Code to support the FLuID Proof-of-Concept publication.
The FLuID Proof of Concept requires only a standard computer with enough RAM to support the in-memory operations.
FLuID has been run on Linux and Windows 10, but it should run in any python environment that is capable of running a Jupyter Notebook. It is possible to use a GPU to speed up the learning process, but it isn't mandatory.
Windows: 10, and 11 Linux: Centos
The FluID dependencies are found in the yml file.
- rdkit
- tmap
- scikit-learn
- imbalanced-learn
- xgboost
- tqdm
- plotly
- ipywidgets
- faerun
- mhfp
- seaborn
git clone https://github.com/LhasaLimited/FLuID_POC
Switch to the release tag Nature_MI_v1.0.
Install the requirements from the .yml file.
Typical Install time is less than 5 minutes, assuming you have a working python environment.
Open the Jupyter notebook in Jupyter lab (we used v7.0.8 but any suitable version should work) and then execute. A seed has been set to control for reproducability, but the results will not be completely identical due to uncontrollable randomness in learning algorithms. Typical runtime on a normal computer is approximately half an hour.