Research map / Probabilistic and scientific modeling / Molecules and proteins
RetroXpert: Decompose Retrosynthesis Prediction Like A Chemist
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 |
|---|---|---|---|
| Decompose retrosynthesis into reaction-center identification and reactant generation | RetroXpert · USPTO-50K, reaction type given | Top-1 accuracy: 63.2 -> 70.4 | Helped |
| Decompose retrosynthesis into reaction-center identification and reactant generation | RetroXpert · USPTO-50K, reaction type unknown | Top-1 accuracy: 52.6 -> 65.6 | Helped |
| Use edge-enhanced attention (ablation) | EGAT · USPTO-50K, reaction type given | EGAT bond disconnection accuracy: 85.7 -> 86.0 | Helped |
| Use edge-enhanced attention (ablation) | EGAT · USPTO-50K, reaction type unknown | EGAT bond disconnection accuracy: 64.3 -> 64.9 | Helped |
| Add the auxiliary task predicting the number of disconnection bonds (ablation) | EGAT · USPTO-50K, reaction type given | EGAT bond disconnection accuracy: 74.4 -> 86.0 | Helped |
| Add the auxiliary task predicting the number of disconnection bonds (ablation) | EGAT · USPTO-50K, reaction type unknown | EGAT bond disconnection accuracy: 51.5 -> 64.9 | Helped |
| Add semi-templates (ablation) | EGAT · USPTO-50K, reaction type given | EGAT bond disconnection accuracy: 84.0 -> 86.0 | Helped |
| Add semi-templates (ablation) | EGAT · USPTO-50K, reaction type unknown | EGAT bond disconnection accuracy: 59.9 -> 64.9 | Helped |
| Train RGN with unsuccessful-synthon augmentation (ablation) | RGN · USPTO-50K, reaction type given, ground-truth synthons | Top-1 accuracy: 72.9 -> 73.4 | Helped |
| Train RGN with unsuccessful-synthon augmentation (ablation) | RGN · USPTO-50K, reaction type unknown, ground-truth synthons | Top-1 accuracy: 71.9 -> 70.9 | Hurt |
| Use both training augmentation and predicted-synthon test inputs (ablation) | RetroXpert · USPTO-50K, reaction type given | Top-1 accuracy: 63.5 -> 70.4 | Helped |
| Use both training augmentation and predicted-synthon test inputs (ablation) | RetroXpert · USPTO-50K, reaction type unknown | Top-1 accuracy: 48.1 -> 65.6 | Helped |
Closest papers by meaning
- RetroOOD: Understanding Out-of-Distribution Generalization in Retrosynthesis Prediction (AAAI 2024 · similarity 0.91)
- Advancing Retrosynthesis with Retrieval-Augmented Graph Generation (AAAI 2025 · similarity 0.91)
- LLM-Augmented Chemical Synthesis and Design Decision Programs (ICML 2025 · similarity 0.90)
- GDiffRetro: Retrosynthesis Prediction with Dual Graph Enhanced Molecular Representation and Diffusion Generation (AAAI 2025 · similarity 0.90)
- RetroBridge: Modeling Retrosynthesis with Markov Bridges (ICLR 2024 · similarity 0.90)
- Retro-fallback: retrosynthetic planning in an uncertain world (ICLR 2024 · similarity 0.90)
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.