This folder contains the experimental protocols, supplementary figures, the original data from screening (Data_Summary.csv), the source code for random generation of RHPs, a notebook of BO and example input and output files of the notebook.

The BO is based on the Botorch and Ax which were developed by Meta Platform Inc and are available at https://botorch.org and https://ax.dev.  
The version of the key Python packages are:
- Python: 3.8.8  
- Torch: 1.9.0 + cu111  
- gpytorch: 1.6.0  
- Botorch: 0.5.1  
- Ax: 0.2.2   
Install of these packages would take about 20-30 minutes depending on the Internet condition.
The optimization was performed on a Lenovo Legion R9000k laptop with AMD Ryzen 9 5900HX CPU and NVDIA GeForce RTX 3080 Laptop GPU (16GB). Each run will take about 120 minutes. It will be slightly longer with the increase of the amount of data. 
The notebook will generate three csv files, including: file with volume of organohalides solution for each RHP, file with composition of predicted RHPs and their predicted GPx-like activity and file that control the Janus workstation. Example input ('ExperimentData.csv') and output ('Predict_RHP_activity.csv', 'Predict_volume.csv', 'JanusAddingFile_next_round.csv') files were included.

The random generation of of RHP was run with python 3.8.8 with numpy. It will generate two csv files, one for general reading and another one for control of Janus workstation.