A Free-Wilson tool with automatic core perception and linear regression.
Core perception is currently done using two strategies.
- Topological Maximum Common Substructure (MCS) i.e. comparing unlabelled graphs
- Fragmentation using a SMARTS pattern.
R-group decomposition is done using the respective code in RDKit. Linear regression is based on the Fujita-Ban method defining the compound with the lowest mass weight residues as the reference. Various linear regressors are available.
Create a suitable Python environment with Python >= 3.10 using e.g. conda or mamba. Then install the software from the repository
pip install .Test the tool was installed correctly
fwanalysis -hA web service creating a very simple graphical user interface is available in service/ and can be started with e.g.
uvicorn main:app --host 0.0.0.0Note, you may want to limit the IP address range the server is listening to i.e. replacing 0.0.0.0 with a suitable IP address.
A notebook querying the web service above is available in notebooks/. The Python file(s) in that directory need to be converted e.g.
jupytext -o web_demo.ipynb web_demo.py