Implementation of BR-SAScore developed by Prof. Yousung Jung group at Seoul National University (contact: yousung.jung@snu.ac.kr).
- Developer
- OS Requirements
- Python Dependencies
- Installation Guide
- Usage
- Data
- Reproduce the results
- Publication
- License
Shuan Chen (shuan.micc@gmail.com)
This repository has been tested on Windows operating systems.
- Python (version >= 3.6)
- Numpy (version >= 1.16.4)
- rdkit (version >= 2019)
pip install BRSAScore
git clone https://github.com/snu-micc/BR-SAScore.git
cd BR-SAScore
pip install -e .
from BRSAScore import SAScorer
scorer = SAScorer()
smi = 'CC(OC1=CC=CC=C1C(O)=O)=O' # Aspirin
score, contribution = scorer.calculateScore(smi)
The expected output of score
should be
0.8789047205405656
See Demo.ipynb
for the examples of estimating the synthetic accessibility of 18 structurally complex molecules shown in the paper.
The reactions of of the training reactions train_data.csv
are downloaded from the Dropbox link available at the GitHub repo of GLN, originally from Figshare.
The zip file of eMolecules building blocks origin_dict.csv
can be downloaded from the Dropbox link available the GitHub repo of Retro*.
Downlaod the data can put them in ./data/
and rename them to uspto.csv
and emolecules.csv
with column titled reactants>reagents>production
and SMILES
.
Feel free to change the (atom-mapped) reaction data and building blocks data as long as the data is correctly formatted. If the reaction data is not atom-mapped, we recommend to use LocalMapper to prepare high-quality the atom-mappings for your reactions.
Prepare the BRSAScores by
python scripts/prepare_Scores.py -r uspto -b emolecules
from BRSAScore import SAScorer
reaction_from = 'uspto'
buildingblock_from = 'emolecules'
scorer = SAScorer(reaction_from=reaction_from, buildingblock_from=buildingblock_from)
To use the BRSAScore on you own data, replace the reaction_from
and buildingblock_from
@article{chen2024estimating,
title={Estimating the synthetic accessibility of molecules with building block and reaction-aware SAScore},
author={Chen, Shuan and Jung, Yousung},
journal={Journal of Cheminformatics},
year={2024},
publisher={Springer}
}
This project is covered under the MIT Liscence.