Research map / Probabilistic and scientific modeling / Molecules and proteins
PiFold: Toward effective and efficient protein inverse folding
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PiFold's one-shot featurizer and PiGNN instead of ProteinMPNN | protein inverse folding · CATH 4.2 all | recovery: 45.96 -> 51.66 | Helped |
| Use PiFold instead of ProteinMPNN | protein inverse folding · TS50 | recovery: 54.43 -> 58.72 | Helped |
| Use PiFold instead of ProteinMPNN | protein inverse folding · TS500 | recovery: 58.08 -> 60.42 | Helped |
| Use one-shot decoding rather than autoregressive competitors | long-chain protein design · 100 long chains (average length 1632) | inference speedup: 70 times | Helped |
| Add edge distance features (ablation) | PiFold · CATH 4.2 all | recovery: 49.61 -> 51.66 | Helped |
| Add edge direction features (ablation) | PiFold · CATH 4.2 all | recovery: 50.86 -> 51.66 | Helped |
| Add AttMLP node messaging rather than GCN (ablation) | PiFold · CATH 4.2 all | recovery: 49.65 -> 51.66 | Helped |
| Add AttMLP node messaging rather than GAT (ablation) | PiFold · CATH 4.2 all | recovery: 45.78 -> 51.66 | Helped |
| Add edge updating (ablation) | PiFold · CATH 4.2 all | recovery: 50.00 -> 51.66 | Helped |
| Add global context attention (ablation) | PiFold · CATH 4.2 all | recovery: 51.22 -> 51.66 | Helped |
| Replace autoregressive layers with a one-shot encoder, increasing encoder depth from 3 to 6 layers (ablation) | PiFold · CATH 4.2 all | recovery: 49.30 -> 50.96 | Helped |
| Increase one-shot encoder depth from 8 to 10 layers (ablation) | PiFold · CATH 4.2 all | recovery: 51.22 -> 51.66 | Helped |
Closest papers by meaning
- Bridge-IF: Learning Inverse Protein Folding with Markov Bridges (NeurIPS 2024 · similarity 0.87)
- ProtInvTree: Deliberate Protein Inverse Folding with Reward-guided Tree Search (NeurIPS 2025 · similarity 0.87)
- Protein Inverse Folding From Structure Feedback (NeurIPS 2025 · similarity 0.87)
- SurfPro: Functional Protein Design Based on Continuous Surface (ICML 2024 · similarity 0.86)
- AlphaFold Database Debiasing for Robust Inverse Folding (NeurIPS 2025 · similarity 0.86)
- Fast Uncovering of Protein Sequence Diversity from Structure (ICLR 2025 · 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.