Research map / Probabilistic and scientific modeling / Molecules and proteins
Efficient Prediction of Large Protein Complexes via Subunit-Guided Hierarchical Refinement
Chixiang Lu, Yunhua Zhong, Shikang Liang, Xiaojuan Qi, Haibo Jiang
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 |
|---|---|---|---|
| use hierarchical subunit-guided refinement instead of end-to-end | AlphaFold3 · Recent PDB | DockQ success rate Top-1 (%): 70.4 -> 69.0 | No clear effect |
| add multi-chain context refinement over pairwise assembly | CombFold+AF3 · Recent PDB | DockQ success rate Top-1 (%): 43.2 -> 69.0 | Improved |
| use HierAFold on complexes beyond memory limit | AlphaFold3 · Large complexes (>5k tokens) | DockQ success rate Top-1 (%): OOM -> 43.9 | Improved |
| use full diffusion model for coarse prediction instead of distilled (ablation) | HierAFold · Recent PDB | DockQ success rate Oracle (%): 73.1 -> 73.3; time 46 -> 125 min | No clear effect |
| use mini-rollout coarse prediction instead of distilled model (ablation) | HierAFold · Recent PDB | DockQ success rate Oracle (%): 73.1 -> 71.0 | Hurt |
| use residue-level segmentation (tau_split=0) (ablation) | HierAFold · Recent PDB | DockQ success rate Top-1 (%): 69.0 -> 67.8 | Hurt |
| vary tau_split between 5 and 20 (ablation) | HierAFold · Recent PDB | DockQ success rate Top-1 (%): 68.9 to 69.1 | No clear effect |
| use PDE instead of PAE for segmentation (ablation) | HierAFold · Recent PDB | DockQ success rate Top-1 (%): 69.0 -> 68.2 | Hurt |
| combine PAE and distance criteria for subunit selection (vs PAE-only) (ablation) | HierAFold · Recent PDB | DockQ success rate Top-1 (%): 66.9 -> 69.0 | Improved |
| combine PAE and distance criteria for subunit selection (vs distance-only) (ablation) | HierAFold · Recent PDB | DockQ success rate Top-1 (%): 66.3 -> 69.0 | Improved |
| loosen selection thresholds (tau_p=10, tau_d=30) (ablation) | HierAFold · Recent PDB | DockQ success rate Top-1 (%): 69.0 -> 69.3; time 46 -> 50 min | No clear effect |
| use full model rather than distilled model in the fine stage (ablation) | HierAFold · Recent PDB | DockQ success rate Top-1 (%): 64.9 -> 69.0 | Improved |
Closest papers by meaning
- PSBench: a large-scale benchmark for estimating the accuracy of protein complex structural models (NeurIPS 2025 · similarity 0.87)
- Evaluating Representation Learning on the Protein Structure Universe (ICLR 2024 · similarity 0.86)
- Inverse problems with experiment-guided AlphaFold (ICML 2025 · similarity 0.86)
- Protein Structure Tokenization: Benchmarking and New Recipe (ICML 2025 · similarity 0.86)
- PPDiff: Diffusing in Hybrid Sequence-Structure Space for Protein-Protein Complex Design (ICML 2025 · similarity 0.85)
- Atom-level Protein Representation Learning Improves Protein Structure Prediction (NeurIPS 2026 · similarity 0.85)
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.
