This is the code repository for optimization of synthesis plans for multiple molecules. the conda environment tested for running the following code is stored as environment.yml.

The code will require external dependency of the ASKCOS software, which need to be downloaded at git clone https://github.com/Coughy1991/ASKCOS.

After downloading the repository, the folder ASKCOS as well as the subfolder ASKCOS/askcos should be added to PYTHONPATH

The pipeline for running the code include the following steps:

1. create a folder to store the results (e.g. case_1), and input the SMILES of target molecules in the "targets.csv" file
2. run KERAS_BACKEND=theano python tree_builder_for_multiple_target.py case_1 to construct the reaction network for multiple molecule targets
3. run KERAS_BACKEND=theano python predict_conditions_and_evaluate.py case_1 to evaluate the reactions

The GUROBI optimizer is needed for running the optimization. The instruction for installation and obtaining a free academic license can be found at https://www.gurobi.com/

Run the following command for optimizing synthesis plans:

python optimize_synthesis_plan_for_library.py [-h] [-directory [DIRECTORY]] [-separate] [-decomposition] [-weights WEIGHTS WEIGHTS WEIGHTS]

e.g. python optimize_synthesis_plan_for_library.py -directory case_1 -decomposition -weights 10 1 0.1

will run combined synthesis planning for case_1 with the decomposition method used, and weights beging 10,1 and 0.1.