Reaction fingerprints, atlases and classification. Code complementing our Nature Machine Intelligence publication on "Mapping the space of chemical reactions using attention-based neural networks" (http://rdcu.be/cenmd).
If anyone else want to run rxnfp on a higher Python version (for me its 3.9), the only thing you need to do is install rxnfp with --no-deps:
conda create -n rxnfp python=3.9 -y
conda activate rxnfp
pip install rxnfp==0.1.0 --no-deps
pip install scipy torch tqdm transformers rdkit # enough for gen fingerprint
# some error message when installing, obviously you need other packages to draw reaction atlas
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
rxnfp 0.1.0 requires faerun==0.3.20, which is not installed.
rxnfp 0.1.0 requires matplotlib==3.2.2, which is not installed.
rxnfp 0.1.0 requires scikit-learn==0.23.1, which is not installed.
rxnfp 0.1.0 requires scipy==1.4.1, but you have scipy 1.11.1 which is incompatible.
The main problem here is the compatibility of scipy==1.4.1 with python=3.9. Full error message has been mentioned at intel/dffml/issues/1334.
If anyone else want to run rxnfp on a higher Python version (for me its 3.9), the only thing you need to do is install rxnfp with
--no-deps
:The main problem here is the compatibility of
scipy==1.4.1
withpython=3.9
. Full error message has been mentioned at intel/dffml/issues/1334.