11785CMU_Final_Project_Innovative-Drug-like-Molecule-Generation-from-Flow-based-Generative-Model

1. Dataset downloaded scPDB: http://bioinfo-pharma.u-strasbg.fr/scPDB/ CrossDocked2020: (datasets and related packages) https://github.com/gnina

2. Preprocessing It is only for scPDB. Firstly, make sure you have the following packages: Biopython, prody, py3dmol Biopython: $ pip install biopython $ pip install biopython --upgrade prody: $ pip install -U ProDy or $ tar -xzf ProDy-2.0.tar.gz $ cd ProDy-2.0 $ python setup.py build $ sudo python setup.py install py3dmol: pip install py3Dmol In dataset, list of scPDB is in "scPDB.txt" file. Run GetPDB_info.py, you can download all information about scPDB. Then, run class_readPDB.py, you can get final txt file including BindingSites_ID_basedLIgand.txt, and BindingStrength_affinity.txt After downloaded all datasets, run Permutation_test/Protein_assembly_beta.py to rebuild structures based on B-factors and biological aseemblies.

3. Evaluate binding affinity Using BindingStrength_affinity.txt, using prodigy server to predict binding affinity. Downloaded in: https://github.com/haddocking/prodigy

4. Training model and molecule generation: The tutorial for both models have been included in their GitHub Readme file. GraphBP: https://github.com/divelab/GraphBP liGAN: https://github.com/mattragoza/LiGAN The best model we trained is in the "Model" directory. The training loss for each model is in "Results" directory

5. Validity In "Validity_check" file. You need to use python2.7 as your environment and download rdkit package. $ conda create -c conda-forge -n my-rdkit-env rdkit python=2.7 $ conda activate my-rdkit-env $ cd [anaconda folder]/bin $ source activate my-rdkit-env $ conda install numpy matplotlib $ conda install cmake cairo pillow eigen pkg-config $ conda install boost-cpp boost py-boost Then, use Check_Validity.py by using your generated small molecule SMILE string and it will return if the SMILE string is valid.

6. Generated small molecules In the directory "Generated_molecules", there are 3 files including protein structures, binding pockets, and generated small molecules. You can check the structure by using VMD or pymol. To install VMD: https://www.ks.uiuc.edu/Development/Download/download.cgi?PackageName=VMD To install pymol: https://pymol.org/2/