[MATRIX]Claimed novelty: NR
[INTERPRETATION]Статья не «доказывает» общее преимущество метода вне своего экспериментального дизайна.
Что представленный evidence не поддерживает
- Predicted confidence, likelihood, pLDDT, docking or developability scores are not experimental affinity.
Ограничения, признанные авторами
[MATRIX]Predicted affinity optimization is not measured improvement; training-data scarcity and epitope representation limit generalization.
Дополнительные ограничения интерпретации
NRПока Methods/SI не прочитаны, нельзя утверждать split leakage, точные гиперпараметры и полноту ablation.