Research map / Probabilistic and scientific modeling / Molecules and proteins
End-to-End Full-Atom Antibody Design
Xiangzhe Kong, Wenbing Huang, Yang Liu
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| end-to-end full-atom dyMEAN vs pipeline baselines | dyMEAN · RAbD CDR-H3 design | AAR: 43.65% vs 37.38% best baseline (MEAN*) | Improved |
| remove side-chain (full-atom) geometry (ablation) | dyMEAN · RAbD CDR-H3 design | lDDT: 0.8454 -> 0.7999 | Hurt |
| remove side-chain (full-atom) geometry (ablation) | dyMEAN · RAbD CDR-H3 design | DockQ: 0.409 -> 0.343 | Hurt |
| remove shadow/native paratope information sharing (ablation) | dyMEAN · RAbD CDR-H3 design | RMSD and DockQ: RMSD 8.11 -> 9.46; DockQ 0.409 -> 0.356 | Improved |
| remove learnable channel weights w_i w_j^T (ablation) | dyMEAN · RAbD CDR-H3 design | AAR and CAAR: AAR 43.65 -> 39.29; CAAR 28.11 -> 25.87 | Improved |
| remove memory term phi_d (ablation) | dyMEAN · RAbD CDR-H3 design | AAR and CAAR: AAR 43.65 -> 40.01; CAAR 28.11 -> 24.37 | Improved |
| remove memory term phi_d (ablation) | dyMEAN · Complex structure prediction | DockQ: 0.452 -> 0.447 | No clear effect |
| remove external distance loss L_dist (ablation) | dyMEAN · RAbD CDR-H3 design | DockQ: 0.409 -> 0.393 | Improved |
| use 2 iterations (T=2) instead of T=3 (ablation) | dyMEAN · RAbD CDR-H3 design | DockQ: 0.409 -> 0.383 | Improved |
| structural initialization followed by AME (SI+AME vs initialization only) | dyMEAN · RAbD CDR-H3 design | TMscore: 0.5072 -> 0.9726 | Improved |
| designing full antibody instead of CDR-H3 only | dyMEAN · Full antibody design | AAR: 74.96% full antibody vs 43.65% CDR-H3 | Improved |
| gradient search affinity optimization with restricted changed residues | dyMEAN · Affinity optimization SKEMPI V2.0 | ddG: -6.48 (MEAN) vs -7.31 (dyMEAN) | Improved |
Closest papers by meaning
- Cross-Gate MLP with Protein Complex Invariant Embedding Is a One-Shot Antibody Designer (AAAI 2024 · similarity 0.89)
- Relation-Aware Equivariant Graph Networks for Epitope-Unknown Antibody Design and Specificity Optimization (AAAI 2025 · similarity 0.89)
- dyAb: Flow Matching for Flexible Antibody Design with AlphaFold-driven Pre-binding Antigen (AAAI 2025 · similarity 0.89)
- Antibody Design Using a Score-based Diffusion Model Guided by Evolutionary, Physical and Geometric Constraints (ICML 2024 · similarity 0.88)
- Antigen-Specific Antibody Design via Direct Energy-based Preference Optimization (NeurIPS 2024 · similarity 0.88)
- Retrieval Augmented Diffusion Model for Structure-informed Antibody Design and Optimization (ICLR 2025 · similarity 0.87)
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.
