Research map / Probabilistic and scientific modeling / Molecules and proteins
Exploring Synthesizable Chemical Space with Iterative Pathway Refinements
Seul Lee, Karsten Kreis, Srimukh Prasad Veccham, Meng Liu, Danny Reidenbach, Saee Gopal Paliwal, Weili Nie, Arash Vahdat
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 bidirectional (BU+TD) iterative cycle instead of unidirectional sampling (ablation) | ReaSyn · synthesizable reconstruction | reconstruction rate | Improved |
| add Edit Bridge refinement to the BU+TD cycle (ablation) | ReaSyn · synthesizable reconstruction | reconstruction rate | Improved |
| use SMILES-based non-hierarchical pathway representation instead of fingerprint-based hierarchical one (ablation) | bottom-up ReaSyn vs SynFormer · synthesizable reconstruction | reconstruction rate | Improved |
| train one model with random BU and TD directions instead of two separate models (ablation) | ReaSyn · Enamine reconstruction | reconstruction rate (%): 95.2 -> 95.0 | No clear effect |
| use Edit Bridge coupling instead of uniform coupling (ablation) | Edit Flow · 10,000 training pairs | aligned rate (%): 2.59 -> 70.56 | Improved |
| use Edit Bridge coupling instead of uniform coupling (ablation) | Edit Flow · 10,000 training pairs | edit operations per sequence: 142.87 -> 30.01 | Improved |
| use bidirectional single model vs BU-only model on ZINC250k (ablation) | ReaSyn autoregressive model, BU inference · ZINC250k reconstruction | reconstruction rate (%): 51.3 -> 38.8 | Hurt |
| use bidirectional single model vs TD-only model on Enamine (ablation) | ReaSyn autoregressive model, TD inference · Enamine reconstruction | reconstruction rate (%): 85.2 -> 82.5 | Hurt |
| add ReaSyn projection to Graph GA | Graph GA · TDC oracles average | average AUC top-10: 0.633 -> 0.633 | No clear effect |
| use ReaSyn projection instead of SynNet in synthesis-constrained optimization | Graph GA · TDC oracles average | average AUC top-10: 0.545 -> 0.633 | Improved |
| use ReaSyn instead of SynFormer for synthesizable projection | synthesizable hit expansion for JNK3 · JNK3 hit expansion | analog, improve and success rate | Improved |
Closest papers by meaning
- Projecting Molecules into Synthesizable Chemical Spaces (ICML 2024 · similarity 0.90)
- Procedural Synthesis of Synthesizable Molecules (ICLR 2025 · similarity 0.89)
- Generative Flows on Synthetic Pathway for Drug Design (ICLR 2025 · similarity 0.88)
- SynFlowNet: Design of Diverse and Novel Molecules with Synthesis Constraints (ICLR 2025 · similarity 0.88)
- Advancing Retrosynthesis with Retrieval-Augmented Graph Generation (AAAI 2025 · similarity 0.87)
- Retrosynthesis Planning via Worst-path Policy Optimisation in Tree-structured MDPs (NeurIPS 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.
