ASYNT-GAN

This is the code for the paper De Novo Drug Design using Artificial Intelligence ASYNT-GAN Computer-assisted de novo design of natural product mimetics offers a viable strategy to reduce synthetic efforts and obtain natural-product-inspired bioactive small molecules but suffers from several limitations. Deep Learning techniques can help address these shortcomings. We propose the generation of synthetic molecule structures that optimizes the binding affinity to a target. To achieve this, we leverage on important advancements in Deep Learning. Our approach generalizes to systems beyond the source system and achieves generation of complete structures that optimize the binding to a target unseen during training. Translating the input sub-systems into the latent space permits the ability to search for similar structures and the sampling from the latent space for generation.

Install TF.GRAPHICS

We are using the Pointnet implementation of TF Graphics we need to clone and install it locally

git clone https://github.com/tensorflow/graphics.git
cd graphics

Windows:

Download and install OpenEXR https://www.lfd.uci.edu/~gohlke/pythonlibs/#openexr

pip install path_to_whl

Install tf graphics

pip install -e . --user

Install Pymol

Pymol is used to generate the 3d representations in .wrl format https://www.lfd.uci.edu/~gohlke/pythonlibs/#pymol-open-source

PREPARE DATA

Create and activate an anaconda env. Install the requirements pip install -r requirements.txt

The file rcsb_pdb_ids_20200628065205.txt holds ids of pdb files related to covid-19 pandemic. The convert_to_wrl_files_pymol.py will fetch all the pdb files and split them into ligands and proteins, centralize them and create .wrl files for each ligand and chain in a particular protein. The convert_to_ply_blender.py will iterate over the files in the data/to_convert folder and create ply files in the converted folder

Train

The ASYNT-GAN.py will sample from converted files train and save the model checkpoints to tf_ckpts_v2 folder and tensorboard logs to logs3d folder. to render tensorboard tensorboard --logdir logs3d The port url and port will be shown in the cmd