ChemCensor

ChemCensor is a precedent-based framework for evaluating reaction chemical plausibility. It separates the reaction center (what changes) from the functional-group context (what must be tolerated), then checks whether similar patterns are supported by known precedents stored in an SQLite database.

The resulting ChemCensor Score is an integer confidence level from 0 to 5, where higher values indicate stronger precedent support.

---

Installation

python -m pip install -e .

---

Download DB and use with ChemCensor

Below is a complete step-by-step workflow with commands.

1) Clone and install ChemCensor

git clone https://github.com/insilicomedicine/ChemCensor.git
cd ChemCensor
python -m venv .venv
source .venv/bin/activate
python -m pip install -U pip
python -m pip install -e .

2) Download the SQLite database archive

mkdir -p data
hf download insilicomedicine/chemcensor \
  --repo-type dataset \
  --include "ChemCensor-DB-U2-1.0.0.sqlite.zip" \
  --local-dir data

3) Unpack the database

unzip -j -o data/ChemCensor-DB-U2-1.0.0.sqlite.zip -d data/

After unpacking you should have:

data/ChemCensor-DB-U2-1.0.0.sqlite

4) Use the database in Python

from chemcensor import ChemCensor

db_path = "data/ChemCensor-DB-U2-1.0.0.sqlite"
censor = ChemCensor(db_path=db_path)

reaction_smiles = "CCO.CC(=O)O>>CCOC(=O)C"
score = censor.score(reaction_smiles)
print(score)

db_path can point to any local .sqlite file location; using data/ is a convenient project convention.

---

License

ChemCensor is released under a license for independent benchmarking and evaluation purposes only. Use in products, pipelines, automated workflows, or redistribution requires prior written permission from Insilico. See LICENSE for full terms.

---

Citation

If you use ChemCensor in your work, please cite:

@misc{zagribelnyy2026singleanswerenoughrethinking,
      title={When Single Answer Is Not Enough: Rethinking Single-Step Retrosynthesis Benchmarks for LLMs},
      author={Bogdan Zagribelnyy and Ivan Ilin and Maksim Kuznetsov and Nikita Bondarev and Roman Schutski
                and Thomas MacDougall and Rim Shayakhmetov and Zulfat Miftakhutdinov
                and Mikolaj Mizera and Vladimir Aladinskiy and Alex Aliper and Alex Zhavoronkov},
      year={2026},
      eprint={2602.03554},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2602.03554}
}